I am in Denmark this week, so there is no new Cohong Lane deep dive today.

Instead, with our first full quarter behind us, here is a look back at what actually landed since I launched in mid-April.

A lot of you have joined recently, so this doubles as a place to start. And if you have been here a while, a mid-year list is a fair chance to catch the one or two you meant to read and never got round to.

The short version first. Cohong Lane is where I invest my own capital in public. Real money, a deliberate China tilt, currently about a fifth of the portfolio, primary work rather than market shorthand, and my mistakes on the record alongside everything else. I am not running a model portfolio or a signal service. I am thinking out loud about how a former CFO tries to compound his own savings without an institution behind him.

Here they are, roughly in the order you ranked them.

Top of the list, and the piece most of you have taken to. It lays out the operating system behind the whole portfolio: two core buckets, a fair-value compass, and the discipline to buy when nobody else wants to.

Next, the most-liked piece of the quarter. Why I hold 4 state lenders for a politically mandated coupon rather than capital gains, and the single signal that would make me sell out.

Then the method piece, and one of the most shared. How I trace a Five-Year Plan mandate down to the handful of companies that actually get paid once the policy lands.

Going by your comments, this next one drew the most discussion: the pace that forces Chinese companies to ship or die, and why I pay to train in that gym.

There was also the week I left the desk and got on a train to see one of these bets on the ground. The trip added what the filings could not.

You restacked this one more than any other. It is where I try to separate the noise you simply have to tolerate from the loss you actually need to fear.

And if you have just arrived and want the why behind all of it: what made me leave a family-office CFO seat to invest my own capital in public, and where the odd little name comes from.

That is the first quarter in the life of Cohong Lane. Whichever one you open, you will have a fair sense of how I think by the end of it.

Back next Saturday with something new. Until then, enjoy the weekend.

Philip

Cohong Lane is a periodical publication made generally available to the public. Nothing here is a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

On 26 June I ran the half-year numbers. Five of the seven indices I follow were green; the two that were red were both in Hong Kong, where I hold individual businesses, not just a fund. My own capital, marked down while the rest of the screen looked cheerful. I did not close the laptop.

Here is the scoreboard, year to date through 26 June.

A word on proportion first. My entire China exposure, including the cash I hold against it, is about 19% of the portfolio, concentrated enough to matter and contained enough to stress-test honestly. The rest sits in the United States, Europe, and across Asia, inside the same income, growth, and venture structure I use everywhere. China is where I do my deepest work, because it rewards it.

Am I being repriced for the businesses I actually own, or dragged by the vehicle they trade inside? Hong Kong’s underperformance is not one drawdown to wait out. It is a sorting problem: which of my Hong Kong holdings are genuinely impaired, and which are sound businesses caught in an unloved index.

I sit on both sides of that line. Hang Seng Tech is a growth position, so a 22% fall is a 22% drawdown, and that was the bargain when I bought it. The banks, utilities, telecoms and insurer are different: those I own one by one. A fund is a basket: a rotten apple inside it is not mine to act on, and I accepted that going in. My own book is a stock-picker’s basket, where a rotten apple is mine to find and mine to cut. So a fall in the fund is the drawdown, full stop; a fall in my businesses is only a mark on what someone will pay today, and the work is deciding, name by name, whether the mark is right.

The easy read is that China has rolled over. That is too simple. Last year the Hang Seng returned almost 28% and Hang Seng Tech 23%. This year both are in reverse, while across the border the STAR 50 and ChiNext indices have run up by more than a quarter over the same months. One country, three completely different tapes. The companies I hold were telling me otherwise.

Retail sales: the obvious read, and the incomplete one

The Chinese consumer is soft and property is still broken. Retail sales fell 0.6% in May from a year earlier, the first annual decline since late 2022; car sales dropped 16% and property investment over the first five months fell another 16%. If you want to sell the idea that China is uninvestable, Hong Kong is the most liquid place to do it.

But uninvestable is too large a word for the evidence. Industrial output still rose around 4.5% in May, and exports grew by double digits over the first four months of the year. The economy is not uniformly weak. It is split: genuinely soft in domestic consumption, sturdy in manufacturing and technology. If the problem were simply a weak China, the mainland boards would not be among the strongest in the world this year. Something more specific is happening inside the Hong Kong vehicle.

My Hong Kong exposure is a bet on particular businesses, not on a retail-sales print: state banks earning and paying out a spread, utilities and telecoms on regulated or near-monopoly cashflows, an insurer rebuilding its model, a platform giant rebuilding its stack around an open-source AI model, carmakers taking European share at higher margins, and a few semiconductor and software names tied to industrial policy. Part of that book lives or dies on the consumer. Most does not. The index draws no such line. I have to.

The vehicle problem: Hang Seng Tech is not the AI trade

The global rally of 2026 has been an AI hardware and capex story, and the Hong Kong tech index is not built to capture it. The names leading the world this year, the chip designers, the server makers, the data-centre landlords, the power providers, are listed in New York, Taipei, or on the mainland, not in Hong Kong.

Standard Chartered’s own wealth research lays it out plainly: Hang Seng Tech is roughly half internet and platform companies, about 15% electric vehicles, and under a fifth upstream AI hardware and semiconductors. It leans on exactly the parts of Chinese tech the retail number hurts, and is thin in the build-out. The AI hardware cycle is real but largely onshore, which is why the STAR 50 and ChiNext have flown while Hang Seng Tech has sunk.

Hong Kong also printed a great deal of new paper. More than 110 listings raised something like HKD 280 billion last year, a more than threefold jump on the year before. That is a supply problem for the old internet complex, not proof that every new line on the exchange is broken.

And the index compiler knows it. On 8 June Hang Seng Tech took in its first two pure-play AI model makers, MiniMax and Zhipu (the company behind the GLM model), dropping Kingdee and Kingsoft to make room. Both listed in Hong Kong only in January and have since run several times over. The gauge built for the last cycle of Chinese tech is being rewired for this one: the index I am marked against in June is not the one I will face by December.

This is the trap, and the June reshuffle proves it. A single quote screen prices very different things as one, and even its own compiler is busy redrawing what counts as Hong Kong tech. I will not read a verdict on my holdings off it.

I should grant the obvious objection: the vehicle story is also a convenient one for someone who would rather not sell, blaming the index compiler instead of asking whether I own the wrong businesses. Fair. What keeps it honest is that each position carries its own kill conditions, written down before the screen turned red.

Three buckets for a drawdown

I have written before, in Volatility Is the Admission Price, Not the Risk, about the four questions I run in any sell-off: has management integrity changed, has the industry structure worsened for good, has regulation impaired the model, has the balance sheet become a real problem. But when a whole sleeve lags at once, I sort the names into three buckets first.

One: the business has changed and the thesis is weaker, which earns a hard look and perhaps a cut. Two: the business is fine and the category, not the company, is being liquidated. Three: the business is fine and now cheaper, where I would add if the underwriting holds. The honest constraint is that I am already fully allocated to every one of these names. Adding means over-allocating beyond the weight I set in calmer weather, a bar I clear only at an exceptional price. The vehicle is cheaper than it was. Not cheap enough yet. So I am not adding yet: price and evidence, not hesitation.

The contradiction, in two names

Take BYD, the sharpest contradiction I own. One metric I track is European registrations, and over the first four months of the year they more than doubled past 100,000 units, enough to overtake Tesla on the continent, and at higher margins than it fights for at home. The domestic price war is real and unfinished, so this is no clean bill of health. But on the metric that matters to my underwriting, export margins rather than the home price war, nothing has yet turned. BYD peaked at around HKD 113 in April and trades below HKD 73 today, tracking the sector down while its own sales climbed.

I can hold it without flinching because I wrote the kill conditions down in calmer weather. The case breaks if Europe shuts the door faster than the factories can localise, through tariffs or local-content rules, or if the price war reaches the export margins. Neither has happened. For now the operating metrics confirm the thesis even as the screen denies it: not proof, and a read that can change with the next print, but when what I track and the quote disagree this loudly, I side with the cars over the screen.

China Merchants Port Holdings makes the same point without the glamour. It is the sort of infrastructure sold indiscriminately whenever China is out of favour. Its first-quarter metrics ran the other way to its shares: revenue up roughly 5% year on year, container throughput up around 4%. Profit was the softer line, and I will not bury that, but for an asset I own over a multi-year horizon the test is whether it is still used and still earning, and rising volume and revenue say it is. One quarter's margin move does not change that. Second bucket: a working business inside an unloved vehicle, sold off with the category.

The rest of the book runs through the same test, more briefly. The platform and EV complex is where earnings pressure is genuinely real, consumption soft, the delivery and ride-hailing fight brutal, the ad recovery patchy, yet the platform giant is rebuilding its cloud around its open-source model, so the next few quarters may read badly while the multi-year case holds.

The banks, insurers, utilities and telecoms are a different animal: liquid, widely foreign-owned, among the first things sold in any China risk-off, which is why they get dragged down alongside Hang Seng Tech while sharing almost none of its fundamentals. The sell-off also lifts their yield. The four state banks I hold as bonds, not stocks sit below book at around 5.5% today, against the 6.8 to 7.4 per cent I earn on cost; the three state telcos I own for the income yield roughly 6% on their Hong Kong lines. Below book on that kind of yield is a bet on balance-sheet survival and dividend policy, not AI capex.

The AI-adjacent software and semiconductor names are the most volatile of the lot, because their thesis leans on a hardware cycle largely not listed here: they can be right about the trend and wrong about the route.

What would make me wrong

Four developments would move a name from out of favour to genuinely impaired: a stalled cloud and AI transformation at the platform giant, with margins compressing and no path back; BYD’s overseas expansion blocked by tariffs or local-content rules moving faster than its factories; a dividend cut at an income name from balance-sheet stress rather than a policy tweak; or a credit event travelling from property into the banks.

None of those has happened. What has happened is duller: the market found better stories to tell elsewhere, and for now it has gone to tell them.

The admission price this time

In Volatility Is the Admission Price, Not the Risk, the admission price was watching real money disappear while every instinct screamed at me to act. This time it is quieter and harder: watching an index I sit inside lag a world that has found a better story to tell. The discipline is to keep that price from rewriting the underwriting.

The real risk is not the drawdown but selling a good business because the wrong index was liquidated first. Sitting with that is harder than it sounds on the page. The screen refreshes every few seconds. The filings are months away. So I hold.

The port operator has already answered its question: volumes and revenue up, so it comes off the open list. The two still undecided are the platform giant’s cloud-margin trajectory and BYD’s first-half European registrations. Those will tell me whether I own the wrong business or merely the wrong index, and I have decided each answer in advance: confirmation at a price this cheap is where I over-allocate, a broken kill condition is where I cut, anything in between is hold. Until the filings land, that distinction is the only one that matters.

As of the date of publication, I hold positions in Alibaba Group Holding (HKEX: 9988), BYD Company Limited (HKEX: 1211), and China Merchants Port Holdings (HKEX: 144), as well as in the Hong Kong-listed banks, utilities, telecoms, insurers, and tech names referenced in the body. Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

The clause everyone skipped on the way to the headline

China’s draft mandatory standard for automated driving systems requires fault-tolerant perception, not a specific sensor, and that distinction changes how the rule hits my portfolio.

I read the primary text riding in a driverless Pony AI robotaxi on the way back from OCT Harbour in Shenzhen, the draft open on my iPad. Four of my holdings depend on it. I had spent the morning walking dealership to dealership, XPeng, Xiaomi, NIO, Hongqi, with a HeyTea somewhere in the middle as a concession to the June heat. By the time I got in the car, I needed to know whether what I had seen on those showroom floors matched what Beijing was actually mandating. Or whether the market had misread the rule, and I had sized for a catalyst that lands in 2027 rather than a mirage that evaporates in 2026. I had a thesis. I was about to find out whether Beijing had just broken it.

It is the Ministry of Industry and Information Technology’s (MIIT) draft national standard for automated driving systems, the mandatory GB standard replacing the voluntary GB/T 44721-2024. The comment window closed on 13 April 2026; the proposed effective date is 1 July 2027.

A word on the driving intelligence ladder the rest of this piece leans on. L2 still needs a human watching the road. L3 lets the car drive itself within defined conditions while a person stays ready to take back control. L4 removes the human inside a defined zone altogether. The draft standard is written for L3 and above, which is where the sensing and compute bill jumps.

The line doing the rounds is “China mandates lidar.” The text says no such thing.

What it actually says is narrower, and for my portfolio far more interesting. Clause B.3.1.5 requires fault tolerance in perception: if one sensor drops out mid-drive, a blinded camera, a lidar that stops returning, the system must still detect the road and bring the car to a safe stop. No single sensor failing may cause harm. Table B.1 sets a minimum forward detection range that climbs with speed, out to 130 metres at 120 km/h, paired with a braking capability of at least 5 m/s². The standard is technology-neutral from the first clause to the last. It never names a sensor. It mandates that the perception system keep working safely when one part of it fails, and it leaves the manufacturer to choose how.

The component reading turns the standard into a lottery ticket on one supplier. What it actually is runs duller and more durable: a performance-and-fault-tolerance bar gated by type approval, a slow, structural tailwind for the companies built to clear it. I own several of those companies. I bought them for other reasons, and this rule now tests every one.

How a safety clause turns into a cashflow

The mechanism is unglamorous, which is why it is mispriced. A car cannot be sold in China without clearing China Compulsory Certification (CCC) type approval. The new standard makes the safety case, and the documentation behind it, the thing you clear. A single-perception-fault tolerance bar, plus a minimal-risk-manoeuvre clause forcing an unaided lane change to a safe stop, is expensive to prove. They reward manufacturers with scale, capital, and regulatory standing, and they raise the sensing and compute content that has to sit on an L3 car to pass.

The standard does not put a number on that content uplift. That it raises content per L3 vehicle is a structural read, mine, built from the clauses, not a figure printed in the text. What I am sure of is the direction, and it reaches the suppliers before any single carmaker’s brand. Who captures it is the question the rest of this piece is really about.

Where each holding sits on the line

I did not buy any of these names for this standard. I bought them earlier, for their own reasons, and the rule is a test of whether those reasons still hold. Four of the six sit in my Ventures sleeve, two in Growth. I own the stack rather than a single badge for a simple reason: I cannot tell you which Chinese carmaker brand wins the autonomy war, and owning the suppliers means I do not have to.

Start with the Ventures sleeve.

Hesai is the global leader in automotive lidar, ranked first in ADAS main lidar at roughly 43 per cent share in 2025, and the first lidar maker to post a full-year profit on a GAAP basis. Its moat is scaled, cost-down, automotive-grade manufacturing with in-house chip design, turning out a unit roughly every 10 seconds, with its ATX sensor heading toward about US$150 in 2026. I own it as the picks-and-shovels supplier that gets paid by sensor content per car, whoever wins the showroom; it is the lidar position I added on the L3 certification path. The 2025 numbers are strong on their own terms: full-year net revenue of RMB3,027.6 million (~US$432.9 million), up 45.8 per cent; a first full-year GAAP net income of RMB435.9 million; and shipments of 1.62 million units, up 222.9 per cent.

Here is the nuance I refuse to gloss. That 2025 revenue is overwhelmingly an ADAS story at a blended selling price near US$260, down from about US$530 the year before as cheaper ADAS units crowded the mix. The L3 content thesis is forward. Hesai is the confirmed lidar supplier for Mercedes-Benz models enabling L3, announced on 19 May 2026. Separately, a week later, it secured a different win: an order exceeding 1 million units from a second, unnamed top European maker across more than 10 of that group’s China joint-venture models. Its multi-lidar design wins with Li Auto, Xiaomi, and Changan start production across 2026 and 2027. The company’s own guidance is 3 to 6 lidars per L3 vehicle, roughly US$500 to US$1,000 of content. That is Hesai’s guidance, not the regulation, and I hold it at arm’s length. I am not buying a 2025 revenue line. I am buying the design-win book that converts as those programmes reach start of production. What the book is worth turns on one number I cannot yet see: how much of the won volume ships at three-to-six-lidar L3 content rather than single-lidar ADAS. So I size for the wins already entering production, not the full book at full content, and I read the start-of-production dates as the conversion happens rather than waiting on a single 2027 print.

Horizon Robotics is the compute leg of the same thesis, one layer up the stack. It is China’s leading domestic ADAS compute and system-on-chip supplier, the Journey series, and the credible domestic alternative to NVIDIA, embedded across China’s top-10 carmakers with switching costs that bite mid-programme. That lock matters more in cars than in most of tech: once a chip is designed into a platform, swapping it mid-cycle is close to restarting the programme, so a design-win is not a single sale but a claim on that model’s revenue for its whole production life. Horizon is sitting on design-wins across more than 400 models, and that book is the thing I am actually buying.

Its share of the domestic-brand basic ADAS market is 47.7 per cent; its share of the broader mid-to-high-end intelligent-driving segment is a more modest 14.4 per cent, just behind Huawei at 15.2 per cent. FY2025 revenue was RMB3,758 million, up 57.7 per cent, with the high-end, NOA-capable hardware reaching 45 per cent of Journey volume, nearly 5 times its 2024 share. There is a second pull underneath all of this that I do not need to forecast to act on: export controls on the most advanced NVIDIA parts, and the wider push to localise compute inside China, move carmakers toward domestic suppliers whichever way the raw-performance race runs. That is a policy-made demand pull, not a call to buy, and it tilts the ground in Horizon’s favour. I own it for the part of the thesis that does not depend on which sensor wins: whether redundancy is met with more lidar or with more cameras and more compute, the compute content rises, and Horizon gets paid. The base is still overwhelmingly L2 and L2-plus, so the L3 uplift is optionality, not current revenue. And the company lost RMB10.469 billion in 2025 with research and development running at about 137 per cent of revenue, while one customer, its Volkswagen joint venture, was 37.6 per cent of revenue in the first half of 2024. I looked at that figure for a while. That is a deliberate land-grab burn, but it is still a burn, and I would be lying if I said the size of it sits easily with me. This is the nature of venture investing.

The car I was sitting in when I read the draft belongs to this position. Pony AI is my L4 position, and it sits on a different clause of the same rulebook. It runs robotaxi and robotruck fleets, its Gen-7 cars built with Toyota and GAC Toyota, and it has reached city-wide unit-economics breakeven in Guangzhou and Shenzhen. The reason I treat L4 as a commercial-fleet business and not a feature that arrives in the car you own is unit economics. A fleet operator can spread one remote safety operator across a growing number of vehicles. In Shenzhen, Pony has pushed that ratio to about one operator per 30 cars, with the wider fleet nearer one to 15 and heading the same way, while an individual owner can never amortise a safety net across anything but a single car. That is why city-wide breakeven on the Gen-7 fleet is the number that matters, and why L4 in a private car stays marketing rather than a business.

I own it as an asymmetric venture bet on L4 commercialisation against genuine failure risk, and it is also a Hesai lidar customer, so it sits downstream of the same sensing leg. The distinction that matters: Pony’s regulatory exposure runs through the L4 framework, Appendix C, not the L3 Appendix B redundancy clauses that drive the Hesai and Horizon content thesis. I will not conflate them. The 2025 revenue was US$90.0 million, up 20 per cent, split robotaxi US$16.6 million, up 128.6 per cent, robotruck US$40.6 million, effectively flat, and licensing US$32.8 million. Q1 2026 revenue jumped to US$34.3 million, up 145 per cent, and the fleet passed 1,700 by late May with a 3,500 target by year-end. It is pre-operating-profit, the one GAAP-positive quarter came from a mark-to-market gain rather than operations, and the robotruck line is going nowhere fast. I size it accordingly.

XPeng is the one I own partly as a hedge against my own thesis. It is a mass-market smart-EV maker, 429,445 deliveries in 2025, up 126 per cent, and its edge is in-house foundational driving models and its own Turing compute rather than sensor content. It has moved to a pure-vision stack and dropped lidar from its new models. Its G7 Ultra runs three Turing chips at about 2,200 TOPS with no lidar at all and is billed as the first L3-compute production car; it holds an L3 road-test permit in Guangzhou from December 2025. The tie-in is uncomfortable and useful, and the discomfort is specific: owning something that argues against my own thesis sits differently in the portfolio from anything else I hold. XPeng adds essentially zero lidar content. If the Chinese market ends up looking like XPeng, my lidar leg is narrower than I would like. Holding it keeps me honest about the counter-case, and it gives me exposure to the software-and-compute path if that is the one that wins.

Then the Growth sleeve.

BYD I have covered at length before, so one line here: God’s Eye reached more than 2.5 million cars by December 2025, but the mass-market tier carrying the bulk of that volume runs no lidar, only the premium Denza and Yangwang tiers do, and the first L3 test licence has not turned into a commercialised L3 product.

Geely is the multi-brand group, Geely, Galaxy, Lynk and Co, Zeekr, that sold 3.02 million vehicles in 2025, up 39 per cent, with NEV sales of 1.69 million, up 90 per cent. I own it for NEV volume and margin recovery, but on this thesis the load-bearing reason is the Zeekr hedge: a mass-market group with a real premium tier sitting on top. Its G-Pilot ladder runs from camera-led mass trims with no lidar up to the ZEEKR 9X, which carries multiple lidars and is described as designed for L3, though L3 there is a future capability, not a certified one. That ladder is why I hold it. If the redundancy bar lifts content at the top of the market, Geely captures it through Zeekr without betting the whole group on L3 arriving on schedule; if the mass market stays camera-led, the volume and margin story still pays. It is the carmaker position that does not depend on which way the sensor-mix question breaks.

Of the six, only Hesai has revenue mechanically tied to autonomy sensor content today, and even Hesai does not break that revenue out by autonomy level. Horizon is the next most exposed, through compute, but mostly at L2-plus. The carmakers are L2-plus volume stories with L3 still aspirational, and Pony is L4. So the L3-content thesis is forward and supplier-side, not a 2025 revenue line.

The risks that matter, and holding through them

That is the base case. Here is what collapses it. The strongest objection to my own thesis is sitting in my own book. The standard is technology-neutral, which means redundancy can be met with cameras, radar, and more compute rather than more lidar. XPeng’s G7 Ultra clears its own L3-compute bar with no lidar at all. The mass market is going the same way: BYD’s mass tier and Geely’s entry trims are camera-led. The data is genuinely split. Aggregate lidar adoption is still rising, around 3.21 million passenger-vehicle installs in China in 2025, up about 110 per cent, with penetration near 18.5 per cent by December, and the premium L3-aspirant tier is stacking multiple lidars. But below roughly RMB200,000 the volume is going camera-led, and falling lidar prices, an L2 unit near US$200 now, cut both ways: they widen the market and they cap the per-car content story at the bottom.

My read is hybrid tiering, not a single winner. Hesai’s lidar leg is validated where it matters commercially, the high-value L3 programmes and the European wins, and capped in breadth by a camera-led mass market. Horizon is the leg that does not depend on which way the sensor mix breaks, though that is not the same as saying its own balance sheet cannot break.

Then timing. The standard is still a draft. The comment window closed on 13 April 2026 and the final text has not been published, so I am treating that April close as the live fact. If the effective date slips past mid-2027 or the redundancy language is watered down, the catalyst dilutes. Since the supplier revenue is already weighted to 2026 and 2027, I am underwriting a wait in any case.

This is where the psychology does the real work, and it is the part I care most about. Four of these are Ventures-sleeve positions, sized at entry so that I can hold them through exactly this kind of uncertainty without ever being forced to act. The catalyst I am waiting on prints in 2027, not this quarter. The temptation with a regulatory thesis is to trade every headline: the draft, the comment close, the next earnings call, the next robotaxi incident. I am deliberately not doing that. I sized these as venture bets precisely so that a slipped date, a soft quarter, or a competitor's bad week, like the temporary regulatory pause that followed a rival robotaxi fleet stalling on 31 March 2026, does not put me in the one position I never want to be in: selling something good at the wrong time because the balance sheet made me. I have been there once, and the balance sheet had nothing to do with it. I owned the position on borrowed conviction, never made the thesis my own, and when it wobbled my nerves sold it for me. The worst feeling in investing is not the money. It is knowing I sold the right thing for the wrong reason, and that the reason was me. Holding power is something I build on purpose, and it runs on conviction I own rather than borrow.

The signal I am actually counting

The rule tells me the bar went up. It does not tell me whether it went up in sensing or in compute. The number that settles it is the average lidar and compute content on the 2027 model-year cars actually approved under the final standard. If those approvals come in at or above the 3-to-6-lidar band, the sensing leg is confirmed and Hesai’s design-win book converts into revenue. If they cluster on camera-led stacks with one lidar or none, I retreat to the compliance-and-capital leg and lean harder on Horizon’s compute, which gets paid either way. That is the test, and it is falsifiable.

Beijing did not mandate lidar. It mandated that no single sensor failing may cause harm, which is a harder bar, and it is the one my suppliers are built to clear. The approval sheets for the 2027 cars will tell me how much lidar and compute the final standard actually forced onto them, and that is the number I will be counting.

As of the date of publication, I hold positions in Hesai (NASDAQ: HSAI), Horizon Robotics (HKEX: 9660), Pony AI (NASDAQ: PONY), XPeng (HKEX: 9868), BYD (HKEX: 1211) and Geely (HKEX: 0175). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

On Friday 6 March this year, I watched around a thousand people queue outside Tencent’s headquarters in Shenzhen. Schoolchildren, retirees, office workers; reports put the ages at eleven to seventy. They were waiting for Tencent engineers to install OpenClaw, an open-source AI agent, on their laptops, free of charge. It was mad. And the question it left me with was how fast Tencent would move now that the queue had formed.

The answer turned out to be four days: Tencent launched QClaw, its own agent built on the same open-source framework, installable in about three minutes. Eight days later QClaw was a mini-program inside WeChat. Within the month Tencent had a full agent suite on the market: QClaw for individuals, Lighthouse for developers, WorkBuddy for enterprises. By late April there was a global beta and a new flagship model.

I cannot name a Western platform company at Tencent’s scale that moves like that. The interesting question is why Chinese ones do.

What is China speed?

Every conversation about Chinese industry arrives, sooner or later, at the same two words: China speed. Almost nobody who says them stops to explain them. So let me.

The Tencent story is the consumer-software version: pavement queue to shipped product family in under seven weeks. The industrial version runs on the same clock, and cars are the clearest place to see it.

A carmaker in Munich or Detroit was built to produce variants of a single model for 7 to 10 years: enough time to stabilise manufacturing, refine a supply chain and amortise the tooling. A Chinese EV maker takes a new model from concept to production in as little as 18 months by industry accounts, and the Wall Street Journal reported in March 2024 that Chinese development timelines run roughly 30% faster than at legacy manufacturers overall. The compression comes from running a different process entirely: development stages in parallel rather than in sequence, virtual testing in place of slower mechanical cycles, smaller and faster suppliers in place of the established names, and a willingness to put a car on sale and improve it in public rather than perfect it in private.

China speed is not the same industrial system executed faster. It is a different operating system, one in which speed is the organising principle and everything else, supplier choice, testing philosophy, launch timing, even pricing, runs downstream of it. I have walked enough mainland dealerships to know the cadence is real. What took me longer to understand is that the cadence is the strategy. The strategy runs on people, and the people running this one have a name for what it costs them: 996.

What does 996 mean?

The number is a schedule: nine in the morning to nine at night, six days a week. Seventy-two hours, against the 40 in a European employment contract. In August 2021 China’s top court and labour ministry published model cases ruling the formal 996 schedule unlawful. The schedule survived the ruling; it just stopped appearing in contracts.

Since I came back to Hong Kong in 2025 I have not stood in a Shenzhen office at nine on a Saturday night counting lit desks. What I can report is my own calendar: I keep more than a 996 schedule myself, and in this part of the world that surprises nobody. The stories do the rest. Xiaomi’s EV engineers, by press accounts, ran what insiders called a 996-plus regime through the SU7 programme, and the AI teams at Alibaba and Tencent are racing agent products out the door on timelines that make the schedule self-enforcing. Nobody has to mandate the hours. The launch calendar does that.

When your competitor’s engineers iterate on Saturday, your five-day development plan is a competitive decision someone else has made for you. Multiply that across every function and every rival, and you get an industry whose clock simply runs faster than yours. That is the gym’s first piece of equipment: time itself, used more aggressively than anywhere else in the world.

What does fierce competition mean?

The word “competitive” gets used loosely about most markets. Here is what it means in China’s EV market in 2026, where the growth end of my China book trains.

First, capacity. Chinese battery production capacity reached roughly 2,500 GWh in 2024 against shipments of about 1,250 GWh, a two-to-one overhang that pushed utilisation toward 50%. A cyclical glut clears on its own. Capacity sustained at twice demand is structural, and it sits underneath every pricing decision in the sector.

Second, pricing through the product itself. When a new model generation arrives every 18 months instead of every 7 years, pricing no longer happens through list-price cuts a CFO can ration. It happens through the launch: each generation lands at a price the previous one cannot defend. That is why the price war does not pause when executives disavow price cuts. The product cadence is the price cut. BYD’s own interim report last August described “industry malpractices such as ‘one-price policy’ and ‘excessive marketing’” weighing on domestic profitability. That is the market leader describing the arena it is winning.

Third, the state had to legislate against losing money. In February 2026 the market regulator issued the Auto Industry Price Behavior Compliance Guide, which explicitly prohibits selling goods below cost.

The hunger behind all of this has a face. Yang Zhilin, founder of Moonshot AI: born 1992, top of his class in computer science at Tsinghua, a Carnegie Mellon PhD finished in under four years, co-author of the Transformer-XL paper, stints at Google Brain and Meta AI. He went home and founded Moonshot in March 2023, naming it for the Pink Floyd album. By November 2025 a company of a few hundred people had shipped K2 Thinking, an open-source model that beat GPT-5 and Claude Sonnet 4.5 on key reasoning and coding tests, on Moonshot's own published numbers. Silicon Valley did not laugh: Cursor was reported to have built its flagship product on Kimi's models, and Elon Musk praised it in public. I was not laughing either: the research desk behind Cohong Lane runs on Kimi Code. Then the cadence kicked in: Agent Swarm in February 2026, K2.5 and K2.6 in April, K2.7 Code on 12 June: four releases in the four months before this piece went out. Asked what he is building toward, Yang talks about marching toward "the endless, unknown snow-capped mountain".

I keep coming back to that sentence. The hunger is for the summit, and the founders in Beijing, Hangzhou, and Shenzhen believe the summit is reachable from where they stand. That is what I am underwriting against when I own anything in this arena: a generation of founders who do not believe in finished.

Why the toughest gym in the world produces global champions

A company that can survive and thrive inside that arena can survive and thrive anywhere in the world. Not automatically, and not unconditionally. The gym builds the muscle: cost discipline under permanent price war, product cadence nobody outside China has matched, manufacturing flexibility, software iteration at consumer-app speed. What the gym does not hand out is the right to compete abroad. That has to be earned separately, through correct integration: local production, local suppliers, local service networks, political acceptability, and economics that survive tariffs.

Strength travels. Membership does not.

In April I re-tested the reason I own BYD in Can BYD Still Become the Future Toyota I Bought?: not a Chinese EV winner, but a globally localised, cycle-resilient, structurally profitable auto-industrial platform at Toyota scale. That sentence is a five-test scorecard, and I wrote it for BYD. The chairman has now named the destination out loud: at BYD’s annual general meeting in Shenzhen on 9 June, with the Hong Kong shares down by a third over the past year, Wang Chuanfu told shareholders that BYD “will truly become the No. 1 automaker globally in terms of scale in five years”. Scale is the one test a chairman can promise from a stage; the other four are why I keep the scorecard. From this piece onward I run it for Geely as well, because the logic does not care which badge is on the bonnet, only whether the gym graduate can belong abroad.

The two are taking different routes up the same scorecard. BYD is building the foreign base itself: trial production at Szeged in Hungary from January 2026, Brazil operational, Turkey scheduled by end-2026, its own roll-on roll-off vessels carrying the cars in between. Geely leans on a foreign industrial base it already owns through Volvo, plus a home machine that held gross margin flat at 16.61% while NEV sales jumped 90% in FY2025, per its March results presentation. Self-built versus acquired: the gym does not care, but the scorecard does, because each route fails differently.

BYD does not file margins by geography, but Dolphin Research’s March 2026 analysis estimates roughly 17.2% gross margin at home against 28.1% overseas in H2 2025, with gross profit per vehicle of about RMB 22,000 domestically and RMB 52,000 abroad. Analyst inference, not filed fact, and I treat it as such. But it rhymes with what the filings verify: overseas revenue grew to 38.65% of BYD’s total in FY2025 while the group margin compressed.

The margin shows up where the competition is not Chinese.

So the question I keep putting to myself, holding both names: if BYD and Geely can survive and thrive in the toughest EV gym in the world, who exactly beats them outside it? The honest answer is nobody on product or cost. What beats them is the part the gym never trained: the markets they fail to integrate into, and the tariffs they fail to localise around. And there is a harder version of that question I owe you: localisation changes the cost base itself. A car built in Szeged is built on Hungarian wages, European suppliers, and none of the financing habits the home arena permits. Whether the trained discipline survives that translation is not something I can assert from my Hong Kong desk. It is exactly what the overseas margin line will measure, and why I will keep reading it.

The membership fee

None of this is free. I own these companies, so I study the bill.

BYD’s 2025 annual report, filed in Hong Kong on 27 March 2026, carries the two numbers that price the gym. Gross margin fell from approximately 19.44% to approximately 17.74%. Operating cash inflow has fallen two years running: roughly RMB 169.7 billion in FY2023, RMB 133.5 billion in FY2024, RMB 59.1 billion in FY2025. Two-thirds of the operating cash engine gone in two years, while revenue kept growing. I read that line at CAPARESH, an Italian restaurant in Shenzhen Bay’s MixC I keep going back to, an iced Americano going down rather more easily than the cash flow statement. It is the line that would have alarmed me, had I not gone looking for the mechanism.

There are two causes compounding, and most coverage stops at the first. The price war is in the number, plainly. But so is a float unwind: since 2018 BYD had paid suppliers through Dilian, its in-house promissory note platform, on terms that, by Bloomberg’s count, averaged 275 days in 2023. When the new payment regulation took effect on 1 June 2025 and BYD joined 16 other automakers in pledging 60-day supplier terms, the structural float that had quietly financed years of expansion reversed all at once. Cash that was always going to leave left simultaneously. And it is not coming back: the float was a working-capital subsidy, and the subsidy has been withdrawn for good. The FY2023 cash baseline therefore belongs to a financing regime that no longer exists.

The second cause sits on the capex line. None of it reads as distress. It reads as a company investing through the price war rather than harvesting from it: Hungary, Brazil, Thailand, intelligent driving, flash charging, and now an announced entry into humanoid robotics, which Stella Li framed in a June 2026 interview with Yicai as a natural extension of the company’s manufacturing and software base. The strongest member pays the highest fee, and keeps adding equipment. The drop is not comforting. It is legible. Those are different things, and the difference is where my sizing lives.

The other names I own in this arena train in the same gym at different weights. In the growth sleeve, Geely’s flat margin through a violent mix shift is the placement test passed, and XPeng repaired its vehicle margin from 5.5% in Q1 2024 to 13.0% in Q4 2025 before easing back into a loss in Q1 2026: real repair, volume-dependent. In the venture sleeve, the fee is existential rather than uncomfortable. Four frontier sets of accounts in one sitting is its own kind of reading: somewhere around the third I stopped asking who wins the technology and started asking who can afford to stay at the table. Horizon Robotics ships a 64.5% headline margin with an adjusted operating loss underneath it; Pony AI runs a 15.7% gross margin on a US$275 million operating loss; Innoscience works out to roughly 7.3%. Three of the four are still paying for their seat. Hesai is the exception: FY2025 results filed 24 March 2026 show a 41.8% gross margin and a first full-year GAAP net income of RMB 435.9 million on revenue up 45.8%. The frontier’s first graduation certificate. The sizing instruction across the sleeve is unchanged: size by margin durability, not by sales momentum.

The full margin and policy work, the scenario weights, the question of whether the state ends the price war, all of that now belongs to the scorecard piece I will publish after the H1 2026 results season, with this version dated so you can mark me wrong later.

Who wins the gym

I have sat across a dinner table from this thesis more than once. Someone, usually someone smart, usually someone who has never read a Chinese filing, says it with the confidence of a conclusion: the toughest arena on earth, whoever survives it wins everywhere. I believe most of it. This piece has been my attempt to show you the machinery behind the slogan: a development clock running 30% faster, a workforce norm that uses time as a weapon, capacity at twice demand, pricing fought through the launch cadence itself, and a public that queues on the pavement to install the next tool. Everyone trains. Nobody rests.

But the gym story is a share-and-survival story, and it skips the question that matters for my capital. The membership fee shows up on the margin line, even for the strongest member, and the prize is collected somewhere else: in Szeged, in Brazil, in the markets where the graduate either belongs or merely exports. The gym makes everyone fitter. The gross margins, not the sales charts, will tell me who actually won it.

As of the date of publication, I hold positions in BYD Company (HKEX: 1211), Geely Automobile Holdings (HKEX: 0175), XPeng (HKEX: 9868), Tencent Holdings (HKEX: 0700), Alibaba Group Holding (HKEX: 9988), Horizon Robotics (HKEX: 9660), Innoscience Technology (HKEX: 2577), Pony AI (NASDAQ: PONY), Hesai (NASDAQ: HSAI), and Meta (NASDAQ: META). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

If I take the lift to the 47th floor of my apartment building, I can see Qianhai across the border. Beijing has built a copy of Hong Kong on reclaimed land in Shenzhen, and I live close enough to look at it before breakfast.

Qianhai is a special cooperation zone on the western edge of the city, built to run Hong Kong’s institutions on the mainland. On the map I am closer to it than I am to Hong Kong Island. Here is the part that surprised me: the better Beijing’s copy works, the more valuable the original becomes, not less.

I went this May as part of a German Chamber of Commerce delegation. After the border I crossed into the zone in a Pony AI robotaxi, no driver, full autonomy. A few minutes later the avenues turned wide, clean and almost empty, glass towers set back behind landscaped green, the air quiet in a way central Shenzhen never is. It does not look like a city that grew. It looks like a city that was specified, then built to the spec.

Qianhai matters because it shows what China can do when policy, capital, infrastructure and institutional design all point at the same target. The skyline is the least interesting part. The wiring is what I came to read: where money, law and state ambition meet, and where my Hong Kong view of China either gets sharper or gets exposed.

My working view after the visit: the institutional layer is real, the wiring to Hong Kong is real, but the experiment is not yet proven. Growth has more than halved from its 2023 peak, the zone still runs on large subsidies, and no published figure yet separates the firms that came for the institutions from those that came for the cheque.

Ground Truth: What I Saw Standing In It

Numbers on a plan are cheap. What changed my read was standing in the place, looking at the parts that do not fit on a balance sheet.

The delegation was not a spectator’s outing. The room read as genuine commercial interest.

The sharpest moment was the aircraft. I sat in a pink eVTOL inside the zone, and I had it filed wrong in my head before I went. This is not a prototype going nowhere. It is a production aircraft you can actually buy, the S-Zero from YIVTOL. Even static, strapped into the seat, the low-altitude economy stopped being an abstraction.

I climbed in on the ground, not in the air. The demo site is not yet approved for passenger flights, so what I watched was a demonstration lift-off, not a ride. I plan to go back to the maker’s production facility for an actual test flight, and I will report on that when I do.

Seated inside the YIVTOL S-Zero.

The YIVTOL S-Zero doing a demonstration lift-off.

The same instinct, frontier technology as marketing, showed up over coffee. In the commercial office of the Shenzhen retail robot maker Anno, where we watched the machines run live, a robot barista ground the beans, pulled the shot and finished the cup with no human hand in the loop, then used its new AI module to print my own face into the foam. It is a gimmick, and it is designed as one: the machine is built to be filmed and shared as much as to pour a flat white. But it is also the cleanest small picture of what the zone is selling, which is automation as spectacle. The low-altitude aircraft and the latte-printing arm are doing the same job, turning “new productive forces” into something a visitor can stand next to and photograph.

The finished cup, my face printed into the foam by Anno's AI module.

The same machine pulling the shot, start to finish with no human hand.

None of this is data. The delegation, the aircraft and the robot barista told me one thing only: which parts of the official story survive contact with the place. That is most of what a visit ever buys you.

By then I could see myself living here one day. I had not gone in expecting that. That pull is the reason I am trying to read it honestly.

What Qianhai Is, and Why Beijing Built It

The geography is the policy. Qianhai is a spit of reclaimed land on the western edge of Shenzhen, directly across the bay from Hong Kong. It is one node in the Greater Bay Area, Beijing’s project to wire eleven cities, Hong Kong and Macao included, into a single economic region of more than 80 million people. Hong Kong sits at one corner of that bay with its convertible currency and its common law; Shenzhen sits across the water with the factories, the engineers and the capital. Qianhai is the seam Beijing chose to stitch the two together.

Fifteen years ago this was mudflats and low-rise. Today it is a finished financial skyline, and the before-and-after carries that part of the story on its own.

The spec is visible underground. Qianhai is the first mainland development zone built from scratch around a centralised cooling system: nine plants, 90 kilometres of underground pipes, ice thermal storage that charges at night and discharges through the day, seawater condensers feeding the loop. Every building connects to it by contract, a condition of the land transfer. From street level you see none of this, no roof-mounted chillers, no cooling towers breaking the skyline. A city built from scratch can bury its own plumbing.

The vision behind it is specific, and Xi Jinping has put his name behind it. He has inspected Qianhai repeatedly since his first visit in 2012 and has personally championed its expansion, and in 2021 the Central Committee and State Council published the Qianhai Plan, growing the zone in a single stroke from 14.92 to roughly 120 square kilometres, close to eightfold. The brief was explicit: make Qianhai the modern-services bridgehead to Hong Kong.

Beijing gave Qianhai a narrower job than another manufacturing park or tech park. Law, finance, accounting and asset management were the point: modern services run to something near Hong Kong’s standard, but on the mainland side of the border. Take the parts of Hong Kong that are hardest to copy, common law, a convertible currency, a century of professional trust, and give them somewhere to operate inside China. None of that is the sort of thing a mainland city can simply build.

So what is Beijing actually testing here? Currency, law, integration, talent: four versions of the same experiment. Can Beijing run Hong Kong’s software inside China’s hardware?

That is the bet: whether the institutional software of an open economy can be imported, run on mainland soil under a different legal system, and kept inside a zone small enough to switch off if it misbehaves. Qianhai is where China is testing whether it can have Hong Kong’s machinery without Hong Kong’s politics.

Is the Growth Real, or Just the Subsidies?

The headline growth is real, but it is no longer explosive. GDP growth has more than halved from its 2023 peak, the zone still runs on large subsidies, and no published figure separates firms that came for the institutions from firms that came for the cheque. Until that changes, the experiment is still unproven. By 2025, Qianhai reported RMB 218.2 billion of modern-services value-added, 212,000 registered enterprises and 183 Fortune 500 firms with a presence in the area. The three series that carry the story, GDP, foreign investment and trade, read better as trends than as single years, so they are charted below.

Qianhai GDP and nominal year-on-year growth, 2021–2025. The left-hand bars are shown in RMB 100 million units; the 2025 figure of 3,318.1 equals RMB 331.81 billion, about 16.6 per cent of Shenzhen's economy. The 26.4 per cent jump in 2023 came as the post-Covid rebound and the new master plan arrived together; 2026 is the first print without those tailwinds, and the first clean read on the zone's own pace.

Actual foreign investment used and its year-on-year change, 2021–2025. After the 2021–2022 peak, FDI fell almost 40 per cent in 2023 and has clawed back only part of the way since.

Qianhai trade and its year-on-year change, 2021–2025. The left-hand bars are shown in RMB 100 million units; total trade reached RMB 757.43 billion in 2025, up 7.2 per cent, after the 43 per cent surge in 2024.

The trend matters more than the level here. Qianhai's GDP grew 26.4 per cent in 2023, when the post-Covid rebound and the new master plan landed in the same year. The 10.3 per cent it printed in 2025 still beats Shenzhen as a whole, but it is well under half that 2023 pace. My read is that the cranes have done their work, and from here the institutions have to compound on their own. The largest single bet you can physically walk into is Tencent's. Penguin Island, the company's new campus, runs to 809,000 square metres of reclaimed land and about RMB 37 billion of investment, with more than 14,000 staff already on site at roughly 30 per cent complete and room for around 80,000 when it is finished. That is a company moving its centre of gravity.

Around it the institutions are accumulating: the Shenzhen-Hong Kong International Financial City now hosts around 522 financial institutions, more than 370 of them Hong Kong or foreign-funded; HSBC has put more than RMB 4 billion into its HYQ Tower, its first wholly-owned office complex in southern China; and the International Legal-services District, open since 2022, has drawn 182 legal-service entities, including seven Guangdong-Hong Kong-Macao joint-venture law firms.

The human numbers tell a narrower story. The zone’s own official boards, as presented to the delegation in May 2026, put its administered population at more than 1.3 million and the professionals it has drawn at 667,000. Against that, it has 3,316 foreign workers and 5,354 foreign residents. That is meaningful by Shenzhen standards, but tiny against Qianhai’s own scale. The zone is filling with domestic professionals much faster than it is becoming an international professional base.

The growth still runs on incentives, and they are large and getting larger: Qianhai’s 2024 measures pay qualifying foreign firms up to RMB 30 million a year, a Shenzhen-wide programme adds up to RMB 50 million, and a regional headquarters earns a one-off RMB 5 million. Competing for capital with a cheque book is fair enough, plenty of places do it. The problem is that no published figure separates the firms that came for the institutions from the firms that came for the cheque, and if the organic share were large and growing, it would be a number worth showing.

Tencent and HSBC prove that large anchors will show up. The harder test is the mid-sized foreign firm that chooses Qianhai for the institutions and nothing else: a law practice, a fund manager, an accountancy. Until that firm shows up without a subsidy attached, the case is still unproven.

Where My Money Already Sits

Here is where I have to correct a sentence I used to tell myself about this place. I used to say I had no money in Qianhai, that I only went to look. That is wrong, and walking the zone is what corrected it. I already own three pieces of how Qianhai works.

The first is Tencent. Penguin Island is the one position in this whole piece that I can physically stand inside. Tencent has genuinely planted itself in Qianhai, and that is not true of every name people attach to the zone.

The second is less obvious and matters more, because it is the plumbing. I own four Chinese state banks, which I hold as synthetic renminbi bonds rather than growth equities, a trade I set out in full in Why I Own China’s State Banks as Bonds, Not Stocks. They belong here because the cross-border renminbi rails Qianhai runs on are theirs. Bank of China is the designated offshore renminbi clearing bank in Hong Kong, and a fast-growing share of cross-border renminbi now clears through China’s own system, CIPS, which moved roughly RMB 180 trillion in 2025. Qianhai does not build its own currency plumbing. It uses theirs. I went to read this experiment directly. I own the rails underneath it.

The third is the most physical of the lot, and I only joined it up standing in the zone. I own China Merchants Port. The Qianhai bonded port zone, the logistics and free-trade spine the whole modern-services pitch is bolted onto, is planned, built and operated by China Merchants Bonded Logistics, a company it wholly owns. The reclaimed land I keep describing trades through a gateway run by a group I already hold. The ground this experiment is built on is in my portfolio, not just in my eyeline.

China Merchants Port’s container terminals in the West Port area of Qianhai. The bonded port zone, the free-trade spine the modern-services pitch is bolted onto, is operated by China Merchants Bonded Logistics, a wholly-owned subsidiary.

I hold no venture position inside the zone itself. The low-altitude economy I sat inside is an area I am actively exploring, written into the 15th Five-Year Plan as part of the “new productive forces” agenda, but exploring is not owning. I will not own a Chinese company unless I can see it in the Plan, and Qianhai is where the next chapters of that Plan get tested before they scale. I set the method out in full in How I Align My China Portfolio with the 15th Five-Year Plan.

So this is a held position, not a trade I am urging on anyone.

How Qianhai Plugs Into Hong Kong

The portfolio point only makes sense because Qianhai does not route around Hong Kong. It plugs into it.

Start with the renminbi. Hong Kong still sits on the world’s deepest pool of offshore renminbi, around RMB 1 trillion, and clears more than 70 per cent of global offshore renminbi payments. In June 2025 the HKMA and the PBoC launched Payment Connect, and in February 2026 the cross-border renminbi facility quota doubled to RMB 200 billion. Qianhai’s cross-border channels run on that Hong Kong infrastructure rather than competing with it.

The legal layer works the same way. The joint-venture law firms exist precisely so that Hong Kong common-law practice can reach onto the mainland. Qianhai can build the offices, the tax incentives and the cooling pipes. It cannot decree a century of legal trust into existence.

This is why I do not read Qianhai as a simple threat to Hong Kong. The obvious read is that Beijing is building a replacement. I do not think that is right. The wiring says something stranger: the copy still runs on the original’s plumbing. Firms choosing Shenzhen’s older finance district over Qianhai are choosing Shenzhen over Shenzhen. For Hong Kong to lose, firms have to choose Qianhai over Hong Kong, and that is not what the wiring is built to do.

The assumption underneath this is simple: Beijing keeps finding it cheaper to borrow Hong Kong’s convertibility and common law than to build its own. What would break that assumption is not a sudden abandonment of Hong Kong, but a quieter path where each successful pilot becomes a step towards replacing the original rather than leaning on it.

For that, I would need the renminbi fully convertible on the mainland and a mainland court trusted the way Hong Kong’s is. Neither is close. Both are earned over decades, not decreed. So I still read the rails as Hong Kong’s, and the signal I am watching is the convertibility pilots themselves.

The next Qianhai year-end enterprise data is the first real read. I will look past the headline registration count to the proxy underneath it: whether operating firms, not just subsidised registrations, are choosing Qianhai for the institutions. Get that, and the copy starts to prove itself. Without it, Qianhai stays an expensive piece of policy theatre.

As of the date of publication, I hold positions in Tencent (HKEX: 0700), China Merchants Port (HKEX: 0144), Bank of China (HKEX: 3988), China Construction Bank (HKEX: 0939), Agricultural Bank of China (HKEX: 1288) and China Merchants Bank (HKEX: 3968). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

I own four Chinese state banks through their Hong Kong-listed shares. They are paying me 6.8 to 7.4 per cent a year on cost while I wait to find out whether the political settlement on payouts holds — or whether the property book finally forces the kind of write-down nobody can absorb on the cohort’s terms. That is the entire trade. Everything else in this piece is the work I did to convince myself the wait is being paid for.

The positions: Bank of China (BOC), China Construction Bank (CCB), Agricultural Bank of China (ABC), and China Merchants Bank (CMB). Three of these, BOC, CCB and ABC, are part of China’s Big Six: the state-controlled giants that also include the Industrial and Commercial Bank of China (ICBC), Bank of Communications, and Postal Savings Bank. CMB is the largest of the joint-stock commercial banks rather than a Big Six member, included here for its wealth-management franchise. For BOC, CCB and ABC I initially paid trailing price-to-earnings multiples between 3.3 and 4.5 times in 2023 and 2024. The order of size today, largest to smallest, is BOC, CCB, CMB, ABC.

I bought them as bonds.

That sentence is the thesis. Not equities priced for growth. Not deep-value cyclicals waiting on a turn. Four lines of state-bank paper that I underwrote the way you underwrite a long-duration sovereign bond — except the coupon is higher and the duration is longer than anything actually available in the renminbi fixed-income market. The credit risk is absorbed by the sovereign, the currency exposure is to a renminbi I think is structurally undervalued over a ten- to twenty-year horizon, and the equity re-rating is optionality the market refused to price two and a half years ago and is only beginning to price now.

The sell signal is the payout ratio. The hold signal is everything else. A yield that falls from 7 per cent on cost to under 5 per cent at today’s price is not the trade breaking, it is the trade working. The chart below shows how rich that coupon has been: trailing yields on the three state banks peaked in double digits around 2021 to 2022 and have compressed to the mid-to-high single digits as prices re-rated, while CMB ran lower throughout and yields 5.8 per cent at the latest year-end.

The pillar that has to break before the trade does

A payout ratio of 30 per cent of net profit is the policy anchor across the cohort, written into the SASAC framework. Actual payouts have clustered tightly around it: across 2019 to 2025 the four banks paid out between 28 and 36 per cent of earnings every year, never once falling toward the 25 per cent line that would break the frame. CCB declared 30 per cent for 2025, its third straight year at that level, and China Merchants Bank’s articles of association set a constitutional floor of “not less than 30 per cent of the net profit attributable to the ordinary shareholders” — language filed at HKEX in May 2025 and unchanged since. The Big Six aggregate payout for 2025 landed near RMB 427 billion, up 1.6 per cent year on year.

That is not the behaviour of an equity board making annual capital-return decisions. It is the behaviour of an instrument with a payout floor written into the articles of association and into the SASAC framework above them. HSBC has now set 50 per cent as the target basis through 2028 on the 2025 annual filing, which is policy guidance, not a constitutional floor. JPMorgan runs in the 25 to 28 per cent range, conservative by choice. The Chinese cohort pays a smaller share of a larger absolute base, and the base is set by state policy. That is what makes the coupon bond-like. The Western banks pay like equities because that is what they are: boards setting the payout afresh each year.

Two falsifiability lines sit on this signal. A cohort-weighted payout below 27 per cent in any annual print is a yellow flag, a hard re-check on my scenario weights. Any single bank below 25 per cent breaks the frame, and the position comes down. Neither has triggered for 2025.

The Reckoning Point

I underwrite the property book more conservatively than the reported numbers invite. Across the cohort I assume the marks are softer than they look, and a synthetic bond is only ever as good as the balance sheet underwriting the coupon.

What lets me hold anyway is that I never paid for clean marks. I started buying in 2023 at a price that was already discounting a stress scenario worse than the one that has since materialised. Sector-wide property exposure has come down from 13.3 per cent of bank assets in 2021 to 10.4 per cent in 2024, and the system-level non-performing loan ratio sat at 1.51 per cent at the end of 2025. The 2025 prints are not clean. Provision coverage declined at five of the six Big Six banks, and retail credit, mortgages and credit cards especially, deteriorated. None of that surprises me. I bought the cohort with that risk in the price. The question I run on the property book is not whether the marks are accurate. It is whether the marks are improving.

There is a second hard fact, and it is the fairest objection to the whole bond frame. When the state absorbs the credit, it does not write a cheque to the H-share holder. It recapitalises the bank, and a recapitalisation can dilute the very equity it is rescuing. The live version played out in 2025. In June the Ministry of Finance subscribed RMB 520 billion of fresh equity into four of the Big Six, funded by special treasury bonds. Bank of China and China Construction Bank, two of the four lines I own, took part; Agricultural Bank and China Merchants did not. Bank of China sold its new shares to the ministry at RMB 5.93. That was only 0.73 times book, dilutive to per-share value, and at the same time a 7.8 per cent premium to the screen price. The H-share register I sit in, which did not take part, fell from 28.4 to 26.0 per cent of the bank. So the protection I am buying is narrower than the phrase first sounds. The bond frame insures me against the equity being wiped to nothing, which is the outcome the historical record actually rules out. It does not insure me against being diluted on the way through a rescue. Those are two different risks, and I only claim the first.

That changes what I watch, not whether I hold. The event that would make me re-underwrite is the other kind: a placement struck at a steep discount to a price that has already collapsed, the dilutive wipe rather than the supportive top-up. That has not happened to the four lines I own, and it sits beside the payout cut as the thing I watch for.

Four layers, one position

The bond frame has four transmission layers.

Currency. Over a ten- to twenty-year horizon I want exposure to the renminbi, not to companies that happen to be denominated in it. State-bank H-shares are the cleanest large-cap renminbi long I can buy from Hong Kong without taking active corporate risk. The dividends are paid out of mainland net profit in renminbi and converted at the time of payment. If the currency works over the horizon, every other layer benefits.

Multiple. The two charts below tell the rest: the three state banks have re-rated off the 2021 to 2022 bottom and still trade at a fraction of book, CMB richer throughout. Western peers trade at multiples of that, HSBC at roughly 1.6 times tangible book on the 2025 annual filing, JPMorgan above 2 times, DBS at 2.1 times on the SGX filing earlier this year. The gap does not need to close to parity for this to work. Almost any compression at all from where I bought is a structural tailwind on top of the coupon.

The reason the gap can close is that there is a large, structural buyer on the other side of the H-share line. The Insurance Asset Management Association of China reported in early 2026 that 63 per cent of mainland insurers planned to expand their Hong Kong-listed holdings, and of nineteen major insurer share-purchase announcements in 2025, thirteen were in Hong Kong-listed names and seven of those were banks. Two policy facts make the bid durable. The risk capital charge on long-duration equity was cut for insurers in late 2025, and the dividend tax on H-shares sits below the A-share equivalent for mainland holders. The mainland insurance balance sheet needs renminbi duration it cannot find in the onshore bond market at a yield that beats its actuarial assumptions, and the state-bank H-share is where it goes to find it.

Credit. I split the credit question in two: whether the coupon is still being earned, and whether the principal is backstopped. Take the coupon first. The full-year 2025 accounts confirm a net interest margin trough flattening, not a recovery. ABC, charted below, is the clearest read in the cohort: the margin has fallen every year this decade, but the decline has narrowed since 2023, and the rest of the Big Six show the same L-shaped trough. I am not waiting for a recovery before I hold; I hold because the bleed has nearly stopped.

What flattens that trough is the deposit funding cost reset, and it is the most important moving part of the trade. A Chinese family that locked savings into a five-year deposit at 5.3 per cent in 2021 can barely get 2 per cent if they renew today, and the banks are the ones who were paying that 5.3 per cent. As those old high-rate deposits come due through 2026 and roll over at today’s far lower rates, the banks’ single largest cost falls. Loan yields are falling too, under the same rate cuts, so this is not profit dropping straight to the bottom line. It is the force holding the floor under the margin. China International Capital Corporation, relayed by China Daily on 28 February 2026, put the scale of it like this:

According to estimates by China International Capital Corp, roughly 75 trillion yuan ($10.94 trillion) in household time deposits will mature this year, including about 67 trillion yuan with maturities of one year or longer. CICC projects that the volume of total household deposits and time deposits of at least one year scheduled to reach maturity in 2026 will increase 12 per cent and 17 per cent, respectively, from 2025 — equivalent to year-on-year rises of 8 trillion yuan and 10 trillion yuan.

Reprice that mountain at today’s rates and the banks’ funding cost falls by something on the order of a percentage point, on my own estimate rather than anyone else’s. They do not have to do anything clever to capture it; they simply stop paying yesterday’s rates. ICBC’s management described the mechanism plainly in September 2025:

Following the LPR cut we also promptly lowered our posted deposit rates, effectively offsetting the impact on our NIM. Through stringent control of debt costs, we ensured the stability of our net interest margin and net interest income.

The maturity wall is not a forecast. It is already moving through the liability side of the balance sheet, quarter by quarter, and it is what tells me the coupon holds.

Now the principal, and the floor under it is a document from 1999. In April that year the Ministry of Finance set up four asset management companies, one per bank, to take RMB 1.4 trillion of non-performing loans off the Big Four. The Bank for International Settlements, in its third Financial Stability Institute paper on East Asian public asset management companies, put the volume in context:

The four AMCs have been specifically mandated to take over approximately RMB 1.4 trillion (USD 170 billion) in distressed assets from the big four banks, equivalent to around 20 per cent of the combined loans outstanding of the big four banks or 18 per cent of China’s GDP in 1998. This mandated NPL transfer, however, represents less than half of the estimated NPLs of the big four banks.

Eighteen per cent of GDP, less than half of what was actually estimated to be bad, and by December 2002 only RMB 300 billion of the 1.4 trillion had been resolved at an average cash recovery rate of around 22 per cent. The carve-out was the political mechanism that made the IPOs of 2005 to 2006 possible. It also established what the Chinese state does when a Big Six balance sheet is in trouble. I am not arguing the 1999 model recurs identically. I am arguing it is the empirical floor under the sovereign-credit half of the thesis: anyone telling you the Chinese state will let a Big Six bank fail in a way that wipes the equity is telling you a story the historical record does not support.

The live worry on top of that floor is LGFV, the local government financing vehicle: off-balance-sheet companies Chinese cities set up to borrow and build infrastructure, and the banking system funds a large share of what they owe. The IMF’s February 2026 Article IV review put the sensitivity like this:

A 5 per cent default rate among LGFVs would equate to roughly a 75 per cent increase in system-wide non-performing loans, though this exposure is heavily concentrated outside the designated global systemically important banks.

The LGFV problem, an estimated RMB 78 trillion of debt, is concentrated at small regional and city lenders, not at the four banks I own. The Big Six are the system, not the risk inside it.

Optionality. This is where Bank of China earns its place as the largest line. I worked through what the plan means for the wider China book in How I Align My China Portfolio with the 15th Five-Year Plan; the piece relevant here is the financial plumbing. The CPC Central Committee’s October 2025 Recommendations for the 15th Five-Year Plan, the upstream document for the Outline adopted at the NPC in March 2026, set the architecture out directly:

Secure and efficient financial infrastructure should be put in place, and the digital Renminbi (RMB) should be steadily developed.

And on the factor markets above that infrastructure:

To promote the efficient allocation of all types of resources and production factors, we should develop a robust unified market for urban and rural construction land, a fully functional capital market, a free-flowing labor market, and a technology market that facilitates efficient industrial application.

The usual proxy here is the renminbi’s share of SWIFT payments, and it is the wrong one. SWIFT carries the messaging, not the settlement, and a fast-growing share of cross-border renminbi now clears through China’s own system, CIPS, which a SWIFT figure never sees. So I watch the pipe that actually carries the flow. Volume through CIPS reached RMB 175 trillion in 2024, up 43 per cent, and roughly RMB 180 trillion in 2025, more than triple its 2020 level. The renminbi settled close to 54 per cent of China’s own cross-border payments by July 2025, double a decade earlier, and in 2023 passed the euro to become the second-largest trade-finance currency. That build-out of the plumbing, not a messaging-share headline, is the second-derivative move the optionality layer was sized for.

The optionality has a domestic leg too, and it shows up first in the fee line. Net fees and commissions turned positive at all six Big Six banks in 2025, ABC and Postal Savings above 16 per cent growth, and system-wide wealth-management balances reached RMB 33.3 trillion by year-end, up 11 per cent. Some of that is cyclical. Some of it is the franchise CMB has spent a decade building finally compounding, dividend coverage from a line that does not lean on the loan book. CMB sits on the middle-class wealth thread: as the plan lifts household income and deepens the capital markets, the strongest retail franchise earns the fees on that migrating money, the deposit wall showing up as fee income rather than a funding-cost saving. ABC sits on the Three Rurals mandate, the rural and agricultural priority the plan keeps funding, which gives it the clearest claim on state-directed credit growth when the cities are slow.

The optionality I bought for free in 2023 has been formalised into policy language in the 15th Five-Year Plan. That is what has changed in two and a half years. The bond coupon was always there. The map now has names on it.

What I actually own

The 3.3 to 4.5 times earnings I paid in 2023 and 2024 is gone, and even after the re-rating these still change hands near half of book value. But the entry multiple was never the point. What I own is not a cheap equity. It is a callable, very long-duration synthetic bond on the renminbi, with embedded equity optionality the bond market does not offer and the equity market did not price.

The coupon is set by political mandate, the credit is held up by the sovereign, and the price is sitting through the optical noise of a banking-sector P&L for a decade or longer. In exchange I get the only structural way I have found to be long the renminbi at a yield that beats a long-only investor's actuarial assumptions. If the equity never re-rates, the downside case is clipping 6.8 to 7.4 per cent for the duration. For a coupon I bought as a coupon, that is a perfectly good outcome. It is the same income trade I run elsewhere in the book, where I am earning about 6 per cent to own China's AI infrastructure while the optionality matures.

Only one thing actually ends it. It is not the property book, which I bought into the price, and it is not a dilutive recapitalisation, which would make me re-underwrite rather than walk away. It is the payout settlement itself, the day Beijing decides these banks should hold capital back rather than pay it out. That is the one line I cannot hedge, and the August 2026 interims are the next read on it. The coupon holds until the mandate that sets it decides otherwise, and not one day before. Until then I run the book as it sits.

I wrote this from Hong Kong, against the HKEX filings of the four banks on the desk, the State Council’s English-language record of the 15th Five-Year Plan Recommendations, and the Bank for International Settlements paper that sits open behind the position.

As of the date of publication, I hold positions in Bank of China (HKEX: 3988), China Construction Bank (HKEX: 0939), Agricultural Bank of China (HKEX: 1288), China Merchants Bank (HKEX: 3968), and DBS Group (SGX: D05). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

I own three Chinese state telcos. They are paying me roughly 6 per cent a year while I wait to find out whether the state’s AI infrastructure build shows up in the earnings — or eats them. That is the entire trade. Everything else in this piece is the work I did to convince myself the wait is worth it.

In the United States, a telco is a pipe. Traffic flows through it, built by others. In China, the operator owns the build site, holds the state procurement, and writes the cheque. That is not a nuance. It is a different business. China Mobile (CM), China Telecom (CT) and China Unicom (CU) are not pipes for AI built elsewhere. They are the licensed infrastructure inside which the state’s computing build is being capitalised.

The last time these three were asked to build national infrastructure at scale, shareholders got less cash for six years. The 2008 to 2013 3G build-out is the analogue. The state has done the naming again. That is the risk I am underwriting against, and it has not gone away. The market has stopped treating these three like dead phone utilities; it has not yet decided whether they belong inside China’s AI+ infrastructure stack. That undecided gap is the trade.

All three are dual-listed: H-shares in Hong Kong, A-shares in Shanghai. Same underlying company, same annual report, same dividend cheque. Different price. A-shares trade at a premium that implies roughly 4 to 5 per cent yield for mainland investors. H-shares imply 5.6 to 6.5 per cent for Hong Kong investors on the same dividend. I hold H-shares for that reason.

I started buying all three in 2023 and 2024, and added again on sell-offs in 2025 and on the 2026 VAT reclassification. The reasoning then was simple: depressed prices, attractive yields, three state-owned enterprise (SOE) backbones inside the national network. A value investor did not need to believe anything else about the future to find that combination interesting. One of my mentors had bought a year earlier, at the 2022 low; his work was done. Mine was not, and I waited.

Why a Chinese telco is not a Western telco

A Chinese state telco is a licensed builder of national infrastructure under a Five-Year Plan mandate, sitting beneath a single shareholder — the State-owned Assets Supervision and Administration Commission (SASAC) — with state procurement directed to domestic providers by both regulation and incentive. A Western telco is a pipe for compute built elsewhere. Same word on the stock screen, different business behind it.

Verizon’s AI infrastructure is a low-latency fibre product for third-party hyperscalers. British Telecom’s (BT) AI thesis is cost-out inside a regulated incumbent with pension obligations. The compute is somebody else’s. Deutsche Telekom (DT) is the closest Western analogue: sovereign-compute pivot, regulated domestic procurement, premium multiple for the digital-sovereignty story. Even DT is buying capacity that others operate.

In China, the 15th FYP names the National Unified Computing Power Network as core infrastructure and writes the operators into the chapter. SASAC owns the controlling stake in all three. The shareholder, the regulator and the procurement counterparty are the same balance sheet.

The 15th FYP screen, applied to the three

China Telecom’s 2025 Annual Results Presentation guides total capex down 9.2 per cent to RMB 73 billion (~US$10.7 billion), and inside that smaller envelope computing infrastructure rises from 25 to 35 per cent of mix while network infrastructure falls from 51 to 41 per cent. Cash reallocated to where the plan says the build will happen, not added on top of where it was already going. That single slide is what Five-Year Plan execution looks like inside a corporate budget.

I have a rule on Chinese equities: I will not own one unless I can see how it rides the current Five-Year Plan. The transmission from chapter language to dividend cheque runs in five steps. The state names a priority. The Ministry of Industry and Information Technology (MIIT) defines the network architecture. The telco moves capex. The segment line shows revenue. The dividend tells me whether shareholders are still being paid while that happens. Break any one and the trade weakens.

The chapter language is unambiguous. Chapter 7 of the 15th FYP mandates the National Unified Computing Power Network: ultra-large-scale clusters, compute-energy coordination, allocation scheduled across regions. The AI Plus action lists model, chip, cloud and application as a coordinated stack. The 2026 Government Work Report names both as priorities. That is step three in the chain: the plan has reached the budget.

Stage three: the growth gets rails

In late April and early May 2026, all three operators stood up within a fortnight and named what they were building: a Token factory. The phrase each used varied. The structure was identical — IaaS compute at the bottom, a scheduling and model layer in the middle, Token billing at the top. Stage three of this thesis is the moment that build went from inferred to publicly committed.

For two years before that, the growth half of this thesis ran on inference. The FYP chapters were clear. The capex mix was shifting. The segment lines were appearing. But no operator had stood in front of an audience and named what they were building.

At the 9th Digital China Summit, China Telecom’s cloud subsidiary Tianyi Cloud launched what it calls a one-stop Token service system, structured around serving AI Token demand at scale. The IaaS layer is already at 91 EFLOPS of self-owned and accessed compute; the stated ambition is a full-stack national infrastructure for Token production. China Mobile followed on 8 May at its annual cloud conference in Suzhou. The chairman announced a Trillion-Token Service Trial Package. The company launched MoMA, a model management platform integrating over 300 AI models with real-time Token billing. China Mobile’s management said explicitly that Token operations cannot yet replace mobile data revenue. I found that last disclosure more useful than the headline number. It tells you the growth runway is still early, and management knows it.

China Unicom announced Token-based computing packages and said it will drive Token volume growth through computing power sales and model access.

The state has provided the framing. In May 2026, Xinhua and CCTV both described the national computing power network as the "computing version of the state grid." Daily Token calls in China exceeded 140 trillion in March, more than 1,000 times the level at the start of 2024. Computing power has been classified as one of China's "six networks," alongside water, power, and logistics. This is the plan being executed in real time.

In 2023, I was mostly being paid to wait. In 2024 and 2025, the growth case was still inferred from policy and capex mix. By 2026, the operators had named the thing they were building. The Token is the unit of account the industry has chosen, and the state has ratified it. Whether this translates to a sustained earnings upgrade is what the August interims will begin to answer. I do not know the answer. That was enough for me to own the sector. It was not enough to own the three names equally.

The sector bet, in order: CM > CT > CU

I size CM largest, CT second, CU smallest. CM is the cohort’s cleanest income position — largest stack, 77 per cent payout, the operator the state would call on first if shareholder cash were needed. CT is the cloud and government-enterprise growth play, where the August 2026 interim results — the first H1 earnings print since the Token factory was named in April — are the first read on whether the Token scheduling layer carries margin. CU is the capital-discipline tail — the thinnest growth disclosure, the most room to disappoint, and the cleanest payout-convergence optionality. I own all three. The differences inside the cohort earn the sizing.

China Mobile is the scale anchor. Backbone operator, 92.5 EFLOPS of computing capacity at FP16, and the largest balance sheet of the three. Payout 77 per cent in 2025, up from 71 in 2023, with 2026 guidance flagged as stable-to-rising at the Capital Markets Day. Its growth play is what the chairman named in Suzhou: the Trillion-Token Service Package and the MoMA platform, sold on the largest stack in the cohort. If the income thesis breaks anywhere in the cohort, it breaks here last. The live alternative is not a revenue miss — it is a payout decision: CM is the operator the state would call on first if the build required shareholder cash, and it has done exactly that before.

China Telecom is the cloud and government-enterprise position. Cloud revenue RMB 120.7 billion (~US$17.8 billion) in 2025, the cohort's first RMB 100 billion milestone. AIDC RMB 34.5 billion (~US$5.1 billion). Security revenue RMB 16.6 billion (~US$2.4 billion) as a standalone line. Its growth play is the Tianyi Cloud Token service named at the Digital China Summit: five layers from compute to application. The useful part is the scheduling layer: guaranteed Token throughput sold with SLA-level commitments, not raw capacity. Whether the margin model follows is the August question. The disclosure says the model exists, not that it is profitable. CT is the cohort name where the story is easiest to believe. That is why I am hardest on it.

China Unicom is the capital-discipline tail. Smallest position. Free cash flow RMB 36.0 billion (~US$5.3 billion) in 2025, up 28.5 per cent on net profit growth of only 1 per cent. I read this as capex restraint rather than revenue momentum; the less flattering read is that CU simply cannot find shovel-ready AI spend at its scale, which would make restraint a tell. The 2027 budget separates the two. Payout sits at 63 per cent against 77 per cent at the other two. That is where the re-rating optionality sits: if CU's payout drifts upward toward 77 per cent, the dividend grows, and the price has to re-rate to hold the yield in line with CM and CT. Its growth play is the Token-based computing packages announced alongside the others, sold for computing power and model access, with smaller scale framed as a flexibility advantage. CU is where I have the least disclosure and therefore the least patience for disappointment.

I am not picking the telco that wins the computing build. I own the listed SOE layer through which part of the build flows: income prices in 2023, growth prices in 2026.

The valuation anchor: CM 11.95×, CT 13.70×, against Deutsche Telekom at 14.02×. The income re-rating from the 2021 trough — CM from 6.85× — is largely in. The next leg of return, if it comes, depends on whether the growth named in April converts to segment-line earnings inside the August interims and the FY2026 Capital Markets Days. CU sits inside the same valuation logic but trades on a free-cash-flow story rather than a clean trailing multiple, which is part of why it is the smallest of the three weightings. The easy income re-rating is mostly gone. The reason I will not pay for it naively is what happened the last time these three were asked to build.

Why 2009 still sits in the middle of my thinking

Across the 2008 to 2013 3G build-out, China Mobile held its payout ratio at 43 per cent for six consecutive years; China Telecom held DPS flat through 2008 and 2009 at an inferred 30 to 35 per cent payout; China Unicom’s 2012 payout sat at 25 to 30 per cent against RMB 99.79 billion (~US$14.7 billion) of capex. Payout suppression of 25 to 40 percentage points lasted six years. The only thing that historically broke the income on these three was the state asking them to build something, and these are the numbers from the last time it happened.

I went to the HKEX primary filings to read the mechanics underneath the headline. CT's DPS sat flat at HK$0.085 across 2008 and 2009 while the C-network integration ran. CU's 2012 DPS was RMB 0.12 (~US$0.02) against that RMB 99.79 billion capex envelope. Board language across all three referenced "investment needs" and "long-term sustainable development". Calm. Administrative. The payout ratio did the talking.

No SASAC or MIIT directive appears in any board disclosure I retrieved. The instruction may have existed. It was not in the public text.

BT cut its dividend in 2020 while funding a national fibre roll-out the UK government had defined as strategic infrastructure. When a company is the national network, shareholder cash becomes the balancing item when the build requires it. China is more direct about that balance.

This is why I will not pay 12× earnings naively.

Why 2026 is not 2009

I do not think 2026 is 2009, for three reasons.

Capex direction. In 2008, the operators were ramping into a build. In 2026, total capex is flat to declining across the cohort. CM cut 2025 capex 8 per cent to RMB 150.9 billion (~US$22.2 billion). CT guided 2026 down 9.2 per cent to RMB 73 billion. CU’s free cash flow rose 28.5 per cent on net profit growth of only 1 per cent, with the capex burden falling. The mix inside the smaller envelope is shifting to AI infrastructure, but the envelope itself is not exploding.

Revenue visibility. The 3G build created the network first and the monetisation later. The AI+ build is already showing up in cloud, AIDC, computing services and security revenue, year by year, in the segment lines. That is a different sequence inside the income statement.

SOE reform direction. The SASAC value-creation framework, often referenced as Document 79, and the later SASAC market-value management push, are distinct documents and I treat them so. The first is a profitability and value-creation campaign for central SOE managers. The second pushes listed central SOEs to manage market value, capital returns and shareholder communication. CM's payout moved from 71 per cent in 2023 to 77 per cent in 2025 inside that framework. The trajectory tracks the framework dates, CT and CU have moved in the same direction over the same window, and management cites shareholder-return language at the Capital Markets Day. Declining capex may be doing more of the work than SASAC. I read the SASAC direction as the most plausible cause; I cannot separate it cleanly from the cash-flow effect on the public disclosure.

I am not treating it as a guarantee. It is the changed incentive set I am underwriting on top of the income floor.

What I am watching for, on a five-year horizon

I am watching three things across three release windows: capex mix at the August 2026 interims and the FY2026 Capital Markets Days; payout trajectory across the same prints; the 2027 budget disclosures, where guided intent either becomes durable or does not. If the capex mix and payout trajectory hold through that window, the cohort thesis is intact. The horizon is the FYP duration, not a single earnings print.

I am a Bayesian investor on long positions. I update on what the data tell me, in proportion to how informative each release actually is. The August interims are one input, not the picture.

On CM, the indicator is the disaggregation of the AI services line. If management starts separating infrastructure services (computing, cloud, security) from bundled content (MIGU, e-commerce), the disclosure quality improves and the underwritten thesis gets cleaner. If the bundle hardens, the underwritten thesis weakens whether or not the headline number grows.

On CT, the indicator is the capex mix continuing to shift toward computing infrastructure inside a flat-to-declining envelope. The 26 per cent to 35 per cent guided shift for 2026 is the signal of intent. The 2027 budget will tell me whether the intent is durable.

On CU, the indicator is payout convergence. If the payout drifts from 63 per cent toward 70 per cent, the yield compresses through price and the re-rating optionality earns out. If the payout stays low while capex obligations rise, the income story weakens before the growth story has had a chance.

Across all three, the indicator that would change my view of the cohort, not the timing, is SASAC’s central-SOE performance assessment. CU’s 2025 Sustainability Report ties performance assessment for central enterprise managers to AI computing power targets. If a future round of criteria encodes computing-power floors as a hard performance measure, the cohort-wide capex discipline is at risk and I would revisit the sector weight before any individual name. That is the single most important unresolved data point in the position.

The trade I am still in

Adding now is harder than buying in 2023. The 2023 buy was easy. Price was low, yield was high, and the underwriting required nothing more than valuation discipline. The 2026 hold is harder. The easy part has played through, and the remaining return needs a state-mandated growth story to deliver where chapter language and filing disclosure currently disagree.

That is the trade I am still in. I bought the position because the work justified it. I keep the 2009 filing on the desk because the last infrastructure cycle did not hurt the thesis first. It hurt the payout.

As of the date of publication, I hold positions in China Mobile (HKEX: 941), China Telecom (HKEX: 728), and China Unicom (HKEX: 762). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

When ServiceNow (NOW) printed Q1 2026, the number was unambiguous. Revenue acceleration. No margin collapse. No evidence that AI-native start-ups were cannibalising the installed base. The print read like the thesis I had been holding for months — and the market sold the stock off seventeen per cent on the news. Analyst downgrades printed alongside the move. The good-till-cancelled (GTC) orders I had set months earlier at the lower end of the fair-value range filled themselves into the drop. The only decision I had to make in that hour was whether to cancel the unfilled tranches. I did not.

The piece of the institutional desk that nobody can buy is not the research. It is the calm. I learned this the slow way: twenty-five years in and around institutional finance, the last six inside a family office, before leaving in 2024 to run the book alone. The question I could not yet answer was whether the discipline had become mine, or whether it still belonged to the desk. The orders that filled themselves while NOW was seventeen per cent lower on a thesis-affirming print are part of the answer. Not the buying — the not-cancelling. Passing that test once does not close the question. The next sell-off may be the one I cancel into.

I am two years in on my full-time investment journey. The research half of the desk is no longer the bottleneck. Five custom AI agents on US$32 a month of reasoning models, sitting on top of a Notion workspace that remembers everything, get me to a depth that used to require several people collecting sources, reconciling filings, and building the dossier. That does not make them a research team. It gives me research-team reach, with one human still responsible for judgement. The other half — the calm that lets the calm-weather version of me do the buying while the in-the-moment version sits on his hands — is not solved. The market is engineered to take that calm away from a retail investor, and the engineering is getting better.

The structural edge of running alone

Running a liquid book alone gives me three structural freedoms an institutional desk does not have: no benchmark to hug, no redemption risk, no career risk. Those are not minor. An institutional desk that tries to run the habits in this piece runs them into a headwind — quarterly performance windows, redemption cycles, the politics of underperforming a peer for two quarters in a row. I have none of those pressures, so the habits compound instead of fighting to survive. That is where the asymmetry comes from, and it is structural — but only when the habits are in place. Without them, the freedoms are just unsupervised drift. The family office taught me the habits. Leaving it taught me they were always the asset, not the headcount. The two scenes below show them under live fire.

Scene one: buying into the SaaSpocalypse

Through late 2025 I held the bull case on three US enterprise-software names — Salesforce (CRM), NOW, and Veeva Systems (VEEV). The market had moved to the opposite view. The narrative was the SaaSpocalypse: large language models would cannibalise the installed base, AI-native start-ups would bypass the incumbents, and the seventy-plus per cent gross margins these businesses had earned for two decades would compress as AI features became table stakes. I disagreed with the market on the business case, not on price. The agentic-workflow case for the incumbents looked stronger to me in an LLM world, not weaker. On price we agreed — these names had been too expensive for me to act on, for months. Valuation was the binding constraint.

What follows is the dossier I built to make sure I was not the one missing something. A logical bear case deserves a proper answer before you bet against it. The answer below is what got me to fair-value ranges I was willing to act on. It is not a full underwriting note for the three positions — I owe that piece separately — and the load-bearing question for this scene is not whether the dossier is right. It is whether the habits held when the print arrived.

The bear case was coherent. That was the problem. A coherent bear case on a name I want to own is the one I have to answer properly, not dismiss or minimise.

The dossier became three tests. I did not trust the bull case until it had cleared all three.

The first test was the data moat. A large language model reasons probabilistically — it works in distributions. Enterprise workflows are not probabilistic. When a customer asks for a refund, the CRM workflow has to be deterministic: same input, same output, every time. The next generation of enterprise systems is precisely that combination — agents reasoning probabilistically, calling deterministic workflows for the actions that have to be exact. CRM, NOW, and VEEV are the systems of record the agentic stack has to call into, and they own the deterministic workflows around that record. Ten years of sales history, customer relationship maps, service interaction logs — context no foundation model can replicate from general training. That is what the sell-off had missed.

Then came the enterprise plumbing: security and complexity at scale. Inside a large enterprise in 2026, point solutions proliferate — one agent for HR queries, another for IT helpdesk, another for sales outreach. Within a year there are fifty independent AI systems from fifty different start-ups running against core data. Every one is a potential prompt-injection vector. Every one requires IT approval, security certification, user retraining, and the political buy-in of the SVP whose department it touches. The CTO ends up insisting on a single system of coordination: a pre-approved, enterprise-grade, security-audited platform the team already knows, one that connects to the systems of record they already run. That is ServiceNow’s AI Control Tower. That is Salesforce’s Agentforce. The bear narrative systematically ignores this constraint because it is modelled by people who have never sat in an IT governance meeting or only care about next quarter’s numbers.

The last test was where the asymmetry actually lived. Today, CRM and NOW compete for IT and software budgets — measured in tens of billions. If AI agents automate work previously done by people — inside sales, tier-one IT support, compliance workflows — the relevant budget shifts. It is no longer the software line item. It is the labour line item. Corporate headcount budgets are measured in trillions. An enterprise deploying agents at multiples of the output-per-dollar of a human employee is not buying software any more. It is replacing payroll. The platform does not automatically take that budget. The incumbents that climb from systems of record into the system of action — the layer that executes tasks rather than stores data — become the toll roads across that flow. Once a customer is hooked on agents that work, the switching cost is the institutional memory embedded in the data model, not the software contract. Existing-book gross margins may still compress as agentic pricing pressure works through. The labour-budget asymmetry is what makes the platform worth owning anyway.

Then the market obliged.

That was the dossier I had on the desk by late 2025. Fair-value ranges, on my numbers, well below market price for months. Sizing decided in calm weather. Then the SaaSpocalypse arrived: a multi-month software capitulation, deepened by Middle East war pressure on risk assets. CRM and VEEV filled close to the peak of the bearish sentiment — bought, in the moment everyone else was selling, by orders the calm-weather version of me had set months earlier. The starter tranches in NOW filled in the same window. That part was uneventful — calm-weather work doing what it is supposed to do.

The eventful part was NOW. The Q1 2026 print landed inside the SaaSpocalypse, not after it. Revenue accelerated. The CFO commentary described exactly the agentic-workflow adoption I had spent six months underwriting. Guidance for the remainder of the year flagged slightly lower margins on the cost of integrating recent acquisitions — acquisitions that, on my reading, strengthened the data moat the case rests on. The market read it as margin pressure. The moat point was ignored. The stock dropped seventeen per cent. Analyst downgrades followed. The GTC orders I had set in calm weather, sitting in tranches at the lower end of my fair-value range, started filling against a tape that was telling me I was wrong.

I did not sit through those hours calmly. A tight pull behind the sternum. The urge to cancel before they embarrassed me further. A low, hot kind of regret — the suspicion that someone else was seeing something I had missed, and the orders were going to keep filling all the way down to a number that would look very stupid in a month.

I named what I was feeling. Then I did the cognitive move the Budapest test had taught me a year earlier: separate the feeling from the state of the business. The print was the business. The price was the print being processed by a market that had decided in advance what the print was going to say. The dossier had not changed. Three questions. Had the data-moat case weakened on this print? No — the CFO commentary and the acquisitions had strengthened it. Had the security-and-complexity case shifted? No. Had the labour-budget asymmetry moved? No. The price had moved. Nothing else had.

I did not cancel the orders. The remaining tranches filled over the next two days. The position is currently above my entry; that is not the point. The point is the decision-shape, which has to be the same whether the stock immediately rewards it or sits underwater for two quarters. I do not know which it will do next, and writing this piece is not predicated on the answer.

What this scene shows is the four habits running into a moment that tests them. Investment policy: the fair-value range was set in calm weather, in writing, weeks before the print. Primary-source research: the bear case got a proper answer, not a dismissal. Calm-weather decision-making: the rule was if a thesis-affirming print arrives and the market sells, let the orders fill inside the range. Execution discipline: the tranches sat in the book and filled themselves while the screen was red. None of this was AI-enabled in the moment. The AI desk wrote the policy. The policy set the orders. The orders built the position. What the moment required of me was not action. It was the discipline not to cancel.

That is the harder of the two registers. The next one is easier — and shows what the system does when I can mostly stay out of its way.

Scene two: scaling China Everbright Water through the 14-to-15 Five-Year Plan bridge

Late 2025. China’s Fourth Plenum hits the wires in the early afternoon, Hong Kong time, on a desk twenty minutes from HKEX. Within the hour the outline of the 15th Five-Year Plan — the country’s economic priorities for the next half-decade — is on my desk. Beautiful China. Water governance named a structural priority. Priority regions made explicit. I am already long China Everbright Water (CEW), a position I initiated in September on work my AI desk ran through August. The Plenum did not trigger the trade. It extended the visibility of the policy the trade was already inside — same direction, longer clock, regions named. That earned the scaling decision.

The interesting moment in this scene is not the Plenum. It is September — buying the position with nothing external yet rewarding me for it.

Under the still-running 14th Five-Year Plan, the research desk worked open-ended. I do not read Chinese well enough to underwrite a policy thesis unaided. The workflow had to be built around that fact. Build a longlist of regulated water-treatment operators that fit the income book. Map the transmission chain from policy mandate to operator cashflow. Work that used to take a team a month. I did it in hours. That speed mattered: it freed August for the policy-to-cashflow work itself, not the data collection. When CEW emerged at the top of the longlist, the agent’s instructions changed: filings only, Plan clauses returned with Chinese alongside English, missing data flagged rather than guessed. The workflow is deliberately suspicious — clause references preserved, any gap treated as a gap rather than filled with fluent nonsense. Conviction does not survive a summary. It survives a source — and on a Chinese-policy thread the source is in Chinese.

By September the numbers were clear. Fair-value range, size band, starter conditions written into the policy in August, while the thesis was still cooling. Looking at the quote screen in September, the in-the-moment version of me wanted to wait. CEW is a smaller operator in a sector I know less well than Chinese banks or insurers; the instinct was to size conservatively and watch another month. Nothing on the tape rewarded the decision — no catalyst, no upgrade, no peer move. Waiting another month felt like the disciplined thing. The August version of me had already overridden that. I bought. Calm-weather decision-making is not the absence of doubt. It is a policy with the authority to override doubt — written when the doubt was not in the room.

When the Plenum outline landed, the same deterministic pipeline ran against the new text. Mandate confirmed and extended; priority regions named; cashflow profile re-tested. The rule fired: if visibility extends, scale to the upper band. I had already decided in August. I executed. By the time the National People’s Congress formally approves the Plan in March 2026, the position has been at its new size for nearly five months. The news prints. I do nothing. Execution discipline, in this case, is the absence of action.

Scene one was holding my nerve while the screen ran red. This was overriding quiet doubt while the screen showed nothing. Same discipline. Different kind of discomfort. The AI gave me depth: policy reading, primary-source checks, and enough Chinese to stop pretending I could underwrite the thesis unaided. The decisions were still mine, twice, and both were made before the screen gave me anything back.

How the desk actually runs

Two reasoning environments. US$32 a month. The tools are not what matters; the architecture is. The architecture turns on one distinction: probabilistic versus deterministic mode. Probabilistic mode is hypothesis work — building the regulated water-treatment operators longlist from a sector universe, ranking by fit to the income book, mapping how a policy mandate plausibly transmits to operator cashflow. The model reasons across distributions; the answer is honestly uncertain and meant to be.

Deterministic mode is the opposite. When CEW emerged at the top of the longlist, the agent’s instructions changed — filings only, Plan clauses with Chinese alongside English, missing data flagged rather than guessed. No improvisation. No fluent inference papering over an absent source. The same model on the wrong mode on the wrong task is worse than useless: it sounds confident about the thing you most need it to admit it does not know.

The full architecture — five agents, the Notion memory layer, how the mode distinction is enforced in practice — is a future piece. What matters here is that the US$32 buys the discipline of mode, not the discipline of the operator.

What the two scenes earned

What the two scenes earned, between them, is the documented version of the four habits twenty-five years in markets installed in me — six of them inside a family office: write the policy in calm weather; read primary sources, not paraphrases; make the hard call when nothing is rewarding it; set the trade up before the moment arrives. The market is engineered against all four — which is why the moment is the wrong place to make them.

The half the US$32 does not buy

The architecture is cheap, deliberately so. When the tool is expensive, I justify using it whether or not it is working. The discipline that decides when to use it is the part that took twenty-five years to build.

What I left the family office with was not a system. It was the knowledge of what good looked like — and a trained instinct for the difference between a market that is genuinely wrong and a portfolio that is genuinely broken. The AI stack is what let me rebuild the depth without rebuilding the headcount.

No Pain to Begin With sets out the structure. Volatility Is the Admission Price sets out the psychology. This is the operating layer that sits beneath both.

But the more I run this desk, the clearer it becomes that the binding constraint is not analytical depth. It never was. The binding constraint is what I described sitting in Budapest a year ago — the gap between what I felt and what I knew, and the willingness to act on the latter while the former was still screaming. That gap is not closed by better models. It is closed, slowly and unevenly, by accumulating enough experience that the subconscious mind starts to recognise the pattern rather than flinch at it.

The test is whether the calm-weather version of me is still in the room when the moment arrives — and whether I can tell, in that moment, the difference between a price moving and a business changing.

The US$32 builds the first. The second is built more slowly, on real tests like the one I described at the top of this piece.

The discipline was always there. The question was always whether I could run it alone. The answer, so far, is yes — but the words so far are doing real work in that sentence.

As of the date of publication, I hold positions in China Everbright Water (HKEX: 1857), ServiceNow (NYSE: NOW), Salesforce (NYSE: CRM), and Veeva Systems (NYSE: VEEV). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

In March 2026, the National People’s Congress (NPC) adopted the Outline of the 15th Five-Year Plan (FYP). The Outline is the screen I will run every name in my China book through for the next five years.

By the time the vote landed, I had been reading its signals for half a year: the Party plenum communiqué in October 2025; the Central Committee’s Recommendations on the 15th FYP in November; MIIT’s AI Plus and Digital Infrastructure action plan in January. The published Outline did not come out of nowhere. It came out of a paper trail that anyone with an indecent quantity of pu-erh tea could read in advance, which is what I did.

The financial press treated the document as a slogan inventory the morning after the vote: AI Plus, Beautiful China, RMB internationalisation, Belt and Road, technological self-reliance. That is not how I read it.

My one-line rule on Chinese equities

I have invested my own money in Chinese equities since 2018. Eight years in, one rule does most of the work: I will not own a Chinese company, public or private, unless I can see how it rides the current Five-Year Plan.

That is not because the FYP is destiny. The 14th FYP correctly called the property deleveraging in writing, and the Hang Seng Mainland Properties Index still fell roughly 75% from its 2021 peak. Plan-right does not equal stock-right. The cautions are part of the rule, not an afterthought, and I will come back to them. The reason the rule still does the work is narrower.

In China, the FYP is the closest thing to an operating manual that the state ever publishes about itself. The chapters tell you what gets funded, what gets licensed, what gets procured, and what the cadres running provincial economies will be evaluated against in five years’ time. If a company I am thinking of owning is not in the chapters, the structural tailwind I am trying to underwrite is not there. If it is in the chapters, I have at least bought myself a fair fight.

The confirmation arrives in chairman letters. I read the FYP first for its house style: the phrases that are too bureaucratic to be accidental, and too specific to be filler. Then I look for the echo. Annual results, capital-markets days, Party committee language, segment descriptions, sometimes a chairman letter doing its best impersonation of a State Council notice. When the same wording appears downstream, I mark it.

That is the check. When China Mobile uses the FYP’s exact framing on AI Plus or computing-power networks in its annual results, I am not reading marketing. I am reading translation from state policy to corporate strategy, in print, and free of charge. The phrase is the company telling its regulator, its shareholders, and the cadre evaluating its Party committee that it has read the manual and is acting on it. That tell carries more signal than almost anything else an outside investor gets in this market.

Why these things bind, and when they don’t

Five-year plans bind, when they bind, because the people executing them get graded on them. The 2019 Regulations on the Evaluation of the Work of Party and Government Leading Cadres are still in force. The mechanics are unromantic. A Party committee evaluates a cadre once a year and at the end of each term. Indicators include implementation of Central decisions and “high-quality development” outcomes, which is the technical phrase for “the things written in the FYP.” A “basic competence” rating triggers a formal admonishment and a deadline to improve. A “poor” rating costs the cadre a rank. Sustained underperformance can cost the principal responsible person the job. That is why the manifesto is a manual. “The state has decided X” in China is not a Western press statement; it is a behavioural instruction tied to a cadre’s evaluation, promotion prospects, and, eventually, job security. The Outline itself tags only some indicators as binding and most as expectative. That is the system telling you which lines a cadre will actually be punished against. The mechanism is real but uneven. Provincial governments game the binding numbers, particularly the environmental ones. Cadre rotation can break term-end accountability before it tightens. Chapters the Party prefers not to enforce get under-resourced rather than rejected. The screen relies on it where it bites.

The horizon is the other thing that makes the FYP serious. The 15th nests inside the 2035 long-range objectives, which were adopted with the 14th FYP in 2021 and which speak in the directional language of “by 2035, China will basically achieve socialist modernization.” Two five-year plans, one strategic horizon. The 16th FYP will be the third leg.

I run my own portfolio with an Income and Growth structure and a five-year minimum horizon on positions, and the FYP-2035 stack tells me the policy direction I am underwriting will not be unpicked at the next reshuffle. That is unusual in most markets, useful in investing, and worth more than people give it credit for.

What the 15th FYP told me to look at, and what it did to my book

Roughly twenty-five things in the 15th FYP would interest an investor. Four have already changed something on my screen. Here they are, in the order they hit my book.

New productive forces, AI Plus, and a computing build that someone has to actually pay for

The 15th FYP carries one phrase that does more work than any other for an investor: “new productive forces”. This is Beijing’s official language for the strategic shift out of the concrete-and-rebar growth model and into technology-led productivity, advanced manufacturing, and the AI-enabled industrial stack. The state has decided to spend the next five years levelling its economy up the value chain rather than rebuilding the apartment blocks.

AI Plus is the most concrete instruction inside it. The Outline’s Chapters 11 and 14 direct the “full implementation of the AI Plus initiative” and the construction of “a nationally integrated computing power network”. The same chapters state explicitly, in the official English translation, that the state will “support meeting compute demand through multiple mechanisms including government procurement of compute services and compute leasing”. That is industrial policy as a purchase order. It is not subtle.

Between an FYP chapter and a corporate result sits the State Council’s implementing notice. In March 2026, the State Council instructed ministries to issue sector plans against the Outline and provincial governments to issue local execution plans. Sector plans become provincial budgets, become procurement, become next year’s loan book at ICBC, become, eventually, the line in China Mobile’s results that says “computing services revenue +11.1%.” I cannot watch every step of the cascade. I can watch the company at the bottom.

The transmission line into my book is China Mobile, the largest of the three Mainland state-owned-enterprise (SOE) telcos and the cleanest example I have of how this works. (For the avoidance of doubt: I own all three; China Mobile is the primary AI Plus position, with China Telecom and China Unicom smaller weights for the same reason.) China Mobile spent RMB 150.9 billion on capex in 2025. It built up to 92.5 EFLOPS of intelligent computing capacity at FP16 (the standard half-precision benchmark for AI compute) and more than 1.5 million standard racks. Computing services revenue grew 11.1% to RMB 89.8 billion. AI services revenue grew 5.3% to RMB 90.8 billion. None of that says “AI Plus mandate” on the income statement; the Chinese state does not insist companies file in the language of state planning. It shows up in the next year’s results as bigger numbers in the sub-segment that does the work.

The cross-check sits next door. China Telecom last year reported industrial digitalisation revenue of RMB 147.3 billion, group capex of RMB 80.4 billion, and 91 EFLOPS of intelligent computing capacity. China Unicom’s 2025 results show the same shape at smaller scale. Three SOEs, comparable direction, comparable cadence. The AI Plus build is industry-wide and Plan-driven, not idiosyncratic to any one name. When I see three SOEs doing the same thing at the same time, I assume someone at the State Council told all three of them to do it. That assumption is rarely wrong.

The demand side is where my auto and sensor exposure earns its place. I covered the supply side in 50,000 Factories: What China’s AI Mandate Means for My Portfolio when MIIT’s January action plan landed. The demand side is the same wave hitting different income statements. BYD shipped over 4.6 million New Energy Vehicles (NEVs) in 2025. Geely crossed CNY 345.2 billion in revenue. Hesai tripled its LiDAR shipments year-on-year on RMB 3 billion of revenue. None of those companies sells to “new productive forces” as a line item. All of them sell into the demand wave the FYP frames at the chapter level. That is the difference between a slogan and a screen.

The Beautiful China indicators and a power generator nobody talks about

The Outline carries a table called Box 1. It lists the Plan’s quantitative majors over the next five years and tags each one as binding or expectative. The bound ones are the numbers a provincial cadre will be measured against at term end. The expectative ones are directional aspirations the system will work towards but will not punish on. The ecological cluster is the bit I take most seriously, and the targets are explicit, sourced, and date-stamped: PM2.5 below 27 micrograms per cubic metre by 2030, and non-fossil energy at 25% of primary energy consumption by 2030. For a utility, that is not mood music. It is the policy mechanism above the income statement: permits, dispatch priority, grid connection, local approvals, and the career incentives of the officials who sign them. The longer-horizon directional commitments, peak carbon before 2030 and carbon neutrality by 2060, sit in the 2035 Long-Range Objectives, which is a different document.

If you have not heard of Beijing Jingneng Clean Energy, that is fine. It does not market itself outside Hong Kong investor presentations and does not need to. It is a piece of Beijing’s municipal-SOE clean-power plumbing that happens to be listed. At end-2025 it held 18,365 megawatts of consolidated installed capacity. Renewables were more than 72% of it. It generated 42.45 billion kilowatt-hours over the year, of which wind contributed 15.97 billion, gas 19.02 billion, and PV 6.49 billion. The chapter says the share of non-fossil in primary energy has to reach 25% nationally. Beijing Jingneng moves the share inside its own book in the same direction, faster, and files its disclosures into HKEX where I can read them. The Plan does not need to name Beijing Jingneng; the filing already shows me which way the municipal-SOE plumbing is being pushed.

People-centred urbanisation and the household balance sheet

The Outline’s Chapter 31 is titled “In-Depth Advancement of People-Centred New Urbanisation.” It is one of the more consequential chapters and one of the easier to misread. People-centred urbanisation is not a return to the property boom. It is the explicit policy frame for moving people from villages into county towns, from county towns into mid-tier cities, with formal hukou (the household-registration system that gates access to schools, healthcare, and other public services in a given city) access, formal credit access, and formal social services. It is the chapter that tells you the next five years of Chinese domestic demand are going to come from a different pool of households than the last five. Four names in my book sit on this line.

China State Construction International (CSCI) is the build-side read. Its order book is heavily Modular Integrated Construction (MiC) work, the urban-renewal modality the Outline names directly. MiC is factory-built construction: more industrial process than old property cycle, more repeatable production line than apartment-boom labour stack. That is the version of construction I want exposure to if the post-property build still has to happen.

The banking read starts lower down the household ladder. Agricultural Bank of China (ABC) has the largest county-level branch network of any Chinese bank, founded on the “Three Rurals” mandate (agriculture, rural areas, rural households), already on file with the village and county households the Outline says will be moving up the urbanisation ladder over the next five years. The county economy is where hukou reform, infrastructure spending, and township-enterprise upgrades land first; ABC banks that economy as a matter of original mandate. The hukou line matters because it is where a migrant household starts becoming bankable as an urban household, not just visible in a migration table.

China Merchants Bank (CMB) sits at the other end of the same household migration. It is the leading retail and wealth-management franchise in Mainland China by client assets under management and by private-banking penetration, structured to capture the wealth-management fees, retail credit spread, and asset-management margin as the middle and upper-middle class that the Outline’s domestic-demand chapters describe continues to grow. The Plan does not have to mention either bank for the screen to matter. It describes the household whose financial life each of them is already structured to serve.

Ping An is the overlap case. One leg is the ageing chapter (Chapter 40), where its life and health insurance franchises sit directly on the structural tailwind: 7.6% operating profit growth in 2025 and resilient new business value in life are what the demographic line is supposed to look like inside an income statement, long before it shows up in the obituary pages. The other leg is the wealthier-household thread that lifts CMB: insurance, retirement, and life-stage savings products for a richer customer. One leg would not earn the size; two legs do.

None of these four is a breakout bet. All four are FYP-aligned. The urbanisation chapter does not promise any of them a return. It tells me they are running into a policy wind the state has decided to keep blowing for the next five years.

RMB internationalisation, the Belt and Road, and the part where boring banks earn their keep

Part Seven is mostly about external opening. Chapter 21 on autonomous opening-up. Chapter 23 on the high-quality co-building of the Belt and Road. Chapter 24 on the shared-future framework. This is the chapter cluster that sounds the most like a slogan inventory and the least like a procurement order, which is precisely why most foreign investors give it the lightest reading and miss what it does in practice.

I own Bank of China (BoC) for it. BoC’s 2025 results show continued growth in cross-border RMB settlement and a higher share of operating income from overseas, which is the literal income-statement signature of Chapter 21. I also own China Merchants Port (CMP). Port assets along the Belt and Road Initiative (BRI) corridors are the listed surface of Chapter 23, and CMP’s 2025 overseas terminal throughput and core-port profitability are what a multi-year corridor build is supposed to put on an income statement. Neither name will ever be a multi-bagger. Both have already done what they are meant to do, which is keep delivering operating cashflow while a chapter of the FYP says, on government letterhead, that what they do is to be encouraged. Names such as these fit my Income book.

If “RMB settlement bank with HK depth” and “overseas port operator with corridor exposure” are not the most exciting sentences in this article, that is the point. That is the job I want them to do in the Income book. The AI line can carry the excitement; these names have to carry cash.

Where the FYP got it right and the stocks still lost money

The FYP being decision-useful is not the same as the FYP being a return engine. Two cautions.

Plan-right does not equal stock-right at the sector level. The 14th FYP correctly diagnosed that the property cycle had to deleverage. It said so in writing, and the system acted on it. The Hang Seng Mainland Properties Index then fell from a peak near 4,000 in early 2021 to below 1,000 by 2024. The macro thesis was right. The equity outcome was a 75% drawdown. If you had bought sector exposure on the official narrative alone, you would have lost three quarters of your money being correct. FYP-aligned is the screen. It is not the position.

Geopolitics overrides the FYP at the firm level. The US Department of Commerce added Yangtze Memory Technologies Co. (YMTC, China’s leading memory-chip maker and a 14th FYP self-reliance flagship) to its Entity List in December 2022, and tooling-export controls have continued through 2024. No FYP can write its way around an export-control regime in another sovereign’s jurisdiction. If you cannot live with that risk, do not own these companies. I can, but I size these exposures as options on the self-reliance thesis, not core income positions.

I write my one-line rule with these caveats inside it, not after it. The rule is “FYP-aligned and worth what I am paying,” not “FYP-aligned.” The first half gets the company on the screen. The second half is the rest of the work, industry by industry, company by company.

The screen at work, and the test that comes next

The 15th FYP matters to me only if it moves from language into behaviour. That is the test.

The names in this article are not a recommendation list. They are the screen at work. The screen keeps me out of the wrong fights. It does not win the right ones for me. My actual book is several dozen Mainland and Hong Kong listings deep, and the screen runs across all of them, every quarter, against every important disclosure HKEX puts out. Names that do not sit in the chapters do not make the book, even when they screen cheap on backward-looking metrics. Names that do sit in the chapters still have to show me the translation: procurement, licensing, capex, loan growth, segment revenue, margin quality. Otherwise I have only found a slogan with a ticker attached.

The first real test of the screen lands over late 2026 and 2027, when State Council sector plans and provincial execution plans start translating Outline language into procurement, licensing, and capital allocation. After that come the corporate echoes: SOE annual results, segment disclosures, loan books, chairman letters, and the awkward phrases that have clearly travelled from the Plan into the company.

That is what I will be reading for over the next five years: not whether the slogans repeat, but whether the wording migrates into sector plans, procurement lines, loan books and SOE chairman letters. If it does, the manual is moving from paper into P&L. If it does not, the screen gets marked down.

As of the date of publication, I hold positions in Agricultural Bank of China (HKEX: 1288), Bank of China (HKEX: 3988), Beijing Jingneng Clean Energy (HKEX: 579), BYD Company (HKEX: 1211), China Merchants Bank (HKEX: 3968), China Merchants Port (HKEX: 144), China Mobile (HKEX: 941), China State Construction International (HKEX: 3311), China Telecom (HKEX: 728), China Unicom (HKEX: 762), Geely Automobile Holdings (HKEX: 175), Hesai Group (NASDAQ: HSAI), and Ping An Insurance Group (HKEX: 2318). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

I went to Chongqing in November to understand a city where three of my positions converge. I wrote about that trip in City Mechanics: What Chongqing Taught Me About My Portfolio. After it published, several readers wrote in about one detail I had folded into a single paragraph: the freight train from Chongqing to Budapest, opened the week I was there.

This piece is the answer. It is also, honestly, an example of the kind of detail I like to understand about a business before I invest. Not trains specifically, but how a company actually operates. That kind of understanding is what lets me tell fact from fiction, and hold through volatility. I own BYD. BYD’s first European assembly plant, Szeged, Hungary, EUR 4 billion, 300,000 vehicles a year at full scale, sits 160 kilometres from where those trains stop. The corridor opened on 30 November 2025: biweekly, eleven days, fixed timetable. Six months old.

When I formed my view on BYD’s European build-out, this line did not exist. There is now a logistics input sitting next to labour, tariffs, and local content that was not there before.

The Budapest Line

Chongqing has been running freight trains to Europe since 2011. The inaugural China-Europe Railway Express train left Tuanjie Village for Duisburg on 19 March that year. Most of those trains run on demand, with variable transit times. The Duisburg route went onto a fully fixed bilateral timetable some years back, China’s first. Budapest, opened on 30 November 2025, is the second. Biweekly, eleven days, run by Yuxinou (Chongqing) Logistics. Chongqing rail officials frame the line as a gateway to Slovakia, Austria, and Serbia, not only Hungary.

China shipped 108 TEUs to Hungary by rail in all of 2025. Not 108,000. 108. The corridor exists on paper and on a biweekly timetable; the volume does not yet exist at all. What follows is the work I did to figure out whether the corridor matters. If the volume gap closes, it adds a logistics input I had missed. If it does not, my view on BYD before discovering the train still holds.

Why Budapest specifically. BYD’s EUR 4 billion Szeged plant is in trial production since January 2026, with mass production targeted for Q2 2026. CATL’s 40 GWh Debrecen gigafactory sits 250 kilometres from the rail terminal, with cells already shipping. On the Chinese side, Chongqing’s automotive parts cluster, over 1,000 suppliers with BYD’s own Fudi Blade Battery factory at its centre, is the natural first-tier supplier base for Szeged. I walked the Chongqing supplier cluster in November. The geometry is unusually clean. One supplier cluster, one assembly plant, one fixed-timetable corridor between them.

As one enterprise representative told state media outlet Xinhua:

“Goods destined for Central and Eastern Europe previously required multiple transshipments or sea freight, with large lead-time variability. Now with fixed-schedule trains, we can even plan production schedules and inventory management by the timetable — supply chain resilience is greatly enhanced.”

Eleven days versus thirty-plus by sea. For just-in-time automotive supply chains, where carrying-cost savings on inventory partially offset the freight premium, the lead-time gap is what makes the corridor commercially relevant. Not the price. The timetable.

The 49% Premium

I built out the door-to-door freight economics myself, using current forwarder quotes and the Drewry container index. The short version: rail is expensive. Rail Chongqing-Budapest runs roughly US$9,600 per forty-foot container, end to end. Sea, routed via the Yangtze to a Northern European port and trucked inland to Hungary, comes in at roughly US$6,425. Rail is around 49% above sea on a like-for-like basis. These are post-subsidy quotes. Central per-container support for China-Europe rail wound down through 2023, and what’s left in Chongqing is land, tax, and infrastructure support to operators, not cash per container.

The premium does not vanish on a spreadsheet, but it does not fully reverse there either. Three weeks of vehicle parts sitting on water is working capital tied up; carrying-cost arithmetic recovers something like a tenth of the rail premium, not all of it. The rest is paid for something else: corridor geometry. A fixed-timetable train between one supplier cluster and one assembly plant lets a 300,000-vehicle plant schedule production against the corridor, not the swell on the Indian Ocean. Whether that optionality is worth a forty-plus-percent net premium is the live question, and the answer probably depends on which parts. Heavy, low-value-density components keep going by sea. Time-sensitive battery and powertrain content moves by rail. That is the split I would expect to see, and the one that will show up in volume data over the next eighteen months if the corridor is going to matter.

The Volume That Has to Show Up

Biweekly is a gesture, not yet a supply chain.

If you are scheduling production for a plant targeting 300,000 vehicles a year, two trains a month does not feed an assembly line. The Chongqing-Duisburg corridor took three years and a month from inaugural train to weekly cadence — March 2011 to April 2014. Budapest opened in November 2025. My base case is weekly cadence somewhere in 2028 to 2030, once Szeged is past trial production and the corridor is pulling volume from Chengdu and Chongqing together.

The math is bottom-up. Take Szeged at half nameplate, 150,000 vehicles, as a deliberately conservative short-term steady state. Per vehicle, roughly 600 kg of China-sourced battery and powertrain content: about 450 kg for a Blade Battery pack, 80 kg for an e-axle, 70 kg for motors and ancillaries. Across 150,000 vehicles, divided by about 13 tonnes per dense-cargo TEU, plus a 20% packaging and consumables buffer. That gets to roughly 8,300 TEUs a year of inbound Chinese content into Szeged.

The split by Chinese origin is the part I am least sure of. BYD has not published a supplier-by-origin breakdown. My working assumption is that around 35% of Szeged’s inbound Chinese content originates in Chongqing: the Fudi Blade Battery factory, the adjacent powertrain capacity, and the broader 1,000-supplier cluster across Chongqing and Chengdu make Chongqing the obvious anchor. That puts roughly 2,900 TEUs from Chongqing alone at the 150,000-vehicle steady state. Drop the share to 15% and Chongqing still carries roughly 1,250 TEUs at steady state, an order of magnitude above all of Hungary in 2025; push it to 55% and you get 4,600. I cannot pin the exact share, and I do not need to: Chongqing-rooted suppliers carry the bulk at either end.

All of Hungary took 108 forty-foot containers by rail from China in 2025. At the BYD plant at Szeged steady state, Chongqing alone could send around 2,900 a year, roughly twenty-five times the entire 2025 baseline, from one inland Chinese city. The corridor is six months old. The plant is not yet in mass production. That is the shape worth watching.

What Would Prove Me Wrong

Two ways the corridor stops mattering.

The 35% Chongqing-origin share assumes Szeged sources predominantly from Chongqing-rooted suppliers. If the Chinese content actually flows from Shenzhen, Xi’an, or BYD’s other hubs, the rail link is largely irrelevant to how I read BYD’s European build-out, relevant to someone else’s read, not mine. A supplier-by-origin disclosure from BYD, or freight-manifest sampling at the Budapest terminal, would close it.

The other way is faster than I expect. EU local-content rules and the Brussels push for European battery and component sourcing could pull Szeged’s parts demand into Hungary, Poland, and Czechia faster than the corridor scales. In that scenario the 49% rail premium loses its carrying-cost offset, because the comparison stops being rail-from-Chongqing versus sea-from-Chongqing and becomes rail-from-Chongqing versus a 400-kilometre truck from a Polish supplier. The logistics input I added washes out.

One Input, Not the Thesis

Not the thesis. I did not buy BYD because of trains, and one biweekly freight line does not move the underwriting on a company I have written about at length elsewhere.

What it changes is one input. How BYD’s European build-out actually runs now includes a Chinese rail-logistics line that did not exist before — small today, potentially material at Szeged steady state, contingent on the volume actually showing up. That is a thing to watch, not a thing to act on.

What it changes more durably is how I read China’s manufacturing reach into Europe. A fixed-timetable corridor between an inland Chinese supplier cluster and a Central European assembly plant is what the policy documents have been describing for years. Now there is a train. Whether it carries 108 containers a year or 8,000, whether the volume comes from BYD or someone else, the geometry has shifted. China spent two decades manufacturing on the coast and shipping out of it. Now it builds inland and puts the parts on a timetable.

This is the kind of work I left institutional finance to do, and the reason I publish what I learn. The longer story: Why Cohong Lane? A CFO’s Bet on Independent Investing.

As of the date of publication, I hold positions in BYD Company (HKEX: 1211) and Contemporary Amperex Technology (HKEX: 3750). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

I went to Chongqing to stress-test 3 China EV positions I already hold: BYD, Geely, and XPeng. I’ve written separately about what that trip taught me about my portfolio — but the question that kept forcing itself into my notebook wasn’t about any of them. It was about Huawei. More precisely: how Huawei gets paid by an industry whose private P&L will never tell you. The answer is on file in Hong Kong — not in Huawei’s accounts, but in the procurement ledger of the assembler that builds the cars Huawei sells.

I keep seeing what people call “Huawei cars” on the street — whether I am in Chongqing, Guangzhou, or Shenzhen. More precisely: AITO vehicles displayed and sold inside Huawei’s own retail footprint. In Chongqing I went into one of those stores. The staff lead with HarmonyOS, ADS, and the cockpit experience. But Huawei doesn’t build a single car. SERES — the Chongqing-based assembler — does.

SERES has now reported its FY2025 audited results, filed on HKEX on 30 March 2026: revenue RMB 164.9 billion (~US$22.7 billion), up 13.6% year-on-year; attributable net profit RMB 5.96 billion (~US$0.8 billion), essentially flat. Overall gross margin expanded approximately 3 percentage points to 26.88%. Top line growing, gross margin expanding, attributable profit flat. Standing in that Chongqing store, the question I couldn’t shake was simpler: if the assembler is generating real gross profit on a far larger revenue base, where is the operating leverage going at the net line? The number, in SERES’s procurement lines, is larger than I expected.

What follows is my attempt to answer that with primary filings and careful inference. Huawei is private, so I cannot prove its unit margin from any primary source — the case I can build is one of dependency rather than profit extraction, and dependency is the more important variable. This is how Huawei captures EV economics without building a car, and what it means for the Chinese manufacturers I hold.

What does a “Huawei car” look like on the ground?

On the ground, a “Huawei car” is an AITO vehicle displayed and sold inside Huawei’s own retail footprint, pitched entirely around HarmonyOS, ADS, and the cockpit experience. The assembler — SERES — is never mentioned. The brand on every wall is Huawei. The product being sold is the cabin, not the car.

The Huawei store in Raffles City Chongqing sits on an upper floor overlooking the confluence of the Jialing and Yangtze rivers — premium real estate, premium context. It’s set up less like a showroom and more like a personalised tech consultation. A staff member walked me through the car as if the vehicle itself were almost a formality: HarmonyOS, the ADS stack, the cockpit display, the seat experience, the audio environment. Horsepower was never mentioned. Neither was the drivetrain. Neither, notably, was SERES — the company that actually built the car sitting on the floor. The brand on every wall was Huawei. The pitch was the experience. The car was hardware.

That observation is not proof of anything on its own. But it calibrates how seriously to take what the industry calls the “third living space” — the idea, now mainstream in Chinese premium EV marketing, that the cabin itself is the product: screens, audio, lighting, near-flat bed seats with heating, massage and cooling, voice and gesture controls, and the ambient sense that the car is woven into your daily life. The drivetrain is assumed. The shell is almost incidental.

That framing tells you who holds the structural advantage. Huawei already wins in two living spaces — the home and the pocket. The car is the next one.

The open strategic question — whether HarmonyOS becomes the premium default in China the way CarPlay and Android Auto became defaults in the West — runs underneath everything that follows.

How does Huawei get paid without building a car?

Through procurement, not profit-sharing. In 2022, SERES purchased RMB 5.8 billion (~US$0.8 billion) from Huawei. In 2023, RMB 7.2 billion (~US$1.0 billion). In 2024, RMB 42 billion (~US$5.8 billion) — 30.2% of total purchases. In FY2025, RMB 56.1 billion (~US$7.7 billion) — 33.78% of total purchases, a new high. The FY2025 annual report names the largest supplier only as “IT, communications, and hardware equipment”; Chinese financial media identify it as Huawei. Cumulative procurement from 2022 through FY2025: approximately RMB 131 billion.

The SERES IPO prospectus is explicit:

“Our collaborations with Huawei do not involve any arrangements regarding profit-sharing, which are consistent with industry norm according to the Frost & Sullivan Report.”

The FY2025 annual report does not amend that arrangement. “No profit-sharing” does not mean “no economic transfer”. It means the transfer happens through ordinary operating lines — procurement and, likely, the retail channel. RMB 56 billion flowing to a single supplier tells me exactly who controls SERES’s cost stack — and it isn’t SERES.

To understand what that number actually means, it helps to put it in context. Apple’s dependence on Foxconn at its peak — the period that prompted Apple to spend a decade deliberately diversifying toward Pegatron, Luxshare, and Indian assembly — ran at roughly 25–30% of COGS. SERES is already at 34%, and rising. The difference is that Apple owned the brand and the customer relationship. In the Huawei store in Raffles City, it was impossible to tell who did.

The strongest counter-argument to this framing is that procurement is not profit. RMB 56 billion is top-line revenue for Huawei's automotive business, not a margin number, and I cannot prove Huawei's unit economics on those components and services from any primary filing. Huawei could in principle sell at or near cost, treating SERES as a platform anchor rather than a profit centre. I doubt it: Huawei's 2024 annual report disclosed that its intelligent automotive solutions unit turned a profit for the first time, and software-and-cockpit revenue carries software-margin economics, not commodity-component economics. What I can prove is the dependency, and dependency is the more important variable. Whether Huawei is currently extracting profit or simply locking in the platform position, the result for SERES is the same: a third of every yuan it spends on inputs goes to one supplier, and that share is climbing.

SERES’s selling expenses ran RMB 24.2 billion (~US$3.3 billion) in FY2025 — 14.7% of revenue, up from 13.2% in 2024. Consistent with what you see on the ground: premium stores, premium service, and a sales story led by the platform brand, not the assembler. SERES still posted a 3.6% attributable net margin in FY2025 (4.1% in 2024) — proof the assembler can be profitable in the Huawei era. But profitable with a cost structure where Huawei is the largest single supplier, where the share of inputs going to that supplier rose by another 4 percentage points in the most recent year, and where revenue grew 13.6% while attributable profit did not move at all.

What happens when Huawei replicates the platform?

Huawei is already doing it, and the cumulative volume across the Harmony Intelligent Mobility Alliance (HIMA) — Huawei’s smart car joint-venture framework through which it supplies ADS, HarmonyOS, and hardware to OEM partners without taking an assembly stake — settles the question. HIMA delivered 112,700 vehicles in Q1 2026 (+41.9% YoY), bringing cumulative deliveries to approximately 1.35 million units by March 2026. The AITO M9 alone has now passed 280,000 cumulative deliveries and outsold BMW X5, X7, Mercedes GLE and GLS in China on a cumulative basis. The AITO M6 took 60,000 pre-orders in the first 24 hours of pre-sale on 23 March 2026. The platform isn’t trying to scale. It has scaled.

The share split tells the same story. Five brands now sell across 4 assemblers: AITO at SERES, LUXEED at Chery, STELATO at BAIC, MAEXTRO at JAC, SHANGJIE at SAIC. AITO’s share of HIMA fell from above 85% in 2024 to roughly 63% by November 2025 as new brands joined; it has since recovered to roughly 69–76% in Q1 2026 as the non-AITO ramp ran slower than expected.

The structural implication is not “Huawei wins”. It’s that platform scale changes OEM bargaining power over time. SERES can be profitable and still become more dependent — and those two things are not in contradiction.

SERES has responded on 3 fronts — none of them the moves of a comfortable partner. They are the moves of a company managing a dependency it knows is structural.

Trademark reclaim. In July 2024, SERES paid RMB 2.5 billion (~US$0.3 billion) to buy 919 AITO trademarks back from Huawei. A company paying to own the customer identity it built in partnership tells you something unambiguous about who the brand equity originally belonged to.

Technology stake. SERES invested RMB 11.5 billion (~US$1.6 billion) for what was originally a 10% stake in Huawei’s Yinwang autonomous driving unit, now diluted to 9.36% as Huawei admitted other partners. The FY2025 annual report recognised approximately RMB 170 million of equity-method income from this stake — accretive at the income line, but a 1.5% return on capital implies the investment is a strategic lock-in, not a financial one. Changan’s AVATR made the same move independently at the same percentage. This is becoming standard practice for HIMA partners who want a seat at the technology table rather than just a supply agreement.

Capital raise. SERES listed in Hong Kong on 5 November 2025, raising approximately US$1.7 billion in fresh capital.

The platform toll and my portfolio

The thesis is simple: Huawei has inserted itself into the value chain, the customer relationship, and the standard-setting layer of Chinese premium EVs — without taking assembly risk.

I didn’t go to Chongqing to find a new stock. I went to stress-test positions I already hold — and Huawei is now inside that underwriting.

The risk isn’t that BYD, Geely, or XPeng can’t build EVs. It’s that Huawei is redefining what Chinese consumers are paying for in the premium segment. Not the drivetrain. The cabin.

Two mechanisms are at work, and they break differently. For HIMA partners — SERES, Chery, BAIC, JAC, SAIC — the toll is procurement, visible at SERES’s 33.78%. For the three I hold — BYD, Geely, XPeng — the risk is competitive displacement: if HarmonyOS becomes the premium default, their in-house cabin stacks become an expense rather than a moat.

BYD has the scale to respond — but only if Denza and YangWang can match the Huawei experience baseline, not just approach it. Geely’s Zeekr is betting in the opposite direction: building its own independent cabin stack (ZEEKR AI OS, Qualcomm Snapdragon 8295) rather than paying the Huawei platform toll. The risk isn’t dependency — it’s irrelevance. If Chinese consumers decide HarmonyOS is the definitive premium standard, Zeekr’s in-house software investment becomes a sunk cost. XPeng built its own stack as a moat; that same stack becomes an expense if HIMA reaches parity fast enough.

I’ll be direct about the uncertainty here: I don’t know which of these resolves first, or how fast. But these are the 3 data points that would change my view:

1. The experience gap. The market is already bifurcating along integration lines. My base case is that the gap widens through 2026 as HIMA launches new models at pace. The break condition for my OEM positions: BYD (Denza or YangWang), Geely (Zeekr), or XPeng demonstrably matching HIMA cockpit benchmarks in independent comparisons by H2 2026. Until then, Huawei is setting the floor.

2. HIMA’s spread. AITO fell from 85%+ of HIMA volume to roughly 63% in November 2025, then recovered to 69–76% in Q1 2026. If it falls below 50% on a sustained basis, Huawei no longer needs SERES to anchor the platform — and SERES loses whatever negotiating leverage it currently holds. If it stabilises above 60%, the relationship is maturing into something bilateral enough to be manageable. The current data point sits in the second camp, but Q1 monthly figures swing widely.

3. The procurement mirror. 33.78% and rising at SERES. As of late April 2026, none of the FY2025 annual reports filed by Changan, BAIC BluePark, JAC, or SAIC show Huawei or Yinwang as a top-five supplier or named related-party transaction at the parent-group level — most likely because the brand-level volumes are still too small to register in parent-group COGS. AVATR has filed for a Hong Kong IPO at a reported US$4.6 billion valuation; when it lists, its prospectus and first annual filing will be the next clean read. Three HIMA partners showing 25%+ procurement dependency would be the clearest possible signal that Huawei has restructured the cost stack of the Chinese premium segment — invisibly, without building a single car.

I hold my positions. But I’m now measuring each of them against one question: can they compete on the thing Huawei is actually selling? SERES’s procurement lines already show what the platform toll costs. The next 18 months will tell me which other OEMs have been paying it all along.

Cohong Lane is where I publish the work I do on the China positions I actually hold — written from Hong Kong, sourced from HKEX filings, calibrated against what I see in the cities where these companies operate. The next piece from the Chongqing trip is I Own BYD. Chongqing’s 11-Day Train Changed My Math. — on the rail corridor to BYD’s Szeged plant that I had not modelled before visiting.

As of the date of publication, I hold positions in BYD (HKEX: 1211), Geely Automobile (HKEX: 0175), and XPeng (HKEX: 9868). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

In a car from Chongqing East station to the hotel, I spotted a vehicle I had never seen before. Low, sculpted, with lines that belonged in a concept video rather than on a road. The badge read AVATR — a joint venture between Changan Automobile, Chongqing’s anchor state-owned automaker, CATL, and Huawei. Three of China’s most consequential industrial players, converging in a single vehicle, designed and assembled in a city roughly 2,000 kilometres from the nearest major seaport.

I hold BYD, CATL, and China Mobile. All three have operations here: BYD’s Blade Battery plant, CATL’s first “factory-in-factory” production line, one of China Mobile’s 8 national data centre hub nodes. The mountain topography over the Yangtze is genuinely dramatic, but I did not come for the scenery. I came to work out what that overlap in my own book actually meant.

What Role Does Chongqing Play in China’s Economic Plan?

My working view after the trip: Chongqing’s logistics stack is a potential structural edge for BYD’s and CATL’s European supply chains that was not in my model. The rest of this piece is the case, and the one signal that would falsify it.

Chongqing is one of only 4 direct-administered municipalities (直辖市) in China — alongside Beijing, Shanghai, and Tianjin. That status means it retains a larger share of tax revenues, issues its own municipal bonds, and competes directly for central government capital without provincial intermediation. When Beijing allocates RMB 800 billion (~US$110 billion) in ultra-long special treasury bonds for infrastructure in 2026, Chongqing sits at the same table as Shanghai.

“Boost the development capacity of the Chengdu-Chongqing economic zone.” — CPC Central Committee 15th Five-Year Plan Recommendations, Section 29 (October 2025)

The policy stack underneath is what matters: direct municipality status, the Liangjiang New Area, a Pilot Free Trade Zone handling roughly 70% of the municipality’s foreign trade across just 0.1% of its land, and the China-Singapore Demonstration Initiative — a government-to-government project giving Chongqing privileged access to Singapore’s capital networks and ASEAN markets. No peer inland city has this many stacked in one place. Not Chengdu, not Wuhan, not Xi’an.

The Numbers That Matter — and Where the City Is Fragile

Chongqing’s 2025 GDP reached RMB 3.376 trillion (~US$485 billion), growing 5.3% against a 5.0% national average — enough, for the first time, to overtake Liaoning, the old industrial heart of the northeast. The city missed its own 6% target by 0.7 percentage points, and has lowered the 2026 target to “above 5%”. The automotive industry kept doing most of the lifting: smart-connected NEV value-added rose 13.4% on the year, and the city’s NEV output has stepped up from roughly 43,000 in 2020 to 1.296 million in 2025, reclaiming the title of China’s top auto-producing city.

Three numbers cut against the headline.

First, the NEV growth rate has already halved. The 90.5% print in 2024 dropped to 36% by 2025, and I expect Chongqing will converge toward the national rate within 2 to 3 years.

Second, the concentration is brutal. The automotive industry alone contributed two-thirds of Chongqing’s entire industrial expansion in 2024. Two-thirds. From one sector.

Third, the population is shrinking. A net loss of 92,000 people in 2024, against an economy that still grew 5.7%. If your China model starts with demographics, Chongqing is the city that breaks it — the growth is coming from productivity and capital upgrading, not headcount, and that is a harder trick to repeat than a demographic tailwind.

Taken together: if automotive growth decelerates to 10–15% before the named replacement sectors — electronics, advanced materials, AI, biomedicine — reach scale, Chongqing faces a growth gap. Not a crisis for the city, but a reason to read every subsequent Chongqing data release more carefully than I used to.

Is the Advantage Structural, or Just Subsidies?

Every major Chinese city offers subsidies, land deals, and tax breaks to attract manufacturers. By CSIS’s Red Ink estimate (Scott Kennedy, 2024), China’s EV industry received roughly US$230 billion in state support from 2009 to 2023, with local governments accounting for about 80% of industrial policies. BYD’s largest megafactory is in Zhengzhou, not Chongqing. CATL’s main cell production hubs are in eastern China.

The answer is logistics — the kind that cannot be replicated by writing a bigger cheque.

Chongqing is the only city in China to hold all five categories of national logistics hub designation simultaneously — port, land-port, airport, production-service, and trade-service. A manufacturer here can route goods east down the Yangtze via Guoyuan Port, west to Europe by China-Europe Railway Express, south to ASEAN via the New International Land-Sea Trade Corridor, or by air through Jiangbei Airport. Four export corridors from a single inland hub. Xi’an moves more CRE rail volume, but Xi’an does not have the Yangtze. Chengdu shares the western rail corridor, but Chengdu does not have the river port. No peer city replicates the combination.

The detail that made me sit up: while I was in Chongqing last November, Xinhua’s Chongqing wire reported (30 November 2025) that China had launched its second fixed-schedule freight rail service to Europe — from Chongqing to Budapest, Hungary, running biweekly in approximately 11 days. The first route, to Duisburg in Germany, has been running since 2011. Why does Budapest matter? Because BYD is building a €4 billion plant in Szeged, Hungary — 160 kilometres from Budapest — its first European passenger-car factory, targeting 300,000 EVs a year at full capacity, with pilot production in Q1 2026. CATL’s Debrecen gigafactory — a €7.34 billion plant, initial capacity 40 GWh with cell production from early 2026 and designed for 100 GWh at full build-out — is 250 kilometres away. When Chongqing’s 1,000-plus automotive parts suppliers need to ship components to those European assembly lines, the timetable now exists. 11 days by rail versus 30-plus by sea makes rail competitive for just-in-time supply chains. For the first time, an inland Chinese city has a direct, timetabled logistics link to the Central European EV cluster.

Ground Truth — What You Won’t Read Elsewhere

To understand Chongqing, you have to leave it.

One morning we drove 90 minutes southwest to the Dazu Rock Carvings. The carvings are extraordinary — a UNESCO site, 50,000 sculptures carved into cliff faces since the 7th century. But it was the drive that changed how I think about the data.

Within 40 minutes of leaving the urban core you are in a different economy. The road narrows. The buildings are older, lower, unfinished. This is the other Chongqing: 82,400 square kilometres of mountain terrain where the main urban district produces 78% of GDP. The per capita income gap between urban residents (RMB 49,778, ~US$6,800) and rural residents (RMB 22,221, ~US$3,000) is more than 2-to-1. An investor assuming the headline applies evenly is making an error I nearly made myself.

Restaurants were packed — not the performative busy of a new development, but the organic busy of a consumer economy actually spending. Wholesale and retail grew 9.5% in 2024. Accommodation and catering grew 7.3%. You feel these numbers in the queue for a hotpot table at 9pm on a Tuesday, not in a spreadsheet. (If you visit: skip the tourist hotpot and try 烤匠麻辣烤鱼. It is a Chengdu-born chain — and their sea bass, buried under a small mountain of dried chillies and green Sichuan peppercorns, is the best thing I ate in China in 2025. 不吃火锅，就吃烤匠.)

One evening walking through Chongqing’s malls is not statistically significant. I know that. But what on-the-ground observation gives you is not data — it is a filter for obvious nonsense in the data. It does not prove a bearish report wrong. But it tells me to ask better questions about which consumers, where.

What Does This Mean for My Portfolio?

I came to Chongqing to understand a city. Two of my positions got sharper. One did not.

BYD operates a dedicated Blade Battery factory here through its Fudi subsidiary: 8 production lines, 20 GWh annual capacity, running at or near full utilisation since 2020. What I had not connected was the logistics. The Budapest rail link means BYD can ship battery components to Szeged in 11 days without routing through Shanghai. For a company building out its first European passenger-car production base targeting 300,000 EVs a year, that is a potential structural cost advantage — contingent on the Budapest schedule going weekly — that was not in my model. It is now.

CATL entered Chongqing in June 2025 with a factory-in-factory model: 2 battery pack lines physically inside the SERES Super Factory in Liangjiang, a 5-year exclusivity agreement for all AITO vehicles to use CATL batteries, planned annual output value of RMB 7 billion (~US$960 million). I knew the concept on paper; being in the city where it runs forced me to focus on what it means — switching costs that are physical and contractual, not just commercial. Beyond that, the Chongqing presence did not change my view of the position; I hold CATL on the global platform thesis, not on a single-site read-across.

China Mobile also runs a Chongqing node — one of 8 national “Eastern Data, Western Computing” hubs. Nothing on the ground changed my view of the position; I hold it on the telco thesis, not on a Chongqing read-across.

The Signal I Am Watching

In that first car ride from Chongqing East, I thought I had spotted an unfamiliar vehicle. What I had actually spotted was my own blind spot. The AVATR was the whole city in miniature: Changan, CATL, Huawei, capital, policy, manufacturing, all compressed into one thing moving down the road.

The question for me now is whether Chongqing's logistics stack can turn that industrial density into a durable edge for my BYD and CATL positions, rather than just an impressive story. The Budapest rail schedule is the test. If it goes weekly by end-2026, I will treat that first sighting as an early clue I was lucky enough to notice. If it does not, the car was still real, but the edge I thought I saw behind it was smaller than I believed. And the deeper question, which I unpack separately in Why Huawei Now Sits Inside My China EV Underwriting, is how much of that value is ultimately being captured not by the assembler, but by the platform layers sitting inside the same machine.

As of the date of publication, I hold positions in BYD (HKEX: 1211), CATL (HKEX: 3750), and China Mobile (HKEX: 0941). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

In April 2025, I was in Budapest, meant to be on holiday, and instead spent an indecent amount of time staring at falling prices on my screen. I felt the drawdown in the only way that matters: as real money, in my own account, disappearing by six figures while every instinct in my body suggested that doing something, anything, would feel better.

Budapest is a beautiful city, which is not especially helpful when tariffs hit. More important than the trades I did or did not make was the thing going on in my own head: I felt fear, I felt regret, and I felt the very human urge to convert uncertainty into certainty, even at a bad price.

The problem is that my instincts do not distinguish particularly well between actual danger and falling share prices. Falling share prices are not automatically danger. They are information, and the job is to work out what kind. The market was not telling me my thesis was wrong. It was telling me sentiment had turned.

I did not sell, though, and the reason was not that I am some sort of monk in a linen shirt who has transcended loss aversion. I am not. The reason was simpler and less glamorous: I knew what I owned, I knew why I owned it, and I knew what I thought it was worth.

What is volatility, and what is it not?

Howard Marks made the distinction plainly in “Risk Revisited Again“: what investors actually fear is not price movement but the possibility of permanent loss. Volatility became the academic proxy because it was quantifiable, but it was always the wrong measure.

I agree with Marks — but I want to take the distinction one step further, because in my own portfolio the useful question is not what volatility is but when it becomes dangerous. The answer, as far as I can tell, is: through only two mechanisms. Either my structure forces me to sell, or my psychology does. Permanent loss itself is narrower than most drawdowns suggest — fraud, obsolescence, terminal competitive decline, or regulatory destruction. Four ways a business becomes permanently worth less. If none of them is in play, the price is moving without the business moving, and moving prices are noise, however unpleasant.

The first mechanism is structural. This is where a lot of institutional pain comes from, and it is also why private capital with the right design has a real edge. If you are running leverage, a drawdown can become a margin call. If you are managing outside money, a drawdown can become a redemption. If you are judged quarterly on relative performance, a temporary decline can become a career problem long before it becomes an investment problem. Time horizon is not a decorative preference in investing. It is often the whole game.

That is why my own portfolio is built to eliminate forced selling. My living expenses are covered by income-generating assets. I do not use leverage. I do not manage external capital. That removes the first channel almost entirely.

The second mechanism is psychological, and this is the harder one because it follows you around. You can design away leverage. You cannot design away being human. If I sell simply because I cannot bear the discomfort of seeing prices down another 10%, even though my underlying thesis is intact, then I have voluntarily converted volatility into permanent loss. Nobody forced me. I did it to myself.

Part of what makes the psychological channel so difficult is that the fear feels rational. A falling price appears to carry information — the market is selling, so it must know something you do not. That instinct is not stupidity; it is pattern recognition misfiring. Price movement is not the same as informational superiority. When the market sells, it tells you about the constraints of the people selling — not about the business.

What actually constitutes permanent loss?

In my experience, most drawdowns have the same basic explanation: the market is selling off for reasons that are only loosely connected to the underlying business. I don’t buy stocks — I buy tiny pieces of what I believe to be great companies, at prices below my estimate of intrinsic value. When prices fall but the business hasn’t changed, that’s not a warning. That’s a price decline worth sitting through, not running from.

When a position I own is in a genuine drawdown, I run four questions before I touch anything. Has management integrity changed — because I have been wrong about management before, and it is usually the first thing I am too optimistic about. Has the industry structure worsened, not cyclically, but in a way that permanently favours someone else. Has regulation impaired the model — not inconvenienced it, impaired it. Has the balance sheet become a real problem, or am I just watching a line item worry me. If none of the four has shifted, I am almost certainly looking at volatility, not impairment.

Why does homework create holding power?

Deep research creates holding power because conviction anchored in business reality can survive a falling share price, whereas conviction anchored in price action cannot. If I understand the industry, the competitive position, and a sensible fair value range, then a drawdown changes my emotional state, not my estimate of intrinsic value.

This is the bit that gets romanticised in investing books and then brutalised by actual markets. People say “just hold through volatility” as if that was a personality trait. Like being tall. It is not.

Holding power is downstream of preparation.

If I have done the work properly, a 30% decline tells me the market’s mood has changed. It does not automatically tell me the business has deteriorated by 30%. Those are wildly different statements, but markets present them in the same font, which is unhelpful. The screen says the bid is lower. Fine. The screen does not tell me whether the bid is rational.

That, to me, is the real reason most investors cannot hold through a continued decline: they do not truly understand what they own. If your conviction is anchored in price action, a further decline dissolves it. If it is anchored in the business, the decline is just the price changing, which is a different thing entirely.

I should be honest about the obvious objection: the diagnostic is only as good as the analyst running it. If I have spent months building a thesis, I am not a neutral examiner of the evidence — I am a motivated one. The risk is not that I ignore the checklist but that confirmation bias shapes what I see when I run it. I do not have a risk committee to overrule me. What I have is the discipline to name, in advance, the specific evidence that would change my mind — and to write it down before the drawdown arrives, when my judgement is not yet compromised by the desire to be right. It is not a guarantee against self-deception, but it is the best structural defence I have found so far.

The time to do the work is in calm weather — deciding what I want to own, at what price, in what size. A GTC limit order is exactly that: a standing instruction placed before the storm that fires automatically when the price arrives, without requiring my emotional state at 3am to agree. When the sell-off arrives, the calm-weather version of you is already at the desk.

Familiarity, Not Heroics

Budapest tested whether that preparation held. The episode mattered because I did not just endure it; I used it as deliberate psychological training. The useful skill in a sell-off is not emotional numbness but the ability to observe fear, regret, and the urge to act without automatically obeying them.

The Budapest drawdown felt awful. I did not enjoy it. I was not floating above it in enlightened detachment, whispering about intrinsic value while the minibar mocked me from across the room.

What I did differently was pay attention. Most investors respond to a sell-off in one of two ways. They either stop looking entirely, or they stare at the screen in a state of rising panic and let each downtick rewrite their beliefs. I tried to do neither.

I watched the losses accumulate. I named what I was feeling: fear, regret, frustration, the desire for relief. And then I separated those feelings from the actual state of the businesses I owned. That gap — between what I felt and what I knew — was the whole exercise. My feelings were screaming catastrophe. My research was saying, in effect, “the market is upset; the businesses are still the businesses.”

Budapest was not my first sell-off. The first time a position of mine dropped 30% I spent three weeks second-guessing every decision that led to the buy. By Budapest I spent a lot less time doing that. The gap is not willpower. It is evidence. My subconscious has seen enough of these chapters to know how they usually end.

I want to be honest about this: it is genuinely difficult. Anyone who tells you they feel nothing when watching large drawdowns accumulate on real money — not paper money, not a simulation, and with no salary arriving at the end of the month to soften the edges — is either exaggerating or has not yet been tested with enough capital at stake. The goal is not to feel nothing. The goal is to feel it, understand it, and choose not to act on it. That, to me, is the real edge of lived experience.

Not heroics. Familiarity. The first major drawdown feels like an emergency. By the fifth, it starts to feel more like weather.

Where I Paid the Admission Price Most Recently

China equities are where the admission price is highest — and where confusing it with risk is most expensive. Sentiment swings are amplified by policy headlines, geopolitical friction, and Western media narratives that default to catastrophe. If you conflate that volatility with risk, you will never hold through the drawdowns that create the best entry points.

In April 2026, when the Iran headlines hit and the Hang Seng sold off, I noticed the familiar sequence almost immediately. The screen made everything feel more urgent than the facts justified. The impulse was the same as in Budapest: do something, reduce the discomfort, convert uncertainty into action.

So I ran the checklist before I let myself touch the order ticket. Had anything changed about the businesses I was watching? Had the competitive position worsened? Had the policy structure shifted? The answers were no, no, and no. What had changed was sentiment.

That distinction matters because the psychological danger in a sell-off is not fear by itself. It is fear smuggling itself into the analysis and pretending to be new information. Once I could see that happening, the job was simple, if not pleasant: slow down, go back to the work, and act only if the gap between price and value had genuinely widened.

That was the setup with Kingdee. I had done the homework months earlier, when I traced how Beijing’s binding factory-automation KPIs were creating a mandate-backed upgrade cycle for enterprise software. The thesis was ready. The fair value range was set. When the Iran sell-off pushed the price into that range, I bought. Same with Hesai — the production wins were already in the work, the certification path was already in the work, and the question in front of me was not whether the headlines felt alarming but whether the business had changed. It had not.

The important part, though, is not that I bought them. It is that the decision did not come from adrenaline. It came from homework done in calm weather, checked again under pressure, and executed only after I had separated what I felt from what I knew. That is the admission price in practice.

If you’re new to Cohong Lane, that is the whole point of the publication: I invest my own capital from Hong Kong and publish the process, the pressure, and the mistakes in public so readers can inspect the reasoning, not just the conclusion.

I have sat through drawdowns that lasted quarters, not days — the 2022 China panic ground on for quarters while every Western headline declared the market uninvestable. I ran the same diagnostic then. I held. The framework held.

Volatility is the admission price. Risk is what happens if I have built a portfolio, or a mind, that cannot afford to pay it.

As of the date of publication, I hold positions in Hang Seng Tech ETF (Xetra: H4ZX), Kingdee International Software Group (HKEX: 0268), and Hesai Group (NASDAQ: HSAI). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

BYD ended 2025 by telling shareholders that China is no longer enough

I own BYD, and this piece is a check on whether the reason I own it still holds.

I did not buy BYD because I thought it would post a neat set of annual results in one particular year. I bought it because I thought it might become something much rarer: a globally localised, cycle-resilient, structurally profitable auto-industrial platform with the scale, staying power, and foreign industrial presence that make the Toyota comparison worth taking seriously. It is the same lens I use across the rest of my portfolio, which I laid out in How I Invest My Own Capital.

BYD sold 4.60 million new-energy vehicles in 2025, exported more than 1 million of them, and lifted revenue to just over RMB 804 billion. On the surface, that still looks like a company scaling.

Underneath it, the year tightened. Gross margin fell to 17.74 percent from 19.44 percent. Net profit fell 18.97 percent to RMB 32.6 billion. Operating cash flow dropped from RMB 133.5 billion to RMB 59.1 billion. Capex jumped to RMB 156.8 billion. Borrowings rose from RMB 28.6 billion to RMB 113.4 billion.

The Toyota comparison demands more than scale. It does not mean BYD gets huge and keeps shipping cars out of China. It means BYD becomes the sort of company that can build at scale across regions, make decent money through a cycle, localise production when politics harden, absorb tariffs and price wars without breaking the balance sheet, and keep compounding because the industrial system underneath it is real.

What management is saying is revealing

Management is clear on one point: the domestic market is brutal. BYD's 2025 annual report describes domestic competition as having entered a "fever pitch" and a brutal "knockout stage" — unusually direct language for a corporate filing. Beijing is signalling the same thing: regulators have moved to curb destructive price competition and below-cost selling in autos. That tells me the company knows the home market is now a stress test, not a comfortable source of compounding.

Second, overseas is not just a volume story. It is an economics story. Overseas revenue rose 40 percent to RMB 310.7 billion while China revenue fell 11.2 percent to RMB 493.2 billion. Overseas now accounts for roughly 38.7 percent of group revenue, up from roughly 28.6 percent a year earlier. Management also confirmed that overseas profitability is structurally stronger.

Those two facts together are enough to tell me the direction is right. The magnitude is the hypothesis I am watching.

I want to be careful with that.

Management commentary is not segment disclosure. BYD does not yet report overseas as a separate segment, which means I cannot verify the full economics directly from the filing. What I can say is that the revenue mix shift is real, the directional commentary is consistent, and the hypothesis makes industrial sense in markets where BYD is not fighting a domestic price war. The first clean test will be whether BYD breaks out an overseas segment in the 2026 interim results. If that happens, the hypothesis becomes a fact or a disappointment. Until then, it is the part of the thesis I watch most carefully.

If that profitability gap holds as the business scales and localises, overseas is not just the faster-growing segment. It becomes the part of BYD that could simultaneously be more profitable and more durable — insulated from domestic pricing dynamics by geography and brand positioning.

That leaves China as the base case: still dominant, still under pressure. And overseas as the part of the thesis that has to prove itself.

You could reasonably argue that BYD does not need to pass this test at all — that a company selling nearly five million vehicles a year with vertical integration and deep cost advantages can compound perfectly well as a domestic champion that exports on the side. That is a defensible case, and it may turn out to be the right one. But it is not the case I am underwriting. What I bought was the potential for something rarer: a company that can build durable economics in foreign markets, not just ship volume into them. If BYD turns out to be only a domestic champion, it will still be a good company. It will not be what I paid for.

Third, management is signalling that the build-out will take precedence over distributions. The proposed 2025 dividend was RMB 0.358 a share, down from RMB 3.974 in 2024, and management tied the cut to operating cash flow and future development needs. As a long-term shareholder, I think that is the right call. This phase has to be funded, not wished into existence.

What management is doing matters more than what it is saying

Global ambition is cheap. Industrial proof is not.

What I look for when I read a chairman’s letter is the gap between what management emphasises and what the numbers actually show. Wang Chuanfu’s language — “fever pitch”, “knockout stage” — is significant not because it is alarming, but because it is honest. It is a chairman telling shareholders that the home market no longer offers a comfortable base from which to fund overseas expansion. That forces you to ask a different question: if domestic economics weaken further, can the overseas business grow fast enough and profitably enough to absorb the pressure? That question is what this piece is really about.

BYD is starting to supply the industrial proof: trial production commenced at the Szeged, Hungary plant in January 2026, with full series production scheduled for Q2 2026 — the first locally manufactured European BYD. The Brazil plant is operational. Cambodia is in progress. Turkey’s 150,000-unit plant is scheduled to open by end-2026, and Spain is under active evaluation as a third European site, favoured for its clean energy infrastructure and manufacturing cost base. The company now runs eight roll-on/roll-off vessels and has established a European headquarters in Hungary.

That is not proof on its own. It is a footprint. Anyone can export. The Toyota test is whether exports turn into factories, suppliers, service networks, and local political staying power. That is a harder path than SAIC’s MG-style push into Europe, and more self-built than Geely’s ability to lean on Volvo’s foreign industrial base. What makes BYD’s path distinctive — and more fragile — is that it is building the industrial infrastructure itself, from shipping to charging to assembly, without an acquired brand to backstop the learning curve.

That is also why capex matters. BYD spent RMB 156.8 billion in 2025 on overseas plants, proprietary shipping, charging infrastructure, energy storage, and R&D simultaneously. The bear case is that management is trying to build five moats at once and cannot fund them all without the domestic cash engine running at full pressure. The bull case is that this is what a real moat often looks like while it is being constructed.

The 2026 test for the Hungary plant is not whether it produces cars. Trial production has already started.

The real test is whether it runs at a utilisation rate and cost structure that can compete with Chinese production after European labour and logistics costs are taken into account. Underutilised overseas plants drag margins. If Hungary starts reporting at meaningful capacity — initial design is 150,000 units, scalable to 300,000 — before the domestic business stabilises, the thesis gains a genuine second leg.

For me, localisation does not mean putting pins on a map. It means local suppliers, after-sales capability, political acceptability, demand that holds up without brute-force discounting, and eventually economics good enough to survive tariffs or policy tightening.

The strategic direction only matters if BYD can export a platform, not just cars

BYD’s technology roadmap matters only if it helps the company win outside China — in markets with weaker charging infrastructure and harder politics.

Fifth-generation DM and longer-range PHEVs matter because large parts of the world still do not have the charging infrastructure that makes pure EV ownership easy. Management is pushing larger-battery PHEVs with up to 210km of pure-electric range and more than 2,100km of combined range. That is aimed at markets where charging is patchy and drivers still need ICE-like convenience. It is a product designed for the world as it actually exists, not the world as EV evangelists wish it did.

The individual technologies — Super e-Platform, flash charging, the second-generation Blade Battery, Gods Eye — are not the point. Management is bundling them: charging speed, battery range and durability, driver intelligence, local assembly, and shipping control, packaged as a single offer that is harder to replicate than any one feature. If you are a competing OEM, you do not get to pick which part to match. I explored a related policy-to-cashflow angle in 50,000 Factories: What China’s AI Mandate Means for My Portfolio, where BYD also appears as part of the EV manufacturing and automation build-out.

One reason BYD might become more resilient than a normal carmaker is that it may not end up being only a carmaker. The company delivered more than 60GWh globally in 2025, ranked first worldwide in energy storage system shipments, and won the 12.5GWh Saudi Electricity contract. That is enough to tell me energy storage is not a side business. What I still cannot see clearly is the economics. Because storage sits inside the broader auto segment disclosure, I cannot tell whether it is already a margin enhancer, a temporary drag while the business scales, or simply too small to matter yet. Management has given me enough to be intrigued and not enough to underwrite. That is the real gap.

Premiumisation matters too, but I want to be disciplined about it. Yangwang, DENZA, and FANGCHENGBAO together approached 400,000 units in 2025 and more than doubled. That is encouraging. It is not yet the same as saying BYD has become a premium company. It has become a company that also sells some premium cars, which is a different sentence.

What 2025 actually changed for me

The annual report did not kill the thesis. It raised the hurdle.

Before this report, I could still tell myself a simpler story: China scale first, exports later, and operating leverage eventually doing the rest. I do not think that is the story any more.

The domestic business weakened, cash flow deteriorated, capex surged, and borrowings nearly quadrupled. Those numbers do not decide the thesis on their own. But they leave management with far less room for strategic error.

I remain invested because BYD has already proved enough to keep the thesis alive. But from here, the next leg has to be earned.

How I am positioned — and what would change that

BYD currently represents approximately 2 percent of my growth portfolio, with additional indirect exposure via the Hang Seng Tech ETF. The combined position is meaningful but not yet sized to reflect full conviction — which is deliberate.

My current approach is to use cash-secured short puts to build toward the position gradually, targeting a lower entry point if the stock comes in to a price that reflects more of the execution risk now visible in the 2025 numbers. If management continues to execute as I expect — overseas volume growing, Hungary demonstrating viable localised economics, and cash flow recovering from what looks like a trough year — I would be comfortable growing the combined growth portfolio stake toward 5 percent over time.

The sizing logic is straightforward: the thesis is intact but unproven at the critical overseas step. A 2 percent position reflects a live hypothesis. A 5 percent position would reflect a thesis entering confirmation. I am not going to 5 percent until I can see the overseas economics in the numbers, not just in management commentary.

The mechanism — short puts rather than direct purchase — is not about being clever. It is about being paid to wait at a price I would genuinely be happy to own more stock at. If the puts expire worthless, I collect premium while the thesis develops. If I get assigned, I own more BYD at a price that builds a larger margin of safety into the position. Either outcome works for me.

If Hungary is approaching 100,000 units of annualised production before Q3 2026, or if BYD breaks out energy storage as a reportable segment, I would treat either as a signal to build the position faster.

What would make me reduce is more specific: overseas revenue share continuing to rise while operating cash flow stays depressed. Revenue mix shifting while cash does not follow is not a healthy transition. It is a sign the overseas business is absorbing costs faster than it is generating returns. At that point, patience stops being a thesis and starts being a habit.

What would make me believe more — and what would kill the thesis

I do not need BYD to look perfect in 2026 or 2027. I need the future path to become more legible.

In practical terms, I want to see three things in the numbers: overseas revenue per vehicle holding up as volume scales, at least one overseas plant moving from symbolic output to economically meaningful utilisation, and operating cash flow recovering without another stretch in supplier payables.

The kill condition is not that China stays difficult for another year. China probably will stay difficult.

The kill condition is that BYD keeps growing abroad and still fails to become more resilient — that overseas volume expands but overseas margins compress toward domestic levels as price competition follows the brand into new markets, or that localisation turns into a permanent cost burden rather than a moat. If that happens, the thesis is not early. It is wrong.

What I think now

The real question is not whether BYD can sell more cars abroad. It is whether it can become locally durable there.

The forward signals I am watching: the Szeged plant’s capacity utilisation in the Q3 2026 interim report — if it begins approaching 100,000 units annualised without a margin deterioration narrative attached, that is the first localisation proof point. And separately, whether BYD breaks out energy storage as a reportable segment in the 2026 interims. If it does, the platform thesis has enough commercial weight to require standalone disclosure. If it does not, I will continue to treat storage as optionality rather than underwriting it as a value driver.

A future Toyota is not just a company that can export everywhere. It is a company that can belong everywhere.

As of the date of publication, I hold positions in BYD Company (HKEX: 1211) and Hang Seng Tech ETF (Xetra: H4ZX). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

On 6 and 7 January, while most of the Western financial world was still shaking off the holidays, Beijing published 2 documents that quietly shifted the ground beneath four sectors I care deeply about: the AI + Manufacturing Special Action Implementation Opinions and the Industrial Internet and AI Integration Empowerment Action Plan.

I’m not writing about this because it’s “news.” I’m writing because this is one of the first times I’ve seen the 15th Five-Year Plan translate from aspirational language into specific ministerial KPIs with binding targets — and the implementation clock is now ticking. That distinction matters — because when eight ministries set binding KPIs with SASAC enforcing SOE compliance, the result is captive demand: revenue that exists because policy mandates it, not because customers chose it freely. That captive demand is now flowing into EVs, telcos, chips, and software. I want to be positioned in the sectors where Chinese firms are already competitive enough to turn mandated procurement into real earnings growth — and honest about where I’m not confident enough to buy.

The short version: I already held EVs and telcos. These documents sharpened my conviction that software and the sensor layer deserve direct exposure — so when the Iran sell-off gave me the price I wanted on Kingdee and Hesai, I executed. I did not add semis. The question running through the whole piece is whether captive demand accelerates dominance or subsidises mediocrity — and the answer depends entirely on the sector.

What Actually Changed

Beijing has moved from telling us what it wants to telling us exactly how much, by when, and who’s responsible for delivering it.

The numbers are concrete:

By 2027: 3–5 general-purpose large AI models deeply applied in manufacturing, 1,000 high-level industrial intelligent agents, 100 high-quality industrial datasets, 500 typical application scenarios, and 1,000 benchmark enterprises. By 2028: at least 50,000 enterprises implementing new industrial network transformation, with high-quality datasets across 20 key industries. — AI + Manufacturing Special Action Implementation Opinions and Industrial Internet and AI Integration Empowerment Action Plan, January 2026

These aren’t the kind of round numbers bureaucrats throw around when they’re being aspirational. 50,000 enterprises. 20 industries. 2027 and 2028 deadlines. This is execution-phase policy with ministerial accountability attached.

For investors in EV, telco, chips, and software, this is where captive demand stops being a concept and starts being an order book. When SASAC directs central SOEs to integrate AI into strategic planning and build intelligent computing centers, that’s not market-driven adoption — it’s mandated procurement with budget lines attached.

Why Now: The 15th FYP Implementation Phase

The 15th Five-Year Plan (2026-2030) was formally adopted in late 2025, but until these January documents, the guidance was directional — strategic priorities, aspirational language, the “China will become a leader in X” type of framing. Useful if you’re trying to figure out which way Beijing is leaning — less useful if you’re trying to size an order book.

These are implementation-level policy. The eight-ministry joint issuance of the AI + Manufacturing Opinions (MIIT, CAC, NDRC, Ministry of Education, MOFCOM, SASAC, SAMR, National Data Bureau) signals whole-of-government coordination. That’s not a single ministry floating a trial balloon—it’s cross-functional accountability with SASAC in the room enforcing SOE compliance and NDRC controlling investment approval. And SASAC doesn’t just coordinate — it controls executive appointments and compensation for state-owned enterprises. When a policy document says “integrate AI into strategic planning,” and the people saying it decide whether executives get promoted or fired, that’s not a suggestion. It’s a career incentive.

Add compliance mechanisms with teeth — data governance certifications, model safety assessments, subsidy eligibility tied to targets — and the enforcement structure has both carrot and stick.

Implementation-phase policy tends to be stickier. Once ministerial KPIs are set and budgets allocated, reversing course becomes bureaucratically expensive. I’m comfortable with multi-year exposure here because the commitment runs through 2030 — unwinding this is not just unlikely, it’s bureaucratically painful for every ministry that signed on.

But there’s a tension worth naming first. Mandated procurement creates revenue visibility — but it can also compress margins, misallocate capex, and produce political winners instead of commercial ones. The test I apply sector by sector below: are the domestic firms capturing this demand already globally competitive, or are they being shielded from the iteration pressure that makes products world-class? Where the answer is “already competitive,” captive demand is an accelerant. Where it isn’t, I want to see margin quality and receivables before committing capital.

Who Gets Paid

4 sectors sit in the direct procurement path: EVs, telco, chips, and software. Each captures captive demand differently — and each has a different way of punishing the wrong positioning.

Electric Vehicles

This policy creates two distinct EV value channels.

The first is factory efficiency. AI-powered production lines — real-time optimisation, predictive maintenance, defect detection — can cut EV assembly times by 15–30%. In a margin war measured in hundreds of RMB, that shows up in gross margins.

The second is autonomous driving — the policy supports L3/L4 with clearer regulatory pathways and calls for “AI chip hardware-software coordination.” XPeng’s CEO is forecasting a 2026 leap in autonomous driving capability from L2 to L4, and policy clarity is part of why. That is also part of why I added Hesai. Lidar is not optional for L3 in practice — new safety regulations already require redundant sensing, and Hesai’s nomination list of 160+ production programs across China’s top ten automakers reflects that reality. L4 certification requirements are still being formalised, but regulators are unlikely to accept a camera-only architecture.

The immediate order flow sits with smart factory suppliers — industrial robots, manufacturing software, machine vision — and leading EV makers who already invested get policy validation and subsidy access.

For my EV holdings, the signals I’m watching are specific to each name: export volumes for BYD and Geely (proof that manufacturing competitiveness travels), and licensing revenue for XPeng (proof that the autonomy stack has value beyond its own vehicles). If those numbers are growing, the factory-efficiency and autonomy theses are showing up where it matters — in revenue lines that don’t depend on domestic policy.

Telecommunications

The telco story is straightforward but underappreciated: 50,000+ enterprises need new or upgraded industrial networks, and state-owned carriers are the ones building them. MIIT is targeting 10,000 5G-powered factories by 2027, with 35%+ 5G penetration in large industrial enterprises.

The interesting part isn’t the capex — it’s the revenue mix. Consumer voice and SMS are in structural decline, but industrial internet services — factory connectivity, computing power leasing, AI platform hosting — are a fundamentally different business. The national computing power network positions the big three as infrastructure providers.

When SOEs are mandated to adopt AI + industrial internet and the carriers are themselves SOEs, the loop is circular — procurement mandates, not consumer adoption.

There’s a second reason: trust. Compliant AI deployment with audit trails and domestic hosting requires politically acceptable infrastructure — and the big three are already inside the gate. I explored why trusted rails may matter more than model performance in My One-Person-Company Bet: Who Gets Paid First in China?. China Unicom alone reported AI revenue growth of more than 147% year-on-year in 2025 — and these policy documents just handed the carriers a mandate-backed order book.

For my positions in China Mobile, China Unicom, and China Telecom, the question is whether mandated B2B revenue converts into sustainable margin — or whether it’s low-margin infrastructure buildout dressed up as a growth story. The signal I’m watching: revenue mix and segment margins. China Telecom now reports Industrial Digitalisation as a discrete revenue line; China Mobile breaks out computing and AI services. What I need to see is the resource-based, infrastructure side of that revenue growing faster than the project-based system integration work — and eventually, margin disclosure to confirm the quality is there, not just the top line.

Semiconductors

Chips are where this policy gets complicated — and where I have the most questions. The documents call for breakthroughs in “high-end training chips, edge inference chips, AI servers, high-speed interconnect,” require “lightweight computing modules” across production equipment and automated vehicles, and push for a “national integrated computing power network” requiring domestic chip supply to work around export controls.

Here’s the math: 50,000+ enterprise upgrades need industrial control chips and edge AI accelerators — procurement budgets, not aspirations. 10,000 5G factories need baseband and RF chips. The computing power network needs datacenter GPUs or domestic equivalents. That’s a multi-year order book for domestic foundries — structural, not cyclical, and distinct from consumer electronics. Whether the supply side can deliver is the harder question.

The policy phrase “safe and reliable supply” is doing a lot of work here.

安全可靠 — “safe and reliable” — is Beijing’s standard procurement euphemism for domestic substitution.

When SOEs are told to use domestic chips where available, that creates a protected market where domestic suppliers can achieve scale without competing head-to-head with TSMC. The funding is substantial: Big Fund III at ~US$47.5 billion, plus the National AI Industry Investment Fund and government subsidies.

There’s also support for chiplet architectures — where smaller chip components are combined rather than fabricated as a single piece — which lets Chinese firms work around older manufacturing nodes. I find this more interesting than the headline GPU race, because it’s a realistic path to “good enough” rather than a moonshot to match TSMC. And a focus on smaller AI models that reduces computing requirements extends the viable market for those nodes.

The performance bar for factory-floor inference is lower than datacenter training — domestic technology is good enough for what’s being asked of it. Captive demand plus “good enough” is how protected industries achieve scale — but this is where the tension bites hardest: a decade of Big Fund money and the gap with TSMC hasn’t closed. The policy creates the order book; whether it produces competitive products is what I’d need to see in margins and design wins before buying.

I don’t hold AI/compute semis. The domestic positioning risk gives me pause: fabless designers dependent on TSMC for advanced nodes face supply constraints policy can’t solve, and firms optimised for consumer chips may find themselves outside the tailwind. What would change my mind: a domestic edge-AI vendor showing improving gross margins and non-SOE design wins. One open question that could shift the picture: how the ~US$70 billion chip support package splits between datacenter AI accelerators and industrial/edge chips. If the bulk flows to edge, the order book for domestic foundries gets materially larger — and the “good enough” thesis strengthens. Until then, I’m watching.

Software

Every traditional enterprise software category — ERPs, manufacturing execution systems, product lifecycle tools — is about to go through a policy-mandated upgrade cycle. If your software doesn’t have AI features, you’re off the SOE vendor shortlist. That’s not a competitive dynamic — it’s a procurement filter, and it reshapes who captures the upgrade budget. That is one reason I added Kingdee. It reported RMB 356 million in AI contract value for FY2025, including SOE clients like Shenzhen Energy, with AI tools cutting bookkeeping time by over 80% and tax filing by 60%. Not a demo — a paid enterprise SaaS motion, and this policy just handed it a mandate-backed upgrade cycle.

The policy calls for vendors who are both “intelligent and industry-familiar” — 赋能应用服务商. Firms that bridge AI with domain knowledge in steel production or chemical processing are suddenly in demand.

Industrial internet platforms are positioning themselves as hosts for “model pools” and intelligent agents. This is part of why I hold Kingdee rather than a point-solution vendor — Kingdee already has the enterprise client base and the ERP infrastructure inside SOEs like Shenzhen Energy. When those clients deploy AI agents on top of an existing platform, switching costs compound with every agent added. A point-solution vendor has to win the client first; Kingdee is already inside the building.

And then there’s the compliance layer. Mandatory data governance certifications, Chief Data Officer requirements, data asset registration systems — all of this creates a software category that didn’t exist before: data cataloging, lineage tracking, governance platforms. Compliance software has reliable demand curves.

Foreign incumbents are under pressure. The “technology self-reliance” language is procurement guidance, not decoration. Data localisation complicates foreign cloud offerings, and domestic competitors are iterating faster — a market share risk that compounds.

Of the four sectors, software is where the accelerant dynamic looks clearest after EVs. The products are already usable, the switching costs are real, and the compliance layer adds recurring revenue that doesn’t depend on the next policy cycle. The signal I’m watching: whether Kingdee’s AI contract value converts into retained enterprise subscriptions — revenue that renews because the product works, not because SASAC said so.

How I’m Positioned

My approach and portfolio structure haven’t changed structurally. The Iran sell-off gave me the entry I wanted on Kingdee (workflow software, measurable labour compression) and Hesai (lidar on the L3/L4 certification path). Full exposure in and around this theme: XPeng, BYD, Geely, China Mobile, China Unicom, China Telecom, Kingdee, Hesai, and Innoscience. I also hold Pony AI, though it is less central here.

The common thread: every position I hold in this space is in a sector where I believe the accelerant dynamic applies — firms already competitive enough to convert mandated demand into real earnings growth. The positions that depend on China’s domestic policy for their thesis — telcos, Kingdee — are balanced by positions where China’s global competitiveness is already proven independent of mandate: BYD, Geely, XPeng’s licensing deals. If policy reverses, the latter group doesn’t need it. I may be wrong about that balance, and I’ll be testing it against the signals I’ve named above.

The Thing I Can’t Figure Out

I’ve been running the “accelerant or subsidy” test through every sector above — and the results are mixed. EVs and software pass; telcos are ambiguous; chips I’m not buying. The broader uncertainty remains: does the SOE mandate structure that makes this policy enforceable also limit the upside? Captive customers who must buy domestic regardless of price can distort market signals and slow the iteration pressure that produces world-class products.

I’ll be watching the execution data — order books, margins, and whether captive demand shows up in earnings with healthy receivables or just in revenue with deteriorating collection cycles. The investment case depends entirely on which dynamic you’re buying into — and being honest about which one you’re holding. The next test arrives with H1 2026 earnings: that’s when the 2027 KPI deadlines should start showing up in procurement pipelines. If the revenue is there with clean margins, the thesis is working. If it’s there with deteriorating receivables, I’ll need to rethink which sectors deserve the capital.

As of the date of publication, I hold positions in XPeng (HKEX: 9868), BYD Company (HKEX: 1211), Geely Automobile Holdings (HKEX: 0175), China Mobile (HKEX: 0941), China Unicom (HKEX: 0762), China Telecom (HKEX: 0728), Innoscience Technology (HKEX: 2577), Kingdee International Software Group (HKEX: 0268), Hesai Group (HKEX: 2525 / NASDAQ: HSAI), and Pony AI (NASDAQ: PONY / HKEX: 2026). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

A Thousand People Queued to Install OpenClaw

I own Tencent, Alibaba, three Chinese telcos, and Kingdee — and I own them because I believe the companies that already control daily habit, payments, and enterprise workflows are the ones that get paid when digital labour becomes routine. That is the bet — and everything in this piece tests whether I should be increasing it or reconsidering the sizing. A thousand people queuing at Tencent’s Shenzhen headquarters to install software that might do work on their behalf is either early evidence that the bet is right, or the kind of demand signal that flatters before it fades.

The Chinese internet gave it a name: “growing your lobster” — feeding a software agent tasks until it starts handling the workload for you. Tencent’s OpenClaw was the software, and those thousand people were not just turning up to play with a chatbot. They were testing whether software could take tasks off their plate.

Once Tencent began placing OpenClaw’s task-handling logic inside WeChat as a native contact rather than a separate destination, the story stopped looking like a consumer fad and started looking like an economic experiment — not curiosity, but delegation.

By one-person company, I do not mean an internet slogan but software that compresses administration, customer service, bookkeeping, and parts of sales support into a much smaller human footprint — enough that one operator can function a bit more like a small firm.

China has not yet proved that works at scale. OpenClaw matters because it makes the question legible. Once the behaviour is plausible, the investable question becomes who owns the rails it runs on. That is the same logic I use in No Pain to Begin With: How I Invest My Own Capital: structure first, then thesis.

Why China May Normalise Agentic Labour First

China may be the first country to make agentic labour feel ordinary because distribution, payments, and commerce are already concentrated inside a few domestic platforms.

A Taobao seller already lives inside Alibaba’s payments, logistics, and merchant services stack. A small business owner already coordinates through WeChat. The fewer new surfaces an agent has to cross, the faster the behaviour compounds into routine.

Policy reinforces the same pattern. Beijing’s 2026 Government Work Report explicitly called for “large-scale commercial application of AI agents”. That language does not guarantee durable economics. But it does lower the adoption friction — and in China that matters, because bureaucratic customers follow the signal. If auditability and data locality become binding constraints, the infrastructure spend tilts toward domestic rails.

Who Gets Paid First?

When I evaluate a technology shift, I start with who owns the surface where the behaviour has to happen. Models are commoditising fast — DeepSeek-V3 cost under six million dollars to train, and Chinese domestic API prices range from five cents to fifty-five cents per million tokens versus five to fifteen dollars for OpenAI. When the model layer compresses by more than 97 per cent, value does not vanish. It re-pools into whoever controls where the task ends. Four layers matter: the habit surface, the transaction bridge, the trust infrastructure, and the workflow data. Those are the rails. Here is where I think the early profit pool sits in China — and why I have my money there.

Nothing here proves the one-person company exists at scale. What it does show is that the rail layer is already getting paid for AI-adjacent activity. Tencent putting OpenClaw functionality inside WeChat suggests this is being treated as a habit-forming category, not a passing curiosity. Alibaba says 140 million users had “experienced” AI-driven shopping by end-February 2026, though “experienced” is doing a lot of work in that sentence.

I may be directionally right and still not make any money from it. The question is whether agentic behaviour improves monetisation inside businesses that already own distribution, trust, and transactions. If usage explodes, Tencent gets paid on activity inside a surface it already monetises. Alibaba earns across the full vertical — chips, cloud, transaction, take-rate. The telcos collect rent on trusted domestic hosting. Kingdee is the one with the most to prove: whether businesses really will pay on outcomes, not seats. That is the real test: does the behaviour show up in revenue, or does it just make existing software slightly better? It is the same question I asked in 50,000 Factories: What China's AI Mandate Means for My Portfolio: not whether policy sounds ambitious, but where the economics actually settle.

Tencent Owns the Habit Surface

WeChat already bundles communication, payments, mini-programmes, merchant touchpoints, service flows, and daily coordination into one behaviour stack. If agents become habitual, Tencent does not need to teach users a new ritual. It only needs to insert delegation into one that already exists. That is what ClawBot does: it appears as a contact in a user’s chat list, sitting between friends and family, handling tasks through the same interface people already use to coordinate their lives.

That makes Tencent more than a distribution winner. It means more agent activity inside WeChat can flow through revenue channels that already exist. The list is straightforward: AI-targeted advertising — Marketing Services revenue grew 17 per cent in Q4 2025, driven by closed-loop ads that convert inside Mini Programmes and Mini Shops rather than sending users elsewhere; e-commerce technology service fees on Mini Shops GMV already in the trillions of renminbi; enterprise subscriptions through WeCom’s 14 million business clients; and cloud compute, where Tencent reached operating profit at scale in 2025. I cannot yet isolate how much of the advertising growth is agent-driven versus ordinary targeting improvements. But the structural point is simpler: Tencent does not need a new business model. It needs agents to increase commercial intent inside a surface it already monetises.

Alibaba Owns the Commerce Bridge

Alibaba matters not because it can win an AI beauty contest but because it already sits where intent turns into transactions — and it owns more of the stack than almost anyone else in China: chips, cloud, models, and interface. If that vertical integration holds, Alibaba can produce tokens cheaper than most and monetise them across merchant tools, marketplace traffic, payments, and enterprise services.

A Taobao seller using agents to rewrite copy, compare suppliers, handle customer messages, prepare VAT paperwork, and adjust campaign spend — that is the one-person-company idea made concrete. Think about what that seller’s cost structure used to look like: a part-time assistant, a freight coordinator, and a tax filing service. Call it RMB 150,000 a year in overhead. If software compresses that to a RMB 20,000 annual subscription, the seller is not choosing software over labour. They are choosing freedom over headcount. Accio Work, launched in March 2026, extends the same logic to cross-border sellers — autonomous sourcing, supplier negotiations, and customs filings inside the same transaction stack. The seller who once needed a sourcing team, a freight forwarder, and a customs broker can now delegate parts of that workflow to software that already sits where the payments clear. That is the one-person company made concrete — and Alibaba is sitting inside every transaction. It is also the version that matters financially.

No other company in this thesis controls the full vertical: chips, cloud, models, orchestration, and now workflow tools that span domestic and cross-border commerce. Alibaba Cloud reported RMB 43.3 billion (~US$6.0 billion) in the December 2025 quarter, with AI-related product revenue posting its tenth consecutive quarter of triple-digit growth. The vertical integration matters for economics: each agent call that ends in a Taobao purchase generates revenue at multiple layers — inference fees on Alibaba Cloud, commerce take-rates on the transaction, and payment processing through Alipay. T-Head has shipped 470,000 chips — if Alibaba is running meaningful inference workloads on its own silicon, the cost per token should be meaningfully lower than renting compute from third parties. I don’t have a clean per-token number, but the directional logic holds. The Alibaba Token Hub binds models to applications across the ecosystem, creating switching costs even for customers running open-weight Qwen. That is why 140 million users completing their first agentic shopping by February is a transaction-linked metric, not a vanity number — it means agents are already driving activity inside the revenue stack.

Telcos May Be the Cleanest Investment Expression

The telco case matters because trusted deployment may end up being the bottleneck that decides where value pools. If an SOE or a bank is, in practice, going to run sensitive workloads only through a domestically auditable stack, that is not a soft factor — that is the gate. And the gate is a telco. If you need audit trails, domestic hosting, compliant deployment, and politically acceptable infrastructure, the trusted rail matters.

I do not own Chinese telcos because they are exciting. I own them because the case is getting harder to ignore: real cash yield, consistent profitability, and now a genuine AI tailwind. China Telecom’s AI revenue reached RMB 12.3 billion in 2025 — that is what it looks like when regulated enterprises pay for compliant domestic deployment. China Unicom reported AI revenue growth of more than 147 per cent year-on-year in 2025. OpenClaw does not create the telco case. It just makes the demand side easier to see.

Workflow Software Is Where the Thesis Becomes Measurable

This is where the thesis has to show up in somebody’s budget, not just in a launch deck. If AI is genuinely allowing businesses to operate with fewer human hands, the proof should appear first in the systems that handle finance, administration, procurement, tax, coding, and customer workflows — the places where labour savings can actually be measured.

That is why I keep coming back to Kingdee. It reported RMB 356 million (~US$49 million) of AI contract value for FY2025. That does not prove autonomous firms. What it does show is that businesses will already pay to remove repetitive but necessary work, especially when those tools sit on top of structured finance and customer data.

The more important question is pricing power. A RMB 1,000 software licence is still software. A RMB 10,000 workflow bill tied to measurable labour savings is something else entirely. If the customer is genuinely saving RMB 50,000 to RMB 100,000 of labour, outcome-priced software stops looking expensive and starts looking rational.

“For the small and micro enterprise market, Kingdee AI achieved bookkeeping efficiency improvements of over 80%, invoicing efficiency improvements of 40%, and tax filing efficiency improvements of 60%.” — Kingdee FY2025 annual results

None of this yet proves durable outcome-based pricing. But it is exactly the sort of evidence I want to see: software moving away from seat-based selling and closer to measurable labour compression.

The bear case that runs through all of these positions is not that model companies will displace the rails — in China, the leading models are open source (for now), so any platform can run them. The real risk is simpler: that none of this amounts to more than incrementally better software. The rails were already getting paid. If agentic AI just makes existing workflows slightly faster without creating a genuinely new category of economic actor, then I am paying for a theme, not a step change. I do not think that is where this lands, but I cannot prove it yet — which is why the retention and operating-proof tests in this piece exist.

What Would Move This from Interesting to Durable?

For me, this thesis becomes durable only if three things show up in the data: retention, monetisation, and operating proof.

On retention, the early signs are encouraging but adjacent to what I actually need. Tencent and Alibaba can each point to rising AI engagement — Qwen’s 300 million monthly active users, 200 million holiday orders, stronger cloud demand — but none of that is the same as repeat agent-led task completion inside real workflows. The proof I want is habitual delegation, not curiosity.

On monetisation, the question is whether anyone is paying for the behaviour to continue — not just for the software to exist. I want to see agent activity that drives measurable commercial outcomes and shows up as recurring revenue, not just in a launch deck.

The third test matters most. Can one human plus software actually function as a one-person company? That is where the thesis becomes an economic claim rather than a technology narrative — and why workflow software matters more to me than the flashiest consumer AI launch.

My conviction in Tencent, Alibaba, Chinese telcos, and Kingdee has increased — not because the evidence is conclusive, but because each earnings cycle narrows the gap between “interesting hypothesis” and “visible in somebody’s revenue.” The specific signal I am watching next: Alibaba’s Q1 FY2027 results in June, where repeat agentic purchase rates — not first-time usage, but whether those 140 million users come back — will tell me whether the commerce bridge is habitual or promotional. If it is habitual, the thesis graduates from inference to fact. If it is not, I will need to revisit the sizing.

That is also, frankly, how I am trying to operate my own research and capital allocation — with fewer intermediaries, more direct ownership of the process, and software compressing the overhead. The thesis is not just about companies. It is a template.

As of the date of publication, I hold positions in Tencent (HKEX: 0700), Alibaba (HKEX: 9988), China Mobile (HKEX: 0941), China Telecom (HKEX: 0728), China Unicom (HKEX: 0762), and Kingdee International Software Group (HKEX: 0268). Positions may change after publication without notice. Cohong Lane is a periodical publication made generally available to the public; this is disclosure of my positions, not a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

In April 2025, my portfolio was down six figures in a week and I felt nothing worth acting on. Not because I’ve transcended fear — I haven’t — but because I’d built the portfolio so that a 40% drawdown changes my mood, not my decisions. That distinction is an operational design principle, and it’s the foundation of everything I do with my own capital.

I spent most of my career in and around financial markets — most recently as CFO of a global family office. Every crisis I’ve sat through — and there have been several — looked terrifying in the moment and obvious in retrospect.

The trick is that it’s only obvious if you’ve done the homework beforehand and if your portfolio structure doesn’t force you to sell.

In 2024, I quit that world to become what I call the CIO of my own life — betting my livelihood on the ideas I write about on Cohong Lane. What follows is the operating system I use to allocate capital.

Why does structure matter?

Holding power is the one structural advantage an independent investor has over professional money managers. I don’t have clients calling to redeem. I don’t have a Sharpe ratio mandate forcing me to reduce volatility. I don’t have a career at risk if I’m down 30% for two quarters.

Professionals cannot buy during maximum fear because their clients won’t let them. I can. That’s the edge. But it only works if I’ve structured my portfolio so that drawdowns are an opportunity, not a crisis — and if I’ve trained myself to feel no pain when the red numbers pile up.

None of this comes naturally. I’m still learning. A mentor told me it takes about five major drawdowns before your subconscious mind stops reacting. The first one is torture. By the fifth, it barely registers. The work is in internalising, truly internalising, that price does not hurt. Only value does. If I own a company generating reliable cash flows at a price I know is cheap, a 30% paper loss doesn’t change anything about that company. It just means I get to buy more at an even better price.

How is the portfolio structured?

My portfolio is built around two core buckets — Income and Growth — with a small, emerging allocation to Venture bets. The assets span multiple geographies — Europe, the US, Singapore, Greater China — but this newsletter focuses on where I do the majority of my own research: China and the companies shaped by its industrial policy. The separation is about cognitive clarity and eliminating forced selling.

Income exists for one reason: to generate enough cash flow that I never have to sell anything I don’t want to sell. If my dividend and coupon income covers daily baseline expenses, every other decision becomes optional. I can hold Growth positions through multi-year drawdowns without being a forced seller. That changes everything.

The threshold is simple: minimum 5% net of withholding tax. If an income asset doesn’t clear that bar after tax drag, it doesn’t earn its place. Beyond that, yield must be real and sustainable, tiered by risk — from European and US corporate bonds to Singapore REITs and dividend-paying Chinese state-owned enterprises. When high-quality income assets sell off alongside lower-quality names during dislocations, I rotate up the quality ladder. Once this bucket is set up, I spend maybe 20% of my investment time here. The real intellectual work happens elsewhere.

Growth is where I concentrate on my highest-conviction opportunities — businesses I understand deeply, where I believe the market is materially mispricing the long-term trajectory. I hold growth positions across Asia and the US, but my concentrated bets are in Greater China. Concentration, not diversification. Every Growth position is underwritten for a minimum five-year horizon. If the thesis doesn’t hold over that timeframe, the position doesn’t belong here. I don’t want to own 50 growth stocks. I want a small number of companies I understand well enough to hold when the price is moving against me — because the edge isn’t information, it’s the ability to sit through the discomfort.

Valuation still matters here. Growth investing isn’t about paying any price for a good story. I model fair value ranges and size positions based on the gap between current price and my estimate. Even the best business is a bad investment at the wrong price. And if my thesis is about where a company will be in five years, I need to be prepared to hold for five years — including through substantial drawdowns along the way.

There’s also a third bucket: Venture. These are small bets on things that could be worth a lot or nothing — new technologies, new competitive dynamics, situations where the range of outcomes is very wide. I size every one for total loss. That’s the price of entry — and it means the only positions worth taking are ones where I can see a credible path to 5–10x over the investment horizon. If the asymmetry isn’t there, the bet doesn’t belong here. It’s deliberately small, deliberately early.

What does fair value actually mean?

Every investment decision I make ends with the same question: what is this asset worth?

I don’t calculate before I understand. First comes the industry analysis, the competitive position, the technology edge, the operational reality. Only then do I have enough context to estimate fair value. The number is the final checkpoint, not the starting gate.

For income assets, that means estimating a fair yield given the risk. For growth assets, it means building cash flow models with explicit assumptions I can stress-test. The output is always a range, not a single number — because I’m not that precise and neither is reality.

The point of doing this work is not the number itself. It’s having a pre-built shopping list ready before the crash arrives. When prices collapse and everyone is panicking, I don’t want to be making emotional decisions. I want to be executing against fair value estimates I’ve already pressure-tested in calm weather. The time to decide what to buy in a crisis is not during the crisis.

One discipline worth highlighting because it trips up nearly everyone: my P&L on a position has zero relevance to what I should do next. If I’m sitting on a 50% loss, that’s irrelevant. The only question is whether, at the current price and my current fair value estimate, I should buy more, hold, or sell. The past is sunk. Acting on it is the single most common mistake I see investors make — and it applies equally to gains. If a position has doubled and is now at or above fair value, the fact that I’m “up 100%” doesn’t mean I should hold. It means I should evaluate it as if I were buying fresh today.

The obvious counterargument is that holding power without external accountability is just stubbornness with better branding. Professional managers have risk committees for a reason — someone to tell them their thesis is wrong before the P&L does. I don’t have that. What I have instead is the fair value framework and the discipline to update it. If the business fundamentals deteriorate — not the price, the fundamentals — that shows up in the model and changes the decision. The check on stubbornness isn’t a committee. It’s the homework.

Why does this matter for China?

China’s market is shaped by forces that most investment frameworks don’t account for — and that’s exactly why I’m writing from Hong Kong instead of London. I’ve been investing my own capital in Chinese equities since 2018, and the longer I do it, the more I’m convinced that industrial policy isn’t background noise here. It’s a primary driver of capital allocation at massive scale. When Beijing sets binding targets for factory automation or EV adoption, that creates something I think of as captive demand: procurement that happens regardless of market sentiment. The money gets spent because a provincial official’s career depends on hitting the KPI — and that policy commitment creates demand visibility that market-driven economies simply can’t match. Understanding the transmission mechanism from a Five-Year Plan target to a company’s order book is where much of my Growth research begins. When the 15th Five-Year Plan set binding factory automation targets, the procurement pipeline for companies in that chain was effectively locked in for years — regardless of what the macro headlines said. Tracing those mandates to specific order books is the work I publish on this site.

Then there’s the information problem. Western mainstream media defaults to “China is in structural decline.” Local media defaults to “China’s rise is inevitable.” The possibility that reality might be somewhere in between — messy, sector-specific, and actually quite interesting if you look at the numbers — that’s somehow never on the menu. Sell-side research is conflicted by design — banks won’t publish contrarian China views when IPO mandates are at stake. So you end up with a dangerous context gap: investors have data but lack the nuance to distinguish structural risk from cyclical opportunity.

That’s why I’m here. Not because walking into a XPeng showroom in Shenzhen gives me statistically significant data — it doesn’t. My sample of one showroom visit is not going to hold up in a peer review. But it gives me calibration. When I read a bearish report claiming Chinese consumers have stopped spending, I can check that against what I see in Shenzhen malls on a Saturday. It’s not proof. It’s a filter for obvious nonsense — and in China coverage, there is a lot of obvious nonsense to filter.

The combination of holding power, fair value discipline, and physical presence in the Greater Bay Area is the operating system. The homework I share on Cohong Lane — the policy analysis, the valuation work, the ground-truth checks — is how I keep that system honest.

But the homework only works if the structure underneath it never forces me to sell. Income buys the freedom. Fair value is the compass. And the goal — the one I keep coming back to — is not to manage the pain of a drawdown, but to have built something where the pain doesn't arrive in the first place. That's the operating system. The positions are just what it produces — and that's what I'll be writing about next. How the income bucket is constructed to clear 5% net. Which growth positions I'm holding through a five-year lens and why. And the live question I can't yet answer: whether this structure holds when the drawdown is not 40% but 60%. That's the test I haven't sat through yet.

Cohong Lane is a periodical publication made generally available to the public. Nothing here is a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.

Who were the Cohong?

There’s a painting I keep coming back to. It shows a narrow strip of waterfront on the Pearl River, sometime around 1800. Foreign flags — British, American, Dutch — flutter above a row of warehouses called the Thirteen Factories. And standing between those factories and the rest of China were a dozen merchants who controlled every bolt of silk, every chest of tea, and every ounce of silver that moved between East and West.

They were called the Cohong.

The Cohong (gōngháng, 公行) held an official monopoly on foreign trade in Canton from 1757 to 1842. If you wanted to do business with China during the Qing dynasty, you went through them. There was no alternative. Historian Jacques Downs called their world a “golden ghetto” — isolated and lucrative. But the Cohong merchants weren’t just gatekeepers. They were guarantors. Every foreign ship, every person connected with that vessel — the Cohong took full responsibility. Their reputation was their licence.

The most famous of them, Howqua, became one of the richest men in the world — and he got there not through political favour but through what the American consul Samuel Shaw called being an “exact accountant” who valued his “fair character”. Not the colonialism that eventually destroyed the system, but the intellectual integrity that preceded it. That’s the tradition I want to honour.

Why am I here?

In 2024, I quit my job as CFO of a global family office. I’d spent 25 years in and around institutional finance — close enough to the allocation decisions to see how real money moves, and close enough to know that the smartest capital in the room is often the most constrained. In 2025, I moved to Hong Kong — the capital markets gateway to China’s Greater Bay Area — to test a hypothesis: that one serious person, doing the work on the ground with their own money at risk, can sometimes see more clearly than a conflicted institutional machine.

If you’re wondering whether that was a rational decision: so am I, most mornings.

My edge, such as it is, lies in being here — in the same Pearl River Delta where the Cohong operated. Physical proximity doesn’t guarantee insight, but it provides something no desk-based analyst has: the ability to check headlines against reality on the ground. Walking into a BYD showroom in Shenzhen doesn’t give me statistically significant data — my sample of one is not going to hold up in a peer review. But it gives me calibration.

When I read a bearish report on Chinese consumer spending, I can check the claim against what I see on the ground — and the picture is almost always more nuanced, more sector-specific, and more investable than the headline. The filter doesn’t replace the data. It tells me which questions to ask next — and in China coverage, knowing the right question is half the work.

That’s the job I’m trying to do — except instead of a warehouse full of tea, my tools are public filings, policy documents, dealership visits, and the kind of conversations you only have when you’re physically here.

What You’ll Find Here

Cohong Lane is the transparent research process of an independent investor whose bills get paid by the markets. My livelihood depends on getting these investment decisions right. This isn’t a newsletter that happens to mention investing — I call it open-sourcing the workflow.

What I publish falls into three areas. Portfolio & Process — how I structure and manage my own capital, with every position, every sizing choice, and every mistake in the open. You can watch each thesis play out in real time. Investor’s Mindset — the psychology, frameworks, and mental models behind investing independently, because the hardest part of this job isn’t the analysis — it’s the discipline. China Mechanics — tracing how a government mandate becomes a line on a company’s P&L, verified on the ground through dealership visits, city walks, and the kind of checks you can only do from the Greater Bay Area.

If you manage your own money — or you’re building towards it — and you’d rather understand how a serious independent investor actually operates than read another set of tips or hot takes, Cohong Lane might be for you. I don’t have all the answers. But I do the work, I put real money behind it, and I share the whole process so we can figure it out together.

The Cohong merchants earned their monopoly one honest ledger at a time. I don’t have a monopoly on anything — but I do have my own capital on the line and no reason to tell you anything other than what I actually think. That’s the licence I’m betting on.

Next, I walk through the operating system behind my portfolio — two core buckets, a fair value compass, and why I structured everything so that a 40% drawdown changes my mood, not my decisions: No Pain to Begin With.

Cohong Lane is a periodical publication made generally available to the public. Nothing here is a recommendation to buy, sell, or hold any securities. Full disclaimer · About Philip.