Five years ago I bought an ETF called Hang Seng Tech, as my bet on China becoming a dominant force in technology. To date, it has been a drag on my growth sleeve.
The index’s three largest holdings today are a phone maker, a shopping site and the company that owns WeChat. This year’s technology trade has been chips, memory, servers, and the electricity to run them. Almost all of it is listed somewhere else: New York, Taipei, Shanghai, Shenzhen.
In June I wrote that I owned the wrong index, not the wrong businesses, and this was the China position I came closest to cutting. Then, on 10 August, Hang Seng Indexes published a consultation paper making the same case against its own rules: they were missing China’s faster-growing technology companies.
In its own language:
Within the Technology Universe, companies with stronger sales growth are often smaller by market capitalisation... This suggests that higher-growth companies with relatively smaller size may not be considered under a purely size-based selection.
Hang Seng Indexes is saying that its size test misses the fastest-growing companies, so it will stop relying on that test alone. Index providers do not often admit to being wrong. In consultation-paper language, “may not be considered under a purely size-based selection” is a shout.
What it published is a consultation, not a new rulebook. Responses close on 18 September, and the advisory committee can amend or reject any part of it. My own weighting is roughly 60 per cent that it goes through substantially as drafted, 30 per cent that it comes back materially amended, 10 per cent that it is watered down or slips past December.
I still own it. The paper changed my decision.
Why the index became the wrong wrapper
Standard Chartered’s wealth research put numbers on my complaint: the Hang Seng Tech index is roughly half internet and platform companies, about 15 per cent electric vehicles, and under a fifth upstream AI hardware and semiconductors. Mainland financial writers have a blunt name for that gap: the “delivery index” (外卖指数). The label stuck while its largest internet holdings were pulled into the food-delivery price war, and the index I owned missed the AI rally. In August, wallstreetcn and Securities Times described it as beginning to shed the name.
Hang Seng Indexes, the index arm of Hang Seng Bank, launched the Hang Seng Tech index on 27 July 2020, and the timing explains the design. Hong Kong was in the middle of a technology listing boom, mainland champions were arriving in numbers, and there was no simple way to buy the lot. So the compiler built one: take the 30 largest, put them in one basket, and let funds track it.
To qualify, a company had to sit in one of five industry classifications and pass one of three innovation tests: R&D worth at least 5 per cent of sales, revenue growth of at least 10 per cent, or a business built on a technology platform. The third is the generous one; a shopping site qualifies, and did. Everything that cleared was ranked by market value, the 30 biggest got in, and no single name was allowed more than 8 per cent of the index.
Inclusion came down to how big the company was. That is a defensible rule. It is also a backward-looking one, because a company’s market value is a record of what it has already become.
I bought the Hang Seng Tech ETF for the growth sleeve of my China book, where it sits alongside a CSI 300 ETF tracking mainland stocks. The individual Hong Kong names I own are a separate matter: they sit in my income sleeve. A fund was the deliberate choice. I wanted the whole complex rather than my best three guesses inside it, and I accepted going in that a rotten apple in a basket is not mine to remove.
The companies at the top of the index then went through a hard few years: regulation, a slowing Chinese consumer, and a price war in food delivery and ride-hailing that ate into profits at the names carrying the most weight. Alibaba is the obvious case. Meanwhile the cycle moved somewhere the index could not reach. The rally of the past two years has been about the hardware behind AI, and an index confined to Hong Kong listings and ranked by size was never going to catch much of it.
The index fell roughly 22 per cent in the first half of 2026, and I owned it while it fell. Over the same stretch the mainland boards ran the other way: the STAR 50 and ChiNext each rose more than a quarter, and the CSI 300, which I also own, was up too.
The compiler had already begun patching by hand: at the June review it dropped Kingdee and Kingsoft and added MiniMax and Zhipu, two AI model developers that listed here only in January. An index committee reaching for names its own rules would not have delivered is a signal about the rules. By June I was not asking whether the businesses were good. I was asking whether the wrapper was still worth owning.
The repair: 30 names become 50, through two doors
What Hang Seng Indexes has proposed is the largest change to the index since it launched. The index goes from 30 companies to 50, chosen through two doors instead of one. Forty seats go to the largest eligible companies by market value, as they do now. The remaining 10 go to the fastest revenue growers among eligible companies that missed the first cut.
The five existing industry classifications are scrapped on the reasoning that “technological endeavors extend beyond the traditional industry classifications”. Artificial intelligence is promoted from a sub-category to a theme of its own. And the pool of candidates narrows from every Main Board listing to the larger and mid-sized members of the Hang Seng Composite. The money tracking the Hang Seng Tech index has grown from about US$1.5 billion at launch to US$40.4 billion. Some of that is mine.
On the compiler’s own simulation, Advanced Hardware companies go from 5 to 15 and AI from 3 to 6. The top ten holdings fall from 70.6 per cent of the index to 66.3 per cent, and the median company shrinks from HK$118 billion to HK$70 billion. On those numbers the fund I own would hold three times the advanced-hardware companies and twice the AI companies it holds today.
Huatai Securities and CITIC have published simulated constituent lists built on the June data, and a multi-broker roundup in the mainland press converges on similar names. Every list is still an estimate.
CATL, the world’s largest battery maker, turns up in most of the market-value lists, which tells you how far the word tech is being stretched. Biren and Tianshu Zhixin, two young Chinese chip designers, turn up alongside it. My favourite detail is Kingdee, thrown out in June and simulated by Huatai to return in December. I own Kingdee outright, as an enterprise SaaS and AI play, and did not need the index’s permission to hold it, but note what the round trip means: Kingdee did not change between June and December; the committee’s rules did. That is how much of this list is made in a meeting room rather than in commerce.
One buried detail gives the compiler away a second time. At each quarterly review, an incumbent is not thrown out the moment it slips past the cut-off rank; it has to fall well past: 20 per cent past for the size names, 50 per cent for the growth names, while new names only get in by ranking clearly inside. The wider band for the growth names is the compiler admitting, in its own structure, that it expects them to swing.
One calendar point: the September review still runs the current 30-name rules. December is the first rebalancing that could use the new rules.
Taking on the Nasdaq, in ambition only
The obvious comparison is the Nasdaq 100, and it is half right.
The Nasdaq 100 holds the hundred largest non-financial companies listed on that exchange, ranked by size, with no thematic test. That is why it includes Costco. It lets winners run: Nvidia sits at roughly 8.7 per cent of the index. There is no hard cap on any single name; only when the companies above 4.5 per cent add up to more than 48 per cent of the index does Nasdaq trim them back.
Hang Seng Tech is reaching for the same breadth: 50 names, no industry test and AI as its own theme. It keeps the 8 per cent ceiling on any one company, 4 per cent for foreign companies, and still sorts the world into two dozen hand-written sub-themes. Caps like these are written by people who have watched a top holding go wrong. What December would deliver is a bigger, better-aimed curated basket.
What would make me wrong
The first problem is the venue. The index still picks only from Hong Kong listings, so Mainland AI chip designers, Cambricon and Hygon among them, stay outside unless they list here too. I found no dated statement from either company about Hong Kong plans. CXMT, China’s leading memory-chip maker, raised 57.9 billion yuan (about US$8.6 billion) on Shanghai’s STAR Market in July, Asia’s biggest IPO this year, and the shares rose 466 per cent on their debut. It listed in Shanghai. The reform is a listing incentive, not a geographic expansion.
The second problem is the size of the growth side. Those 10 fast-growing names carry 2.4 per cent of the simulated index between them, with a median market value of HK$29.1 billion. At 2.4 per cent, they do not yet change the index enough: it could keep 40 names and almost nothing would change. The real work is done by the 10 extra large companies and the change in theme mix. A trailing-year revenue screen will also favour recent listings off low bases, and rapid growth says nothing about the quality of the business, an objection I take seriously.
What I decided
I was close to giving up on Hang Seng Tech. The paper moved me from leaning-sell to hold. I am giving the index provider a second chance.
I am not adding. The reform changes what sits inside the basket, not whether the basket is cheap, and I was fully allocated before the proposed reshuffle arrived.
Three clocks now run. September gives the final rulebook. December gives the first rebalancing under it. Next year gives the real test: does the rebuilt index move when AI hardware moves? If not, it failed at the thing it was written to fix, and I re-underwrite. The question I cannot answer yet is whether the venue problem is ever solved; a Cambricon or Hygon listing here would be the first sign.
If the split, the scrapped industry test, and the caps survive the committee, the wrong index got righter, and I hold. If they do not, I will say so before December.
If you got something out of this, pass it to one person who owns a tech index and has never checked what is inside it. Word of mouth is how this publication grows.
As of the date of publication, I hold positions in Alibaba, Tencent, a Hang Seng Tech index ETF, a Nasdaq 100 index ETF, a CSI 300 index ETF, and Kingdee. 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.