Why the AI Rally in Asia Hides a Concentration Trap That Few Are Naming
Since the end of 2022, Asian markets have undergone a quiet but profound reconfiguration. The emergence of generative artificial intelligence not only transformed the narrative of global markets, but also reordered the specific weight of regional indices around a handful of names. Three companies — Taiwan Semiconductor Manufacturing Company, Samsung Electronics, and SK hynix — came to account for more than half of the returns of the FTSE Asia ex-Japan index. That is not market leadership. It is structural dependency disguised as momentum.
HSBC recently published a strategy note that puts its finger directly on that wound. The bank identified ten companies it calls "forgotten gems": businesses with high return on equity, growing market share, sustained profitability, and consistent dividends. Names that, according to the bank, are being overshadowed not by weak fundamentals, but by the noise of a market pointing in a single direction. The list includes the operator of the Hong Kong Stock Exchange, Korean food manufacturer Samyang Foods, Indonesian telecommunications operator PT Telkom, automotive glass producer Fuyao Glass Industry Group, pharmaceutical contract research and manufacturing organisation WuXi AppTec, and Indian real estate developer Godrej Properties, among others.
The argument is not that AI is going to collapse. The argument is more precise: when all capital points toward the same vector, the valuations of the rest of the market compress through abandonment, not through real deterioration. And that creates, for those who know how to read the structure, re-entry opportunities with a more favourable risk-return relationship than the headlines suggest.
The Mechanics of Concentration and Its Hidden Cost
There is a difference between a market that rises broadly and a market that rises because three assets are dragging it upward. The latter is far more fragile — not because the three assets are bad businesses, but because their weight in the index becomes a systemic liability for any manager who does not want to deviate too far from the benchmark.
TSMC, Samsung, and SK hynix are each, for distinct reasons, businesses with real structural advantages. TSMC dominates the most advanced nodes in semiconductor manufacturing. Samsung combines scale in memory with foundry capability. SK hynix leads in high-bandwidth memory, the type of chip that directly powers the training clusters for large-scale AI models. None of that is fantasy. But when these three companies concentrate more than half of the returns of a regional index, the market begins to operate on a feedback logic where the price rises because the price rises — not because the gap between value and price justifies it.
HSBC names it with precision: "Everyone holds the same stocks." That phrase, spoken by a global financial institution, is not rhetoric. It is an operational signal. When active and passive managers converge on the same positions, the market's capacity to absorb a reversal is reduced. It does not require the AI thesis to be wrong for the valuations of these three assets to adjust; it is sufficient for the speed of capital flow toward them to diminish.
The hidden cost of that concentration is twofold. On one hand, investors who buy the index are, without realising it, betting more than half of their Asian exposure on a single sectoral theme. On the other hand, businesses that do not belong to that theme — but which have solid fundamentals — accumulate discounts not due to weakness but due to invisibility. That is precisely what HSBC is signalling.
What the List Reveals When Broken Down by Its Parts
The six companies the bank names explicitly are not homogeneous. They are businesses with very different architectures, which makes the list more interesting than a simple collection of "cheap stocks."
Fuyao Glass Industry Group holds approximately 70% of the Chinese automotive glass market and has manufacturing presence in the United States, which gives it geographical coverage in a context where global automotive supply chains are being redesigned. HSBC notes that the market is undervaluing its growth margin and its margin resilience. The thesis is not that Fuyao will grow like a software company. It is that analysts are not updating their models to reflect how the geopolitical fragmentation of automotive manufacturing plays in its favour: having installed capacity in multiple geographies is no longer a luxury — it is a competitive advantage that takes years to replicate.
WuXi AppTec presents a different dynamic. As a contract research, development, and manufacturing organisation in the pharmaceutical sector, it operates in a market where the demand for outsourcing grows in a structural manner. Its manufacturing segment grew 11% in 2025, and the company guided for growth of between 18% and 22% in 2026 for its continuing operations. HSBC projects that this pace can be sustained for two or three more years, supported by capacity expansion in Singapore, the European Union, and the United States. This is a business that is deliberately building geographical redundancy, which reduces its exposure to regulatory shocks concentrated in a single territory.
Godrej Properties operates at the opposite end of the sectoral spectrum: residential real estate development in India. The bank acknowledges that sector shares have faced pressure due to moderation in market appetite, but distinguishes an important nuance: demand in the premium segment remains firm. Godrej has national presence in India, which is unusual among developers, and a balance sheet with sufficient depth to sustain large-scale projects. HSBC expects the projected deliveries to translate into reported profitability and solid cash generation. It is not a high-explosive-growth business; it is a business where execution discipline generates structural advantage over more fragile competitors.
Samyang Foods, PT Telkom Indonesia, and the Hong Kong Stock Exchange complete the picture with distinct profiles but a common denominator: predictable cash flows, market positions with real moats, and dividends that make them total-return assets — not merely price-appreciation plays. PT Telkom has EBITDA margins above 45%, backed by being the incumbent operator in the most populous market in Southeast Asia. HKEX is, in practice, a market infrastructure with a regulated monopoly structure. Samyang Foods builds its growth on exports of strongly branded products with proven demand in international markets.
The Concession That Is Implicit in This Analysis
There is a moment in HSBC's analysis that deserves more attention than it normally receives. When the bank says that the focus on AI is "generating market dislocations and, in some cases, drawing attention away from other growth themes," it is describing something more than a pricing phenomenon. It is describing a structure of incentives that pushes managers toward the same assets because not doing so carries a career risk, not just a financial one.
An active manager who underweighted TSMC over the past two years likely had to explain their underperformance versus the benchmark in every quarterly review. That pressure generates its own dynamic: buying what has already risen because not buying it is riskier for the manager's career than for the client's portfolio. That mechanism is what keeps concentrated rallies alive longer than the fundamentals would justify in isolation, and it is also what makes the eventual rotation more abrupt when it arrives.
HSBC's list is not betting that AI collapses. It is betting on something more subtle: that there are businesses with high return on equity, real cash generation, and defensible market positions that trade at a discount because the market's attention is directed elsewhere. That is a structurally different proposition from saying that AI is a bubble.
The implicit concession in that thesis also matters. By selecting these companies, HSBC is explicitly forgoing the short-term upside that would come from maintaining greater exposure to the three major beneficiaries of AI. That concession has internal logic: the bank is not saying that TSMC, Samsung, or SK hynix are bad investments. It is saying that their weight is already embedded in everyone's portfolios, and that the room for multiple expansion from here is more limited than in the names that nobody is looking at.
What Distinguishes a Structural Opportunity from a Tactical Rebound
The distinction that matters for any investor who considers this argument is the following: there is a difference between buying an asset because it is temporarily cheap and buying an asset because its business mechanics generate value in a sustained manner but the market is not reading it that way. The first case is a tactical timing trade. The second is a structural position with a longer horizon.
The companies on HSBC's list appear to fit better into the second category, at least for those the bank describes with sufficient detail. Fuyao Glass is not cheap because there was indiscriminate selling this week. It is undervalued, according to the bank's argument, because analysts are not updating their models to incorporate the value of having geographically distributed manufacturing in a world where that is increasingly scarce. WuXi AppTec does not have a growth problem; it has a regulatory risk perception problem that is probably over-discounted in the price given its active international expansion profile.
Godrej Properties is not a bet that the Indian real estate market suddenly overheats. It is a bet that the consolidation of the sector toward developers with solid balance sheets, national presence, and the capacity to execute large projects favours the best-positioned operators when the general market moderation normalises. Those are long-term patterns that do not depend on the cycle turning this week.
What makes HSBC's argument more robust is not the list of names in itself, but the diagnosis that precedes it. When more than half of the returns of a regional index are concentrated in three assets from the same sectoral theme, the excess of undistributed exposure in the rest of the index leaves compressed valuations in businesses that, on their own merits, would not deserve that discount. That is the mechanic that generates the opportunity. And that mechanic persists as long as capital continues flowing toward the same destination.
The architecture of the AI rally in Asia is not broken. But its narrowness is a measurable fact, not an interpretation. And the businesses that accumulate discounts through abandonment — not through deterioration — are precisely those that tend to offer the best entry point when the flow of capital begins, gradually, to seek breadth.
