NextFin

Emerging-Market Stocks Fall as AI Concerns Expose Index Concentration

Summarized by NextFin AI
  • Emerging-market diversification is increasingly compromised by semiconductor and AI-infrastructure concentration, with TSMC, Samsung Electronics, and SK Hynix comprising over 28% of the MSCI EM Index.
  • The Nov. 16 selloff was primarily a cyclical, liquidity-driven reaction ahead of Nvidia’s earnings, but it exposed a structural vulnerability linking benchmark performance to the global AI capital-spending cycle.
  • Equity weakness can spread through currencies and capital flows, particularly in Taiwan and South Korea, while global funds may trigger broader emerging-market selling through hedging, redemptions, and risk-limit reductions.
  • The outlook depends on whether AI demand remains robust: improving orders, pricing, inventories, and guidance could reverse the correction, while weaker earnings and rising inventories would confirm deeper structural repricing risks.

NextFin News - The emerging-market selloff is exposing a contradiction that investors have treated as diversification: the MSCI Emerging Markets Index now carries a large semiconductor and artificial-intelligence infrastructure bet, so doubts about the return on AI spending are travelling through the same benchmark that was supposed to spread risk. The index fell as much as 2.8% on Nov. 16, while Taiwan’s benchmark lost 6% and South Korea’s KOSPI fell more than 3% as investors reduced technology exposure ahead of Nvidia’s earnings.

The immediate question is whether this is a short-lived positioning shock or the beginning of a deeper repricing of the growth engine behind North Asian equities. The evidence points to both, but not in equal measure: the day’s move is cyclical and liquidity-driven, while the vulnerability it revealed is structural. Emerging markets are not becoming a single AI trade. Their benchmark, however, has become sufficiently concentrated in a few AI beneficiaries that a technology correction can masquerade as a broad emerging-market crisis.

The market reaction was not evenly distributed. Taiwan Semiconductor Manufacturing, Samsung Electronics and SK Hynix occupy three of the largest positions in the official MSCI EM index. As of Oct. 30, TSMC represented 15.87% of the benchmark, Samsung 6.99% and SK Hynix 5.20%. Together they accounted for more than 28% of the index before the selloff, and all three are tied to the global semiconductor investment cycle. The consequence is mechanical: when investors question data-center demand, chip pricing or the pace of capital spending, they do not merely sell a U.S. technology basket. They also sell a substantial slice of the emerging-market benchmark.

That concentration matters because the index is otherwise broad. MSCI says the benchmark contains 1,165 large- and mid-cap companies and captures approximately 85% of each country’s free-float-adjusted market capitalization. Its $11.28 trillion index market capitalization gives the move a global footprint, but the headline breadth can conceal narrow return drivers. The broad universe and narrow leadership now coexist.

The largest exposures are not identical businesses. TSMC is a contract manufacturer, Samsung spans memory and consumer electronics, and SK Hynix is heavily exposed to memory demand. That difference should ordinarily diversify company-specific risk. It does not eliminate the common factor: each company’s earnings are sensitive to the pace of global technology investment and to the willingness of customers to keep funding AI capacity. The market is therefore not simply pricing three firms; it is pricing the durability of a capital-spending chain.

The currency reaction follows a second channel. Technology-heavy economies such as Taiwan and South Korea are sensitive to foreign portfolio flows, and the same investor who reduces semiconductor exposure can hedge or sell the local currency. A risk-off move therefore hits equities first and then tightens financial conditions through the exchange rate. The dollar does not need to surge for that process to occur; the marginal buyer simply needs to demand more compensation for holding assets whose earnings are tied to a single global investment cycle.

The shock also arrives after a period in which the AI narrative had become a central support for emerging-market returns. Goldman Sachs Research said earnings per share in the region’s technology hardware and semiconductor sectors were expected to increase 37% in 2026, compared with nearly 15% growth expected for internet, media and entertainment companies. Those estimates explain why investors were willing to pay for the region’s upstream position in the AI supply chain. They also explain the asymmetry now appearing in prices: a small reduction in confidence about future spending can overwhelm a large amount of already-recognized earnings growth.

The valuation backdrop makes the reaction more complicated, not less. MSCI lists the emerging-market index at a forward price-to-earnings ratio of 8.70 and a price-to-book ratio of 2.20 as of Oct. 30. Those levels are not evidence that the index cannot fall. They do show that the market is not applying the same aggregate multiple to all emerging-market assets. Investors can simultaneously view the index as inexpensive relative to developed-market technology and still decide that the highest-growth companies deserve a lower multiple if the capital cycle looks less reliable.

That is the setup for the deeper question. If the AI trade is merely taking a pause, the selloff can reverse when earnings validate demand. If the market is discovering that index-level “emerging markets” means a concentrated bet on a few hardware champions, then diversification will not protect benchmark investors from the next AI shock.

The First-Order Shock Is Cyclical, but the Exposure Is Structural

The day’s decline is best classified as a cyclical risk-off event. It was triggered by positioning ahead of a known corporate catalyst, not by a new law, supply embargo or confirmed collapse in semiconductor demand. Investors are waiting for earnings evidence, and that makes the first-order move mean-reverting in principle. A credible outlook from Nvidia or its customers could restore confidence quickly; a weak outlook could extend the drawdown. Either outcome is consistent with a cyclical market reaction.

History supports caution against declaring a regime change after one session. The MSCI EM Index posted annual losses of 20.09% in 2022 and 14.57% in 2018, yet also gained 18.31% in 2020, 37.28% in 2017 and 33.57% in 2025, according to MSCI performance data. Those reversals show that emerging-market equities have repeatedly moved between global risk aversion and risk appetite. A single AI-led decline fits that historical pattern better than it proves a permanent break.

A second comparison is more revealing. The 2022 decline occurred during a broad tightening and growth shock, while the 2018 decline coincided with a global trade and dollar-liquidity squeeze. The 2020 and 2017 gains came during periods when global liquidity and growth expectations supported risk assets. Those episodes show three distinct cyclical drivers—rates, trade and liquidity—rather than a single permanent rule. They also show why the present decline should not be read as proof that emerging markets have lost their capacity to recover.

But the historical analogy has a limit. Earlier emerging-market cycles were often described through commodities, dollar liquidity or China’s domestic demand. Today, the benchmark’s largest company exposures are increasingly connected to a global technology capex chain. The cyclical shock is therefore passing through a structural channel. The market can mean-revert after a disappointing AI week, but the composition of the benchmark will not mean-revert on its own.

MSCI’s current weights make that channel visible. TSMC, Samsung and SK Hynix together exceed the weight of many entire country markets in the index. Add MediaTek at 1.54%, Samsung Electronics preferred at 0.93% and Hon Hai at 0.82%, and the semiconductor and hardware complex becomes an even larger source of common factor risk. Tencent and Alibaba, at 2.66% and 1.84%, add a separate Chinese internet and AI sensitivity. The benchmark is not an AI pure play, but it is no longer safe to model it as a neutral collection of unrelated economies.

The structure creates a feedback loop. A fall in chip shares reduces index performance. Passive and benchmark-aware investors then rebalance, which can increase selling in the same large constituents. Currency hedging adds pressure to the won and Taiwan dollar. Local investors see foreign selling and become more defensive, while higher currency volatility raises the discount rate applied to domestic assets. The first-order event is a lower equity price; the second-order event is a more expensive funding environment for the economy that produced the equity winners.

“What’s striking in the current cycle is how the breadth of emerging markets is providing investors with a way to moderate the sudden reversals in the US stock market stemming from the concentration in artificial intelligence and technology stocks,” said Kamakshya Trivedi, chief foreign exchange and emerging-markets strategist at Goldman Sachs.

That claim remains useful, but the current session shows its boundary. Geographic breadth can moderate a shock only when capital actually moves into less correlated markets. If investors sell the benchmark as a package, the diversification benefit is delayed or lost. The asset class may be diversified in economic reality while remaining concentrated in portfolio implementation.

Why the AI Debate Moves Currencies and Capital Flows

The currency weakness is not an afterthought to the equity story. It is the transmission mechanism that turns an equity correction into a broader emerging-market event. The AI concern starts with a question about future revenue and ends with a question about the funding currency of the companies, households and governments exposed to foreign capital.

South Korea and Taiwan sit near the upstream end of the AI supply chain. Their exporters benefit when demand for advanced logic, memory and equipment rises, but their markets also become more sensitive to changes in global inventory, customer concentration and capital-spending expectations. A U.S. software company can delay a project; a semiconductor supplier may carry the adjustment through orders, utilization and pricing. That makes the equity market a forward-looking gauge of industrial demand. When investors cut exposure, the currency can weaken before the economic data fully reflects the change.

Global funds add another layer. The IMF said cumulative cross-border portfolio inflows to emerging markets approached $4 trillion in 2025 and that nonbank financial investors, particularly investment funds, were major drivers of those flows. The IMF also found that risk-sensitive investors pull back more aggressively during global stress, while stronger institutions, reserve buffers and lower fiscal risks reduce the damage. This distinction matters: an AI-led shock should not affect every emerging market equally, even if the headline index falls.

Countries with deep reserves, credible policy frameworks and domestic investor bases can absorb a foreign-flow reversal more easily than markets dependent on short-term external financing. A semiconductor correction may therefore produce a relative rotation inside emerging markets rather than a uniform exit. The exposed side includes North Asian technology exporters and their currencies. Potential beneficiaries include domestic-demand markets, commodity producers and bond markets where lower imported inflation gives central banks more room to respond. But the rotation must be funded; if dollar liquidity tightens at the same time, even the less exposed markets can initially decline.

This is where the market’s consensus can be misleading. The obvious view is that AI concerns hurt chip stocks and help everything else. The second-order reality is that the same concern can reduce the willingness of global funds to hold emerging-market risk generally, even when a country has little technology exposure. A falling benchmark can trigger risk limits, redemptions and currency hedges. In that sequence, Brazil or South Africa may be sold not because their earnings depend on AI, but because their positions are liquid enough to fund losses elsewhere.

The relevant distinction is between a rotation and a liquidation. In a rotation, capital leaves technology and enters domestic demand, commodities or local fixed income; currencies outside North Asia can stabilize as the proceeds are redeployed. In a liquidation, funds reduce total risk, raise cash and hedge emerging-market currency exposure broadly. The first looks like a change in leadership. The second looks like a breakdown in the asset class. Equity weakness alone cannot identify which process is under way; currency breadth and flow data must confirm it.

Goldman Sachs Research said more than $45 billion flowed into emerging markets between April and October after concern about tariffs eased. That figure illustrates both the strength and fragility of the recent recovery. New money can lift a broad set of assets, but it can also leave quickly when the common global narrative changes. Portfolio flows behave like a tide: they raise boats together, but they do not ask whether each boat is powered by semiconductors or domestic consumption before retreating.

The immediate currency signal should therefore be read through funding conditions rather than country headlines. If the won and Taiwan dollar weaken alongside equity declines, the market is pricing a reduction in future export earnings and foreign demand. If equities fall but currencies stabilize, the move is more likely to represent a valuation reset concentrated in stocks. If both weaken while U.S. yields and the dollar rise, the AI concern has become a global liquidity event.

The Strongest Counter-Thesis: This Is a Buying Opportunity, Not a Regime Break

The strongest argument against a structural-vulnerability thesis is that the AI buildout remains early, earnings expectations are still rising and emerging markets control indispensable parts of the supply chain. Goldman Sachs expects 37% EPS growth in regional technology hardware and semiconductor sectors in 2026. That is not the profile of an industry whose demand has already collapsed. It is the profile of an industry whose valuation is vulnerable because the growth rate is high and the capital cycle is visible.

On this view, the selloff is a positioning flush before a major earnings test. Nvidia’s results can restore the market’s confidence in data-center spending; TSMC, Samsung and SK Hynix can benefit from a renewed focus on actual orders rather than abstract fears about AI monetization. The broader emerging-market index also trades at a lower valuation than the growth leaders that dominate developed-market technology benchmarks. MSCI lists the EM index at a forward P/E of 8.70, a fact that gives the bulls a valuation-based defence even if the headline decline is uncomfortable.

The counter-thesis also has a macro foundation. A weaker dollar, easing inflation and lower commodity prices can improve emerging-market financial conditions. Goldman’s research argues that lower commodity prices help contain inflation and give emerging-market central banks more room to reduce interest rates. If that cycle continues, domestic demand and local bonds can offset weakness in technology exporters. The AI trade would then be one engine of emerging-market returns, not the whole vehicle.

The bullish case has a further advantage: supply-chain bottlenecks and strategic demand can make the current cycle less like a traditional consumer-electronics boom. AI customers may continue to spend because computing capacity has become a competitive requirement, not a discretionary upgrade. If that is true, temporary investor anxiety can create a sharp but brief drawdown in companies whose long-term order books remain intact. The market would then be punishing uncertainty rather than discovering an earnings collapse.

That argument is credible, but it does not remove the concentration problem. Lower forward valuations do not guarantee protection when earnings estimates are revised together. A cheap index can become cheaper if the market decides that the 37% growth forecast requires an unsustainably large increase in capital expenditure. The question is not whether AI demand exists. It is whether the incremental spending produces returns high enough to justify the next round of spending.

The counter-thesis also underestimates the difference between corporate resilience and index resilience. TSMC or SK Hynix can remain financially healthy while the MSCI EM index performs poorly if investors reduce the multiple assigned to the entire theme. A strong balance sheet protects a company’s survival; it does not prevent a benchmark from being repriced when leadership becomes crowded. The market can be right about long-run technology demand and still wrong-footed about near-term returns.

My central judgment is therefore conditional. The Nov. 16 decline is cyclical and likely to be tested by corporate earnings, but the market’s sensitivity to that test is structural. The view would be wrong if the next two reporting cycles show that AI infrastructure demand is broadening rather than concentrating, with semiconductor orders, memory pricing and customer capital expenditure all accelerating without a material rise in inventories. A concrete falsifying signal would be two consecutive quarters in which major Asian chip suppliers report order growth above guidance while cutting inventories and maintaining or raising full-year operating-margin targets. That combination would show that the market has mistaken a valuation reset for a demand problem.

The opposite signal would confirm the structural-risk thesis: two consecutive quarters of downward revenue guidance from major AI hardware suppliers, accompanied by inventory increases and a decline in expected sector EPS growth from 37% toward the mid-teens. Such a print would make the concentration visible in earnings, not just in index weights.

What the Selloff Means Across Time Horizons

In the short term, sentiment and liquidity dominate. The next market session will be governed by Nvidia’s earnings, U.S. technology futures, the dollar and the behaviour of the won and Taiwan dollar. A strong earnings outlook could produce a reflexive rebound in the same large-cap names that led the decline. A weak outlook could force benchmark funds to reduce exposure again. Short-term beneficiaries are the least crowded liquid markets; short-term exposures are the semiconductor-heavy indices and currencies whose forward earnings depend on foreign demand.

Over the medium term, the question shifts from whether AI spending continues to whether profits keep pace with spending. Semiconductor earnings can rise while equity returns fall if investors believe the capital cycle is peaking or if margins compress under new capacity. The distinction matters for Taiwan and South Korea because their markets are linked to both volume and pricing. A rising shipment number is not enough if memory prices fall, inventories build or customers lengthen procurement cycles.

For domestic-demand emerging markets, the medium-term outcome depends on the second-order flow response. If capital rotates rather than exits, countries with stronger current accounts and lower technology concentration can outperform. If funds meet redemptions by selling the most liquid assets, the first phase of the rotation may hurt almost everyone. The IMF’s finding that stronger buffers reduce the impact of risk-off flows suggests that country fundamentals will matter more after the initial liquidation wave, not necessarily during it.

Over the long term, the structural issue is benchmark design. Investors who access emerging markets through market-cap-weighted products are buying the winners of the AI supply chain in proportions determined by their market values. That is not the same as buying a balanced exposure to emerging-market growth. The benchmark will remain vulnerable until either the semiconductor leaders shrink as a share of the index, other sectors grow faster, or investors use approaches that reduce single-theme concentration. None of those changes is automatic.

The long-term implication is not that technology should disappear from emerging-market portfolios. Semiconductors are a real industrial advantage for the region, and the official index weights reflect genuine cash flows, not a fictional theme. The implication is narrower: benchmark exposure now carries a factor risk that country labels conceal. An investor can believe in South Korea’s industrial competitiveness, Taiwan’s manufacturing position and the growth of AI while still recognizing that all three can be repriced by the same global discount-rate and capital-cycle shock.

The base case is a volatile but contained correction: earnings keep AI demand alive, the dollar remains manageable and capital rotates within emerging markets. The upside case requires a positive earnings surprise together with stable currencies and evidence that domestic-demand sectors are attracting flows; in that scenario, the current selloff becomes a reset inside a continuing growth cycle. The downside case requires a negative earnings sequence, rising inventories and a stronger dollar; then the equity decline would become a broader tightening in emerging-market financial conditions.

Investors and policymakers will have several measurable tests. Corporate guidance will show whether spending is broadening. Memory and advanced-chip pricing will show whether supply is catching demand. The won and Taiwan dollar will show whether foreign investors are hedging an equity trade or questioning the export outlook. Finally, cross-border fund-flow data will show whether the $45 billion inflow pattern has reversed. The structural thesis fails if these indicators improve together; it strengthens if they deteriorate together.

Emerging markets have not become an AI bubble by definition. But the benchmark has become a faster route for AI fear to reach currencies, capital flows and domestic financial conditions. The Nov. 16 selloff is cyclical in trigger and structural in vulnerability: the earnings can reverse the move, but earnings alone will not diversify the index.

Data cutoff: Nov. 16, 2026.

Explore more exclusive insights at nextfin.ai.

Insights

What drives MSCI EM concentration?

How do chip stocks shape EM returns?

Why do AI concerns hit currencies?

What triggered the Nov. 16 selloff?

Are AI worries cyclical or structural?

How does passive selling amplify losses?

What do Nvidia earnings signal next?

Can domestic-demand markets outperform?

How do 2026 forecasts affect risk?

What could disprove structural risk?

How does EM compare with past crises?

What role do currency flows play?

Why are chip leaders linked globally?

Can cheap valuations prevent losses?

Can index design reduce concentration?

What signals rotation over liquidation?

How could AI reshape EM markets?

What are the main risks to chip demand?

Search, ask or analyze anything...