Sector Investors News and Insights

World Wide Wednesday: EM’s AI Split — Why China Is Slumping While Taiwan and Korea Soar

Artificial intelligence may be the dominant global equity theme of 2026, but it is producing some remarkably different outcomes across emerging markets.  Taiwan and South Korea increasingly look like direct equity-market expressions of the global AI capital-spending boom. China, despite possessing one of the world’s deepest AI ecosystems, continues to trade more like a struggling domestic economy burdened by property weakness, uncertain consumer demand and a persistent policy risk premium.

That divergence raises a bigger question for international investors: Have emerging-market equities effectively become an AI proxy?  At the benchmark level, the answer is increasingly yes. At the country level, emphatically no.

The EM Index Is Looking More Like a Technology Index

Consider the composition of the iShares Core MSCI Emerging Markets ETF (IEMG). As of September 10, Taiwan represented roughly 27.7% of the portfolio and South Korea another 20.9%. China, once unquestionably the center of gravity in emerging markets, accounted for only about 18.3%. Information Technology represented approximately 40.7% of IEMG.  That weighting matters because the performance dispersion has been extraordinary. BlackRock data through September 9 showed the iShares MSCI Taiwan ETF (EWT) up 75.7% year to date and the iShares MSCI South Korea ETF (EWY) up 94.3%. By comparison, the iShares MSCI China ETF (MCHI) was down 10.6%. The broader iShares MSCI Emerging Markets ETF (EEM) was up approximately 25.8% over the same period.

The underlying economic data reinforce the point. Taiwan reported a record $103 billion of export orders in August, up 71.4% from a year earlier, with the country’s Economy Ministry pointing to AI servers and related semiconductor demand as major drivers. South Korean customs data showed exports during the first 20 days of September up 78.3% year over year, with semiconductor exports surging 259.4% and accounting for nearly 48% of total exports.

For Taiwan and Korea, AI is increasingly visible not only in stock prices, but in orders, exports, earnings and manufacturing utilization.

So Why Isn’t China Participating?

China certainly does not lack AI exposure.  Alibaba this week unveiled a new internally developed AI chip, plans for an even larger next-generation model and a substantial expansion of its data-center infrastructure. Its shares initially rose on the announcement. Chinese developers such as DeepSeek, Moonshot AI, Z.AI and MiniMax have also demonstrated that competitive frontier models can be built with substantially fewer resources than some U.S. counterparts.  The problem is that owning Chinese equities involves owning far more than that AI story.

China increasingly looks like two economies operating simultaneously.  The production side remains relatively strong. Industrial output increased 5.2% year over year in August, while high-tech manufacturing output rose 16.7%, according to China’s National Bureau of Statistics.  The domestic-demand side looks much weaker. August retail sales increased just 0.4% year over year, while fixed-asset investment fell 7.2% during the first eight months of the year. Real-estate development investment plunged 19.9%, new housing starts fell 24.8%, and new commercial-property sales declined 13%.  That creates an important distinction for investors. Taiwan’s and Korea’s AI champions are monetizing an external global capital-spending cycle. China is trying to develop an AI economy while simultaneously absorbing the consequences of a multiyear property correction and weak household demand.  Even a booming AI industry cannot instantly offset those problems across an entire equity market.

China’s AI Strength Also Comes With an AI Risk Premium

There is another difference: policy.  Wednesday’s morning headline pack highlighted reports that Chinese authorities were investigating DeepSeek and Moonshot AI over data-security issues while also examining state data centers’ dependence on Broadcom equipment.  Those developments immediately hit Chinese technology shares. Alibaba fell more than 4% in Hong Kong, while Z.AI, MiniMax, Xiaomi and other AI-related names declined. The Hang Seng Tech Index fell as well.  The episode illustrates an enduring problem for China’s technology sector: Beijing simultaneously wants rapid AI development, greater semiconductor self-sufficiency, strict data sovereignty and control over strategic technologies. Those objectives do not always produce the stable regulatory environment public-equity investors prefer.

China has also reportedly been slowing the rush of humanoid-robotics IPOs amid concern that valuations and state-supported revenues have moved ahead of commercial demand.  That doesn’t make Chinese AI development less important. It does mean technological progress does not necessarily translate one-for-one into expanding valuation multiples for listed companies.

China May Be Winning the Cost War Without Winning the Stock-Market War

There is also an economic paradox inside China’s AI success.  Chinese model developers have increasingly competed through efficiency, open-weight models and dramatically lower prices. That competitive pressure has helped push the global AI industry toward cheaper inference and greater cost efficiency. Reuters analysis recently noted that several Chinese developers are operating surprisingly close to frontier U.S. capabilities despite tighter hardware constraints and dramatically smaller budgets.

That is strategically impressive.  But cheap AI can be better for AI adoption than for AI-company margins.

When models become less expensive and increasingly interchangeable, the economic value can migrate toward semiconductor manufacturers, memory suppliers, networking equipment, cloud infrastructure and businesses capable of embedding AI into existing profitable customer relationships.  That helps explain why the most straightforward EM beneficiaries have been places such as Taiwan and South Korea. They occupy critical positions in the physical AI supply chain.  China’s developers, by contrast, are competing in a market where low prices, open models, large R&D budgets and regulatory constraints could make monetization considerably harder.

Is EM Now Just an AI Proxy?

Not quite. But investors should understand how much the benchmark has changed.  Taiwan and Korea together now represent almost half of IEMG, while technology alone represents more than 40%. Taiwan’s EWT is itself approximately 75% Information Technology, while SK Hynix alone recently represented roughly a quarter of EWY.  That means the marginal driver of broad EM index performance today is unusually connected to semiconductors, memory, data centers and global AI capital expenditures.

Still, emerging markets remain much broader underneath the index.  Brazil provides an obvious counterexample. The iShares MSCI Brazil ETF (EWZ) was up roughly 20.5% year to date through September 9, despite having little direct connection to the AI hardware boom. Financials alone still represent more than 18% of IEMG, while India, Brazil, South Africa, Saudi Arabia, Mexico and other markets introduce very different domestic-growth, commodity, interest-rate and currency exposures.  So EM has not become a single-factor asset class.

What has happened is arguably more important: the headline emerging-market benchmark is no longer a particularly clean representation of the average emerging economy.  Its largest performance contributors increasingly reflect one powerful global industrial cycle — AI infrastructure.

China Is the Exception That Proves the Rule

China’s underperformance therefore should not be interpreted as evidence that investors doubt its AI capabilities.  If anything, the opposite may be true. China’s low-cost models, semiconductor ambitions and enormous domestic engineering base increasingly look capable of competing with the U.S. technology ecosystem. Global investors are still willing to fund that development, and AI-related Chinese IPOs continue to attract capital.  The stock market is discounting something broader.

Weak household demand, falling property investment, regulatory uncertainty, restrictions on advanced technology and doubts about corporate monetization continue to overwhelm the positive contribution from AI.

For international ETF investors, that creates an unusually important distinction. Broad EM exposure such as EEM or IEMG currently embeds substantial exposure to the AI hardware cycle through Taiwan and Korea. EWT and EWY represent even more concentrated versions of that trade. MCHI, despite considerable AI exposure within its holdings, remains primarily a China macro, policy and domestic-demand exposure.  Those trades may all sit inside the same emerging-market category, but they are responding to very different economic forces.  And that may be the most important lesson of this year’s international market: “emerging markets” increasingly describes where companies are listed, not what actually drives their returns.

Sources

Disclaimer: This article is for informational and educational purposes only and does not constitute investment advice, an offer to buy or sell securities, or a recommendation of any investment strategy. ETF holdings, exposures and performance figures change over time. Investors should consider their own objectives, risks and circumstances before investing.

Patrick Torbert

Editor | Chief Strategist

Patrick Torbert is a veteran financial market analyst who is currently the Editor and Chief at ETF Insight a NY based full-service content, TV, video podcast and digital marketing firm that represents several ETF issuers. Patrick brings 20+ years of experience from Fidelity Asset Management where he most recently served as an equity and multi-asset analyst.
Scroll to Top

Subscribe to our Newsletter

Stay updated with the latests analysis and insights from etfsector.com

If you haven’t received your newsletter email, check your spam/junk folder and add us to your contacts to ensure delivery.