China’s artificial-intelligence rally is starting to generate its own ETF product cycle, as the country’s technology market shifts toward semiconductors, memory and AI infrastructure. The KraneShares China Technology & Semiconductor STAR 50 Index ETF, KSTR, offers one measure of the scale and volatility behind that shift.
KSTR tracks the SSE Science and Technology Innovation Board 50 Index, covering 50 large and liquid companies listed on Shanghai’s STAR Market. The exchange accounted for approximately 62% of capital raised through new mainland Chinese IPOs in 2025, according to Deloitte data cited by KraneShares. As of August 26, 2026, KSTR held $386.6 million in net assets and charged a 0.65% net expense ratio.
KSTR produced a 131.8% trailing one-year NAV return through June 30. By July 31, its trailing one-year return had fallen to 64.0% after a 25.9% monthly decline. The fund was still up 27.22% year to date.
CXMT Gives the Rally an Earnings Engine
The equity rally is increasingly being supported by earnings from China’s semiconductor industry. CXMT reported RMB150.3 billion, or $22.4 billion, of revenue for the first half of 2026, almost 10 times the year-earlier level and more than double its sales for all of 2025, according to Bloomberg on August 28. Net profit reached RMB77.6 billion, reversing a loss a year earlier. The company is now the world’s fourth-largest DRAM producer, according to Bloomberg. The Wall Street Journal separately reported first-half revenue of RMB150.31 billion and net profit of RMB77.61 billion.
The numbers followed one of China’s largest technology listings. CXMT raised RMB66.6 billion in its Shanghai IPO including the overallotment option, and its shares have since pushed the company above Tencent by market capitalization, according to Bloomberg. CXMT is developing high-bandwidth memory, the type of advanced memory used alongside AI accelerators, while directing IPO proceeds toward capacity expansion.
Its arrival in public markets matters for ETFs because investors can now obtain direct exposure through some Chinese semiconductor portfolios. The Global X China Semiconductor ETF, ticker 3191 in Hong Kong, held CXMT at 8.25% of net assets as of August 26, making it the fund’s fifth-largest position. The ETF had $461.6 million in net assets and was up 37.95% year to date on a NAV basis at the same date.
ETF Issuers Are Slicing China’s AI Market More Finely
The product response now extends well beyond KSTR. VanEck launched the VanEck China Semiconductor ETF, SMHC, on June 23, 2026. The Nasdaq-listed fund tracks the MarketVector China Semiconductor 25 Index and charges a 0.65% expense ratio, targeting 25 large and liquid Chinese semiconductor companies.
KraneShares followed on August 19 with the KraneShares Asia AI Technology ETF, KAIT. Unlike the China-focused semiconductor products, KAIT extends across Taiwan, South Korea, Japan and China, targeting companies involved in foundries, high-bandwidth memory, optical networking, data-center infrastructure and other parts of the AI supply chain. KAIT had $5.0 million in net assets as of August 27 and charges a 0.65% net expense ratio.
A week later, EMXETF launched the China AI Tigers LLM ETF, TGRZ, on Nasdaq. Its mandate moves away from semiconductor manufacturing toward companies developing Chinese large language models and generative AI. The issuer describes TGRZ as focused on the model layer of the AI stack, giving it a different exposure from hardware-heavy KSTR, SMHC and 3191.
Sources: KraneShares, VanEck, Global X and EMXETF. Data checked August 28, 2026.
This expansion changes what a “China AI ETF” can mean. KSTR combines technology and semiconductor exposure through the STAR Market, while SMHC and 3191 move deeper into chips. KAIT broadens the geographic scope to Asian infrastructure, while TGRZ moves in the opposite direction by narrowing exposure to the model layer.
Asia’s Semiconductor Base Explains the Product Rush
The industrial geography supports that segmentation. Asia accounts for approximately 75% of global semiconductor capacity and material supply, based on Oxford Economics data from May 2026 cited by KraneShares. China, Taiwan and South Korea accounted for 79% of global semiconductor manufacturing-equipment spending, according to 2025 SEMI data cited by the issuer.
The figures also show why these ETFs carry risks that broad technology funds can dilute. Semiconductor portfolios can concentrate exposure to capital-spending cycles, memory prices and Chinese industrial policy. Funds focused on AI models introduce a different set of risks around valuations, competition and the commercial durability of individual platforms.
KSTR provides a useful warning about how quickly returns can reset. Its trailing one-year NAV gain fell from 131.8% at June 30 to 64.0% at July 31, even though the fund remained firmly positive for the year. Global X 3191 offers another example: it was up 37.95% year to date as of August 26, yet had fallen 12.78% over the preceding month.
For investors accessing these products from the GCC, domicile and trading venue also matter. KSTR, SMHC, KAIT and TGRZ are U.S.-listed ETFs, while 3191 trades in Hong Kong and holds predominantly mainland Chinese securities. That creates different considerations around market hours, currency exposure, liquidity and tax treatment for UAE and Saudi-based investors.
China’s AI ETF market is therefore beginning to resemble the industry it tracks. CXMT’s rise has expanded the listed memory-chip universe at the same time that issuers are separating semiconductor manufacturing, regional AI infrastructure and model developers into distinct portfolios. The next test is whether investor assets follow the proliferation of products after the extraordinary equity returns that prompted many of their launches.
