As the US continues to pour capital into AI infrastructure, Chinese model developers are competing with lower development costs. Kevin Carter, founder of EMXETF, says the China AI Tigers LLM ETF (TGRZ), which launched last month, focuses on Chinese open-weight model companies. It can also hold larger companies such as Alibaba and allocate up to 15% of assets to private businesses. The fund offers exposure to China’s AI model layer, but its holdings face uncertainty around profitability and commercialisation.
TGRZ focuses on China’s AI model layer
Carter frames TGRZ’s strategy using Nvidia CEO Jensen Huang’s “layered cake” analogy for AI, with the fund focused on the model layer and, in particular, China’s open-weight models. He says some Chinese models are approaching the performance of leading US models at significantly lower development costs.
The “Tigers” in the fund’s name refers to six Chinese AI companies: Z.ai, MiniMax, Moonshot, whose products include Kimi, StepFun, Baichuan and 01.AI. The companies have research backgrounds linked to computer science laboratories at Tsinghua University. The fund can also invest in larger businesses, including Alibaba.
TGRZ may allocate up to 15% of its assets to private companies. Carter says he is watching for pre-IPO investment opportunities in DeepSeek and Moonshot. That describes areas of interest; it does not mean the fund has invested in either company.
Carter flags the lack of profits at model companies
Carter describes his investment style as somewhat “Omaha”-oriented, but stresses that TGRZ is not a value fund. The model companies it targets are not yet profitable, and as model capabilities become more widely available, they may face pressure from product commoditisation and intensifying competition. The fund’s thematic nature makes progress toward commercialisation an important consideration alongside technical capabilities.
For investors seeking value in emerging markets, Carter points to TSMC, Samsung and SK Hynix. He says the three companies have together driven about 80% of emerging-market returns since January 2025. Combined earnings for the group are expected to reach about $965 billion in 2026 and 2027, he says, while their combined price-to-earnings ratio is below 5. SK Hynix trades at about 3.5 times earnings.
Carter says investors continue to view these semiconductor companies as cyclical, while AI demand may prove more sustained over the longer term. He also identifies memory, optics and power supply as potential constraints in the AI supply chain, with electricity in particular likely to become a key bottleneck as computing capacity expands.
A tactical thematic fund, not a core holding
Carter positions TGRZ as a thematic, tactical tool for more active investors, rather than a core emerging-markets holding. For advisers seeking broader, long-term exposure to AI, he says EMX plans to launch a fund covering the AI value chain across emerging markets. A narrower product such as TGRZ could serve as a more specialised allocation within a portfolio.
TGRZ is designed to capture opportunities among Chinese AI model companies, while carrying uncertainty around their path to profitability and the effect of competition on commercial value. Investors assessing the fund’s actual exposure will need to track whether model capabilities translate into recurring revenue and whether its permitted private-company investments are made.