The man behind China’s most talked-about AI lab has a side hustle, and it involves betting big on the country’s hottest tech IPOs. Liang Wenfeng, who founded DeepSeek and built it into a genuine competitor to OpenAI, is now steering his quantitative hedge fund High-Flyer toward a new mission: capturing value in China’s turbulent IPO market before the rest of the crowd shows up.
From AI trading to IPO hunting
Liang co-founded High-Flyer Quant around 2015-2016, building it into one of China’s most successful quantitative funds. At its peak, the firm managed somewhere between $8B and $13B in assets, powered by AI-driven trading strategies.
Now, that same infrastructure is being redirected. In 2026, High-Flyer and its affiliated entities secured roughly $26 million in pre-IPO allocations for CXMT, a Chinese memory chipmaker. The fund also made smaller bets on Unitree Robotics, a company riding China’s push into humanoid and industrial robotics.
DeepSeek cuts the cord
In June 2026, DeepSeek completed a $7.4 billion external funding round that valued the company at approximately $52 billion. Liang himself contributed around 20 billion yuan, roughly $3 billion, to that round.
The fundraise set the stage for a potential IPO on the Shanghai STAR Market, targeted for 2027. The STAR Market was designed specifically to attract innovative technology companies. DeepSeek, with its $52 billion valuation, would be exactly the kind of company the exchange was built for.
The bigger picture
High-Flyer’s investments in companies like CXMT and Unitree Robotics build expertise and relationships in exactly the sectors where DeepSeek’s AI tools could find commercial applications. Meanwhile, DeepSeek’s growing independence frees up High-Flyer’s capital to pursue more aggressive IPO-related bets.
China’s IPO pipeline has been constrained by tighter regulatory oversight since 2021, when Beijing launched a sweeping crackdown on tech companies, particularly for companies in sectors the government considers strategically vital: chips, AI, advanced manufacturing, and robotics.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.

1 hour ago
21








English (US) ·