Liang Wenfeng

the richest AI founder in the world right now didn't come from Silicon Valley, didn't raise a Series A from Sequoia, and for a long time, most people outside of Chinese quant finance circles had no idea he existed.

Liang Wenfeng grew up in Guangdong, studied electrical engineering at Zhejiang University, and started his career doing what a certain kind of mathematically gifted person does when they want to understand complex systems — he applied machine learning to financial markets. He co-founded High-Flyer, a quantitative hedge fund, and quietly built it into one of China's most sophisticated trading operations. the interesting thing about that, in retrospect, is what he did with the infrastructure underneath it. as High-Flyer scaled, it accumulated serious computing power — the kind you'd need to run large-scale AI research. so when Liang founded DeepSeek in 2023, he wasn't starting from scratch. he was redirecting.

what made DeepSeek land with force — internationally, not just in China — was the release of R1 earlier this year. the model posted results that analysts said were competitive with the leading American frontier models, but achieved with what appeared to be a fraction of the compute. that's the part that caught people's attention and made certain executives in San Francisco uncomfortable enough to raise the issue with the US government. the implication wasn't lost on anyone watching: if you can get comparable performance with less hardware, the assumptions underpinning the entire chip-export control strategy start to wobble.

then, recently, two things happened more or less simultaneously. DeepSeek closed its first outside funding round — 51 billion yuan — and Liang's net worth crossed $36 billion, putting him ahead of the founders of both OpenAI and Anthropic. the richest AI model startup founder in the world is operating out of Hangzhou, runs a company that has never really courted Western press, and built his fortune on infrastructure that most people didn't think to look at until it produced something they couldn't ignore.

there's also a quieter news thread worth noting. High-Flyer participated in the IPO subscription for ChangXin Memory Technologies, a Chinese memory chip manufacturer — through 153 of its quant funds. that's not a small bet, and it fits a pattern: Liang isn't just building models, he's positioning across the stack, including the hardware layer that Chinese AI development depends on as access to Nvidia's best chips remains constrained.

what makes him interesting isn't the wealth number, though that's the hook everyone's leading with. it's the architecture of how he got here — a quant fund that became a compute operation that became an AI lab, each layer funding the next, all of it remarkably self-contained. DeepSeek has taken outside money now for the first time, and there's real conversation about whether an IPO follows, and what that would mean for a company that has operated with unusual independence. those questions don't have answers yet.

what we do know is that Liang Wenfeng spent years building something that most of the world wasn't watching, and the moment it surfaced, it changed the terms of a conversation that everyone thought they already understood.