When DeepSeek stunned the world in early 2025, everyone wanted to know the same thing: who built this? The answer surprised people. Not a Silicon Valley veteran, not a Big Tech spin-off — but a low-key Chinese hedge-fund founder who had quietly stockpiled thousands of AI chips years earlier, and who ran his lab more like a university research group than a startup. Liang Wenfeng is his name, and his path to the center of the AI world is one of the most unusual founder stories in tech.
Here’s who Liang Wenfeng is, told straight: the trader-turned-AI-founder behind DeepSeek, how he got there, and why his approach unsettled the assumptions of much richer, more famous competitors.
From a Small Town to Zhejiang University
Liang Wenfeng was born in 1985 in Wuchuan, a city in China’s Guangdong province. He studied at Zhejiang University, one of China’s top engineering schools, earning a bachelor’s degree in Electronic Information Engineering in 2007 and a master’s in Information and Communication Engineering in 2010. His graduate research was refreshingly practical — a thesis on low-cost object tracking with steerable cameras.
The formative moment came during the 2008 global financial crisis. Still a student, Liang and some classmates started experimenting with using machine learning to trade financial markets — teaching algorithms to spot patterns in price data. It was an early, hands-on fusion of two fields — AI and finance — that would define the rest of his career.
Building High-Flyer, an AI-Driven Hedge Fund
In February 2016, Liang co-founded a quantitative hedge fund called High-Flyer with two of his Zhejiang classmates. “Quantitative” means the fund trades using mathematical models and algorithms rather than human hunches — and High-Flyer leaned into AI hard, shifting to fully AI-driven trading strategies by 2021. It grew into a major player, reportedly managing on the order of $10 billion in assets.
Running an AI-powered hedge fund meant one thing above all: Liang needed serious computing power. And this is where the story takes its most consequential turn.
The Chip Stockpile That Changed Everything
To feed its AI trading models, High-Flyer built its own supercomputers — Fire-Flyer I in 2020, and the far larger Fire-Flyer II in 2021, constructed around roughly 10,000 Nvidia A100 GPUs, the top-tier AI chips of the era. At the time, this was simply infrastructure for a hedge fund.
But timing turned it into something more. Liang amassed those chips before the United States tightened its export controls in October 2022, restricting sales of the best AI chips to China. When the door closed, High-Flyer was already on the other side of it — sitting on a stockpile of exactly the hardware that would soon be hard for Chinese companies to get. That pre-restriction cache became one of the quiet advantages behind DeepSeek’s later rise, and a key detail in the debate over whether those export controls actually worked.
From Trading Profits to Chasing AGI
In April 2023, High-Flyer announced it was spinning up a dedicated AI research lab — one aimed not at trading, but at artificial general intelligence. That lab, formally incorporated in July 2023, was DeepSeek. Liang personally holds a controlling stake, and the profitable hedge fund bankrolled the venture, freeing DeepSeek from the pressure to chase quick commercial returns.
That funding structure matters. Because DeepSeek didn’t need to sell a product tomorrow, it could pour resources into fundamental research — the efficiency breakthroughs like Mixture of Experts and 8-bit training that let it build a competitive model without a frontier-lab budget. A hedge fund’s war chest, aimed at research instead of trading, turned out to be a formidable combination.
A Different Philosophy
Liang runs DeepSeek in a way that stands out from both Chinese tech giants and Silicon Valley. In interviews he has argued that keeping AI closed-source “can only temporarily delay” competitors rather than stop them — a striking stance in an industry where the leading labs guard their models closely. DeepSeek’s decision to release its models as open weights, free for anyone to download and build on, flows directly from that view.
He’s also expressed an unusual read on China’s tech ambitions: that what the country lacks “is not capital but confidence and knowledge on organizing talent.” In keeping with that, DeepSeek reportedly hires for ability and curiosity over pedigree, and organizes work in small, flat, project-based teams rather than rigid hierarchies — closer to an academic lab than a corporate org chart.
The Man of the Moment
DeepSeek’s 2025 breakout thrust Liang into the global spotlight. He was named to Time’s 100 Most Influential People in AI and Fortune’s 100 Most Powerful People in Business. His paper wealth soared with DeepSeek’s prominence — by mid-2026, Bloomberg estimated his net worth in the tens of billions of dollars, ranking him among the wealthiest founders of any AI-model company (though such figures move quickly and should be taken as a snapshot, not a fixed number).
He also received notable acknowledgment at home, invited to high-level symposiums with China’s top leadership as DeepSeek became a point of national pride. Yet by most accounts Liang remains far more comfortable talking about research than fame — a former quant trader who, almost by accident of timing and conviction, ended up reshaping how the entire world thinks about who gets to build cutting-edge AI.
❓ Frequently Asked Questions
Who is Liang Wenfeng? Liang Wenfeng is the Chinese entrepreneur who founded DeepSeek, the AI company whose models shocked the tech world in early 2025. Born in 1985 and educated at Zhejiang University, he first built an AI-driven quantitative hedge fund called High-Flyer, then used it to launch and fund DeepSeek as an AI research lab.
How did a hedge-fund manager end up building a leading AI model? Liang’s hedge fund, High-Flyer, used AI for trading and had already built large GPU supercomputers and a strong machine-learning team. In 2023 he redirected that expertise and hardware toward AI research, founding DeepSeek — and because the fund financed it, DeepSeek could focus on fundamental research rather than immediate profit.
Why was Liang’s GPU stockpile so important? High-Flyer acquired roughly 10,000 top-tier Nvidia A100 chips around 2021, before the US tightened export controls on advanced AI chips to China in October 2022. That pre-restriction stockpile gave DeepSeek access to hardware that later became difficult for Chinese firms to obtain.
Is Liang Wenfeng a billionaire? As of mid-2026, Bloomberg estimated his net worth in the tens of billions of dollars, largely on paper through his controlling stake in DeepSeek and High-Flyer, making him one of the wealthiest founders of an AI-model company. Such valuations move quickly, so any specific figure is a snapshot in time.
Why does Liang support open-weight AI? He has publicly argued that keeping AI closed-source only temporarily delays competitors rather than preventing them from catching up, and that China’s real gap is in organizing talent, not funding. DeepSeek’s release of its models as free, open weights reflects that philosophy.
📚 Keep Learning
- The Day DeepSeek Broke Nvidia — the market shock Liang’s lab set off.
- Learn AI and ML Concepts Through the Lens of DeepSeek — the technology his team built.
- Did DeepSeek Really Cost $6 Million? — the real economics behind the story.
📚 Recommended Resources
Disclosure: some links below are Amazon affiliate links. If you buy through them, I earn a small commission — at no extra cost to you — which helps keep this site running.
Artificial Intelligence: A Guide for Thinking Humans
Melanie Mitchell · Book · ~$28 on Amazon
Liang Wenfeng bet his lab on the idea that fundamental AI research beats hype. This is the book that helps you understand what that research actually is — a clear, no-math guide to what modern AI can and can’t do, so you can follow the DeepSeek story on the merits rather than the drama.
Hands-On Machine Learning with Scikit-Learn, Keras & TensorFlow
Aurélien Géron · Book · ~$70 on Amazon
Liang’s whole career started with using machine learning to trade markets. If that fusion of AI and real-world data appeals to you, this is the best practical on-ramp — the same fundamentals, taught hands-on with runnable code, from the ground up.