Up until last week, Liang Wenfeng was a relatively unknown figure. The 40-year-old founder of the Chinese AI startup DeepSeek had kept a low profile, just one of many passionate quantitative minds in tech. Now, he has burst onto the global stage, with Silicon Valley heavyweights like Elon Musk, Sam Altman, and OpenAI investor Joshua Kushner eagerly watching and anticipating Liang’s next move. In a matter of two weeks, DeepSeek’s AI assistant has overtaken rival ChatGPT as the top-rated free app on Apple’s U.S. App Store.
Liang’s DeepSeek breakthrough has propelled him to the forefront of China’s AI industry, challenging the narrative that the country lags behind U.S. tech giants. DeepSeek’s latest AI model is reportedly on a par with—or even superior to—industry-leading models in the U.S., all at a fraction of the cost. A report from Liang’s firm claims that training DeepSeek-V3 required less data and significantly lower expense than its competitors, though some skeptics have questioned these findings.
DeepSeek’s emergence as a cheaper AI alternative has sent shockwaves through the U.S. market, wiping over $1 trillion in market cap. But little is known about the man behind it.
Math-minded entrepreneurship
Raised on the southern coast of China in Zhanjiang, Liang was a straight-A student with an aptitude for math. At 17, he enrolled at Zhejiang University, majoring in electronics and communication engineering, before earning a master’s degree in information and communication engineering in 2010.
Soon after, he reportedly began developing AI algorithms for stock trading and, in 2013, opened his first solo venture, Jacobi, an investment firm named after German mathematician Carl Jacobi. Two years later, he and several college friends cofounded High-Flyer Quantitative Investment Management, a hedge fund specializing in machine-learning-driven quantitative models for computerized stock trading.
High-Flyer was an early adopter of advanced AI techniques in China, allowing computers to process vast data sets and identify key market patterns.
By the end of 2021, the fund’s portfolio had grown to over $13.79 billion. According to financial data provider Simu Paipaiwang, at least five of High-Flyer’s funds produced an average excess return of more than 20% over market benchmarks in the past five years. The New York Times noted that High-Flyer found success by capitalizing on the impulsive nature of China’s retail investors, who dominate the market.
Like many Chinese quantitative traders, High-Flyer suffered losses after regulatory crackdowns in recent years, though it reportedly still manages $8 billion in assets.
Facing additional regulatory pressure, the firm pivoted to align with China’s growing focus on advanced AI. In April 2023, High-Flyer announced plans to expand beyond the investment sector and “explore the essence of AGI.” A month later, DeepSeek was created as an independent research group, backed by High-Flyer, with Liang at the helm.
Curiosity rather than cash
Despite Liang’s meteoric rise, colleagues say he isn’t motivated by wealth or fame, according to the Wall Street Journal. The bespectacled and reserved mastermind behind AI’s new hottest venture is described as a hands-on leader who cares little about his clothing and appearance, relies on formulas and calculations to solve problems, and has a passion for soccer. Peers claim he once expressed a desire to earn the respect of the U.S.-dominated tech world.
His interest in AI dates back to 2019 when his team at High-Flyer began acquiring and developing computing systems powered by Nvidia graphics-processing units, well ahead of most Chinese firms. By staying ahead of the curve, he positioned himself among the few Chinese companies with more than 10,000 high-end chips—a move that proved critical when the U.S. restricted sales of these chips to China.
“It is like buying a piano,” Liang told Chinese tech publication 36Kr in 2023, speaking about the chip purchases. “Firstly, it’s because you can afford it. And secondly, it’s because you have a group of people who are eager to play music on it.”
At the time, those in Liang’s orbit wrote off his stockpiling of chips as little more than a passion project.
“He was this very nerdy guy with a terrible hairstyle talking about building a 10,000-chip cluster to train his own models,” one of Liang’s business partners told the Financial Times. “He couldn’t articulate his vision other than saying: ‘I want to build this, and it will be a game change.’ We thought this was only possible from giants like ByteDance and Alibaba.”
But Liang has seemingly accomplished what these giants could not. For two years, while major Chinese technology firms trailed Silicon Valley’s AI—releasing chatbots and running flashy marketing campaigns—DeepSeek focused on research rather than relying solely on making consumer-facing AI products for revenue.
“In the past thirty years, [China’s tech industry] has only emphasized making money and ignored innovation,” Liang told Chinese media outlet Waves in July.
He believes that prioritizing research over profits has allowed DeepSeek to catch up with American innovation and gain a competitive edge. Last year, the company significantly lowered prices for developers using its AI model, sparking a price war with larger Chinese rivals. Unlike OpenAI, DeepSeek has opted to make all its models open source, allowing any developer to use and modify them freely.
U.S. skepticism
Liang’s claim that DeepSeek constructed its latest model in just two months for under $6 million—roughly 3% to 5% of the reported cost for OpenAI to create its next model—has raised skepticism across the tech industry, especially given reports that DeepSeek only had access to older, less powerful Nvidia chips. Some experts argue that these claims are likely exaggerated or outright false.
While OpenAI CEO Sam Altman hailed DeepSeek’s model at price point as “impressive,” he was quick to assert that OpenAI would “obviously deliver much better models” and emphasized that greater computing power remains a crucial advantage. The American company has since launched an investigation into whether DeepSeek trained its new chatbot using OpenAI’s model.
Alexandr Wang, CEO of Scale AI and the world’s youngest self-made billionaire, took a more confrontational stance. Last week, he told CNBC his company’s tests confirmed that DeepSeek’s R1 model could outperform, or at least compete with, top U.S. equivalents. However, he alleged that the Chinese firm was using more cutting-edge Nvidia chips in violation of U.S. export restrictions.
Elon Musk appeared to agree with Wang’s comments, responding with a blunt “Obviously” to a clip of the CNBC segment on X.
Josh Kushner of Thrive Capital, which led OpenAI’s $6.6 billion fundraising round last year, also raised concerns about user privacy data.
“Pro-America” technologists openly supporting a Chinese model that was trained off of leading US frontier models, with chips that likely violate export controls, and—according to their own terms of service—take US customer data back to China,” Kushner posted on X.
Despite the naysayers, the once under-the-radar DeepSeek has become an overnight sensation globally, achieving what venture capitalist Marc Andreessen called a “Sputnik moment”: the Soviet Union’s surprise launch of the first satellite into space, which reshaped global competition.
