Instead, Liang hails from the world of finance. After graduating from Zhejiang University, he cofounded the quantitative hedge fund High-Flyer in 2015 and incorporated AI in its trading strategies to predict market trends and help make investment decisions.

According to the Financial Times, he began buying thousands of Nvidia graphics processors in 2021—before the Biden administration began limiting U.S. exports of AI chips to China—as an AI side project. At the time, acquaintances viewed it as a quirky hobby that didn’t look like it would go anywhere.

“When we first met him, he was this very nerdy guy with a terrible hairstyle talking about building a 10,000-chip cluster to train his own models. We didn’t take him seriously,” one of Liang’s business partners told the FT. “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.”

High-Flyer and DeepSeek didn’t immediately respond to requests to interview Liang.

Then he started DeepSeek in 2023, aiming to develop artificial general intelligence, or a level of AI that matches human intelligence.

DeepSeek is now laying waste to the U.S. stock rally, as its development of cutting-edge AI at a fraction of the cost that rivals like OpenAI and Google are spending has raised doubts about the hundreds of billions of dollars that are slated for investment in the sector and elsewhere.

On Monday, shares of Nvidia sank 17%, leading a selloff that sent the Nasdaq down 3% and the S&P 500 down 1.8%. That’s after the so-called Magnificent Seven tech companies have driven the bulk of U.S. stock gains, after the release of OpenAI’s ChatGPT in late 2022 sparked an AI boom.

Companies outside the tech world eventually joined in the rally but are now retreating too. Constellation Energy, which has been the beneficiary of AI data centers sucking up massive amounts of electricity, plunged 20% on Monday.

DeepSeek’s moat, according to Liang

While DeepSeek was initially seen as a side project, Liang, 40, is personally engaged in the company and its research, according to the FT.

He also is dedicated to making DeepSeek a homegrown leader in AI, recruiting talent from top Chinese universities and paying handsomely, on the scale of local tech heavyweights like TikTok parent ByteDance.

In a July 2024 interview with Waves republished in the China Academy, Liang spelled out some of his ambitions for DeepSeek and China’s overall AI strategy.

“For years, Chinese companies have been accustomed to leveraging technological innovations developed elsewhere and monetizing them through applications,” he said. “But this isn’t sustainable. This time, our goal isn’t quick profits but advancing the technological frontier to drive ecosystem growth.”

He acknowledged that innovation requires a lot of money and said China previously adopted existing technology because its economy was in an earlier stage of development.

But that’s not the case now, as ByteDance and Tencent have become global players and are generating immense profits.

“What we lack isn’t capital but confidence and the ability to organize high-caliber talent for effective innovation,” Liang said.

DeepSeek’s open-source model contrasts with OpenAI’s, but he doesn’t consider sharing his company’s breakthroughs as a disadvantage.

Adopting a closed-source model won’t prevent competitors from catching up, Liang added, and being open-source actually provides an edge.

“Therefore, our real moat lies in our team’s growth—accumulating know-how, fostering an innovative culture,” he explained. “Open-sourcing and publishing papers don’t result in significant losses. For technologists, being followed is rewarding. Open-source is cultural, not just commercial. Giving back is an honor, and it attracts talent.”

Innovation is a matter of belief

Liang also described DeepSeek as a bottom-up company where a natural division of labor emerges without any preassigned roles or rigid hierarchy, facilitating free collaboration. But once an idea shows potential, management allocates resources from the top down.

DeepSeek also doesn’t place limits on accessing computing resources or personnel, he added, saying anyone with a good idea can tap into “training clusters” anytime. Even meeting rooms can be easily accessed to encourage serendipitous exchanges between colleagues and creative connections.

“I believe innovation is, first and foremost, a matter of belief,” he said. “Why is Silicon Valley so innovative? Because they dare to try. When ChatGPT debuted, China lacked confidence in frontier research. From investors to major tech firms, many felt the gap was too wide and focused instead on applications. But innovation requires confidence, and young people tend to have more of it.”