DeepSeek, a Chinese AI startup, is making waves with its claim of rivaling OpenAI and Meta's large language models for a mere US$6 million. While this assertion has drawn skepticism, it has also heightened curiosity about DeepSeek's founder, Liang Wenfeng.
Western media have dubbed Liang the "Chinese Sam Altman" and described his transformation from a mathematical genius into a global AI disruptor. However, a more fitting narrative might be his evolution from a "stock market prodigy" into an "AI visionary."
The path from quant trading to AI leadership
Born in 1985 in Wuchuan, Zhanjiang, Guangdong Province, Liang excelled academically from a young age. He reportedly self-studied calculus in junior high school before earning admission to Zhejiang University, where he majored in information science and electronic engineering.
During his university years, Liang began developing AI-driven stock selection algorithms. After graduating in 2013, he founded an investment firm, and in 2015, he co-founded the quant hedge fund High-Flyer with two university classmates. The firm became one of China's earliest pioneers in applying AI to stock trading, using algorithms to analyze data beyond market prices and identify hidden patterns.
However, Liang's dream of stock market success faced significant setbacks. In 2021, High-Flyer apologized to investors for misjudging market trends, and later, China's financial regulators cracked down on quantitative trading, dealing a significant blow to the company.
The pivot to AI
As early as 2019, Liang's team was already using Nvidia GPUs for computing. By the time OpenAI launched ChatGPT in late 2022, High-Flyer was among the limited number of Chinese companies possessing over 10,000 high-end Nvidia chips.
Liang had revealed that he accumulated 10,000 GPUs by 2021. Yet at that time, he saw AI only as a side project, and the Chinese tech industry treated his AI ambitions as just another billionaire's hobby.
Everything changed in 2023 when Liang officially founded the DeepSeek research lab, setting his sights on developing general AI. DeepSeek leveraged High-Flyer's early GPU acquisitions, giving it a strategic edge over other companies. The release of DeepSeek V2, known for its low-cost pricing, earned it the nickname "the Pinduoduo of AI." By December 2024, DeepSeek V3 secured a top position in China's mainstream large-model rankings.
Innovation or disruption?
Many believe that Liang's pivotal decision was open-sourcing DeepSeek's code, aiming to challenge the dominance of major tech firms. Industry insiders have noted DeepSeek's AI development strategy mirrors the principles of quantitative trading—achieving maximum results with minimal resources, including a lean team, limited capital, and fewer chips.
Those familiar with Liang describe him as a hands-on leader obsessed with refining algorithms, making them more intelligent in stock trading and AI models. Colleagues say he embodies the typical Chinese engineer--unconcerned with fashion or appearance, relying on formulas and calculations to make decisions, and privately passionate about soccer.
Despite his low profile, DeepSeek's rapid rise thrust Liang into the spotlight. The overwhelming surge of users led to multiple system crashes and cyberattacks, forcing him to delay his 2025 Lunar New Year holiday to manage the crisis. He is expected to return after the holiday to lead the development of DeepSeek's next-generation model.
Meanwhile, DeepSeek's sudden ascent has drawn scrutiny in the US. Public sentiment has shifted from admiration to suspicion, with some branding the startup a "serious threat" to American AI dominance.
The company's success has raised questions about the effectiveness of Washington's efforts to restrict China's access to advanced chips. Experts speculate that the US government may soon launch an investigation into DeepSeek's impact on the AI ecosystem, and the Department of Commerce could pressure Apple and Google to remove its app from US app stores.