The Insane Journey of a Quant Genius Who Bet Everything to Build “Human-Level AI” for the Masses
- Wall Street’s Loss, AI’s Gain: Liang Wenfeng abandoned a lucrative finance career to solve a bigger problem—making AI accessible to everyone, not just elites.
- 1/20th the Cost, Equal the Power: DeepSeek’s AI models rival OpenAI and Anthropic at a fraction of the price, thanks to groundbreaking efficiency hacks.
- Mission Over Money: Liang rejected a $10B buyout, proving DeepSeek isn’t just a company—it’s a revolution.
Five years ago, Liang Wenfeng walked away from Wall Street—a world where his AI-powered hedge fund, Huanfang Quantitative, had mastered the art of predicting markets with “scary accuracy.” By 2015, at just 30, he’d launched High-Flyer, a hedge fund now managing $8 billion in assets. But Liang wasn’t chasing wealth; he was chasing a dream. “Human-level AI for everyone,” he insisted, even as Wall Street skeptics scoffed.
His pivot wasn’t impulsive. Liang had spent years observing how AI’s transformative power was gatekept by financial elites and tech giants. “AI shouldn’t be a luxury,” he argued. In 2021, he made a jaw-dropping bet: purchasing 10,000 Nvidia H800 chips—a move that stunned competitors. Leveraging his hedge fund team’s expertise in GPU optimization, Liang laid the groundwork for DeepSeek, an AI startup born in 2023 with a tiny team and an audacious goal.
The Underdog Strategy: How DeepSeek Outpunched Silicon Valley
US export restrictions on advanced chips threatened to derail China’s AI ambitions, but Liang’s team turned constraints into fuel. Recruiting dozens of PhDs from Peking, Tsinghua, and Beihang universities—and paying salaries rivaling tech titans like Bytedance—DeepSeek developed radical methods to train models like DeepSeek-V3 and r1 LLM. Their secret? Relentless efficiency.
By reimagining training protocols and hardware workflows, DeepSeek slashed costs to ~1/20th of Western counterparts while matching performance. Critics questioned the speed of their rise, but Liang responded with transparency: the company co-authored a research paper with 200+ contributors, openly sharing methodologies. “Innovation thrives in the light,” he said. The result? Models now used by hedge funds, startups, and everyday users—topping the App Store and defying geopolitical barriers.
The $10B Rejection and the Depths of Ambition
In 2023, Liang faced a defining moment: a $10 billion acquisition offer. He turned it down flat. “DeepSeek isn’t for sale. It’s a mission,” he declared. For Liang, the stakes transcend profit. With 38% of top-tier U.S. AI talent being Chinese (per 2022 data), he sees DeepSeek as a bridge—a “local” champion with global impact, democratizing AI while navigating a fractured tech landscape.
When he’s not coding or strategizing, Liang explores deep-sea caves—a metaphor he embraces. “Answers lie beneath the surface,” he says. It’s this mindset that drives DeepSeek’s relentless execution. While rivals chase scale, Liang’s team focuses on depth: refining algorithms, empowering users, and proving that AI’s future isn’t just for the few—but for anyone daring to dig deeper.