In a leaked four-hour investor presentation, DeepSeek founder Liang Wenfeng argues that compute access—not software moats—defines the US-China AI gap, while claiming Nvidia's CUDA dominance is eroding.
DeepSeek's reclusive founder laid out his vision of artificial intelligence development in rare remarks to investors, painting a picture of the competitive landscape between US and Chinese AI firms.
Liang's core argument centers on a straightforward constraint: the primary disadvantage facing Chinese AI companies stems from restricted access to computing hardware, particularly high-end chips. This hardware limitation, rather than software capabilities or algorithmic innovation, represents the main barrier to closing the gap with American competitors.
On Nvidia's position, Liang contends that the semiconductor giant's long-standing advantage through its CUDA software ecosystem is weakening. As alternative frameworks and tooling mature, the lock-in effect that has protected Nvidia's market position diminishes, according to the DeepSeek founder.
The leaked remarks reveal Liang's philosophical approach to AI development. He frames artificial general intelligence as an inevitable tide—a force no single company can monopolize or control. This perspective suggests a view of AGI as an outcome of broader technological progress rather than proprietary advantage.
The four-hour talk provides unusual insight into how China's leading AI startup assesses competitive dynamics. DeepSeek has emerged as one of China's most capable AI labs, and Liang's comments reflect the company's positioning within ongoing US-China technology competition.
The leak underscores ongoing tensions around AI development and export controls. US restrictions on chip sales to China have been a central policy tool, making Liang's emphasis on compute access particularly relevant to current geopolitical discussions.
DeepSeek's public profile has grown substantially as Chinese AI capabilities advance. The company's technical achievements, combined with Liang's rare public commentary, highlight the accelerating pace of AI development outside the US market.
Bipartisan lawmakers introduced the AI Kill Switch Act on Thursday, which would grant the Department of Homeland Security authority to shut down or throttle AI models deemed dangerous.
Major AI companies are recruiting top university researchers at an accelerating pace, shifting research from public institutions to private industry labs. This brain drain is transforming once-open academic work into proprietary, closed-door projects.
Moonshot AI founder Yang Zhilin's departure from the US reflects a broader shift: Chinese entrepreneurs increasingly see greater opportunities in their home market than in Silicon Valley.
Chinese AI models have cut their performance deficit against US counterparts to 6% in June, down from 9% in May, according to Bloomberg Intelligence analysis. The rapid convergence challenges assumptions about American technological dominance in artificial intelligence.