After more than a week of uncharacteristic silence, Elon Musk has finally spoken about DeepSeek, the Chinese AI startup that has been making headlines for its low-cost, high-performance AI model. But instead of praise, Musk has mocked DeepSeek’s claims, casting doubt on its rapid success and suggesting that the company isn’t being entirely transparent about its hardware resources.
While many tech leaders have celebrated DeepSeek’s AI breakthrough, Musk remains highly sceptical. When Salesforce CEO Marc Benioff applauded DeepSeek’s achievement—calling it proof that AI’s true value lies in data, not expensive hardware—Musk responded with a blunt and dismissive “Lmao, no.”
Musk also took a jab at DeepSeek’s Chinese origins. When a user joked that DeepSeek’s AI model, R1, was “leaked from a lab in China”, Musk replied with a laughing emoji, an apparent reference to past controversies surrounding China’s role in the spread of Covid-19.
Accusations of hidden GPU power
One of Musk’s biggest criticisms revolves around DeepSeek’s GPU usage. The company claims to have trained its model using around 10,000 Nvidia A100 GPUs, a relatively modest amount compared to what OpenAI or Anthropic require. However, Musk and Scale AI CEO Alexandr Wang believe the real number is much higher.
Wang, during an interview with CNBC, speculated that DeepSeek actually has around 50,000 Nvidia H100 GPUs, but cannot publicly admit it due to US export restrictions on advanced chips. Musk agreed with Wang’s theory, responding with a simple “Obviously”, implying that DeepSeek isn’t telling the full story about its hardware resources.
Musk’s dismissive attitude toward DeepSeek contrasts with the reactions of other industry leaders. While Benioff praised it as a “Deepgold” moment, Microsoft CEO Satya Nadella suggested that cheaper AI would only accelerate global adoption, and OpenAI CEO Sam Altman admitted that DeepSeek had built an “impressive model” for its price.
But Musk—who has his own AI company, xAI, which recently launched Grok AI—seems unwilling to accept DeepSeek’s success at face value. Whether his scepticism is justified or just competitive rivalry remains to be seen.