NVIDIA CEO Jensen Huang warned on April 18, 2026, that China is poised to become superior to the United States in artificial intelligence (AI) if future models are optimized differently than the American tech stack.
As of early Tuesday, Huang highlighted significant developments involving DeepSeek, a Chinese AI chatbot. DeepSeek plans to shift its operations from Nvidia’s CUDA platform to Huawei’s CANN platform. This transition signifies a pivotal moment in the ongoing tech rivalry.
Within hours of Huang’s comments, it was revealed that DeepSeek is set to roll out its V4 multimodal foundation model on Huawei’s Ascend 950PR processor. This processor reportedly possesses around 60% of the power compared to Nvidia’s H100.
Huang pointed out that despite hardware deficiencies, China has an abundant supply of energy and a large pool of AI experts. These factors could significantly bolster its domestic development in AI technologies.
Huang also noted that U.S. export policies aimed at curbing China’s AI capabilities have inadvertently boosted its domestic development. “If future AI models are optimized in a very different way than the American tech stack, China will become superior to the U.S.,” he stated emphatically.
During a recent debate about selling chips to China, Huang nearly lost his composure, reflecting the high stakes involved. He remarked, “You’re not talking to someone who woke up a loser,” underscoring his confidence in Nvidia’s position.
But he cautioned against creating two separate ecosystems—one open-source and reliant on foreign tech stacks, and another closed off to American technologies. Huang questioned, “Why are you causing one layer of the AI industry to lose an entire market so that you could benefit from another layer?” His comments have sparked discussions across tech circles about the future of AI development.
Details remain unconfirmed regarding how these shifts will impact global AI dynamics. However, it is clear that the competition between the US and China in this field is intensifying rapidly.