Nvidia CEO Jensen Huang has raised alarm about a potential shift in how China develops artificial intelligence, warning that if Chinese AI startup DeepSeek optimizes its models on Huawei's homegrown chips instead of American technology, it could fundamentally alter the global AI power balance. The concern centers on what happens when AI development moves away from American-designed systems and toward Chinese alternatives, potentially creating a technological divide that could reshape the industry for decades .

What Would Happen If China Built AI Models on Its Own Chips?

Huang's warning stems from a specific technical reality: AI models are typically optimized for the hardware they run on. When a model like DeepSeek's V4 foundation model is fine-tuned for Huawei's Ascend 950PR processor rather than Nvidia's graphics processing units (GPUs), it creates what engineers call a "different tech stack." This means the model learns to work best with Chinese hardware, making it harder to port to American systems and vice versa .

The implications extend beyond a single company's product. If future AI models are optimized in fundamentally different ways than American technology, and as "AI diffuses out into the rest of the world" with Chinese standards and technology embedded in them, Huang argued that China "will become superior to" the United States in artificial intelligence capabilities . This isn't about one model being better than another; it's about which country's technological ecosystem becomes the global standard.

Huang

Why Is DeepSeek's V4 Model Such a Flashpoint?

DeepSeek's upcoming V4 model, expected to launch later in April 2026, represents a critical test case. According to reporting from The Information, V4 would run on Huawei's latest Ascend 950PR processor, marking a significant shift from the company's previous approach . DeepSeek's earlier V3 model, launched in late 2024, relied entirely on 2,048 Nvidia H800 graphics processing units (GPUs), which are older Hopper-generation chips that were specifically designed to circumvent US export controls .

The timing matters because regulatory restrictions on chip exports to China have recently eased under the Trump administration. Nvidia restarted production of the H200, a more powerful chip, for sale in China, according to statements Huang made in March 2026 . Yet even as American chips become available again, Chinese companies are racing to build alternatives that don't depend on them at all.

How to Understand the Strategic Stakes of AI Chip Independence

• Technology Lock-in: Once AI models are optimized for specific hardware, switching to different chips becomes expensive and difficult, similar to how software ecosystems create vendor lock-in that lasts for years.
• Global Standards Setting: Whichever country's chip architecture becomes the foundation for most AI models effectively sets the technical standards that the rest of the world must follow or adapt to.
• Energy and Research Advantages: China has abundant energy resources and a large pool of AI researchers, giving it structural advantages if it can develop models that don't depend on American technology .
• Export Control Circumvention: By building its own chips and optimizing models for them, China reduces its vulnerability to future US export restrictions, creating a more resilient technology ecosystem.

The broader context reveals why Huang's warning carries weight. The US has spent years implementing export controls on advanced chips to China, restricting sales of Nvidia's most powerful processors. Yet these restrictions have proven difficult to enforce completely. Reuters reported last month that DeepSeek's V4 had been trained on Nvidia's Blackwell chips, which would violate US export controls if confirmed . The cat-and-mouse game between American restrictions and Chinese workarounds continues to escalate.

What makes the Huawei chip scenario different is that it represents a permanent solution rather than a temporary workaround. If DeepSeek and other Chinese AI companies successfully optimize their models for Huawei hardware, they create a self-reinforcing cycle: more models optimized for Chinese chips attract more investment in Chinese chip development, which in turn makes those chips more attractive to other AI companies. Over time, this could establish a parallel AI ecosystem that operates independently of American technology.

"If future AI models are optimized in a very different way than the American tech stack, and as AI diffuses out into the rest of the world with Chinese standards and technology, China will become superior to the United States," said Jensen Huang, CEO of Nvidia.

Jensen Huang, CEO at Nvidia

The stakes extend beyond corporate competition. The country whose technology standards become embedded in the most widely used AI models gains significant geopolitical leverage. It influences which companies can participate in the AI economy, which nations can develop advanced AI capabilities, and ultimately which technological vision shapes how artificial intelligence develops globally. Huang's warning suggests that American policymakers should view the Huawei chip question not as an isolated technical development but as a potential inflection point in the global AI race.