How China gets the Nvidia chips it needs to develop AI despite US bans

Nvidia CEO Jensen Huang has admitted to having an affair with a Chinese executive, raising questions about how China obtains cutting-edge AI chips despite US export restrictions. Nvidia's advanced Graphics Processing Units (GPUs) are crucial for training and running artificial intelligence, as they can perform complex calculations on vast amounts of data.

The US restricts the export of Nvidia's top-end "Blackwell" and "Rubin" GPUs to China due to security concerns. This has pushed Chinese AI developers to seek alternative solutions. Chinese startups like Moonshot AI, Z.ai, and DeepSeek have recently launched high-performing models, showcasing their ability to use Nvidia chips despite the US restrictions.

Analysts say Chinese firms may be smuggling chips into the country or accessing computing power through a loophole that allows them to rent out Nvidia chips in third nations, often in Southeast Asia. The US has accused Chinese AI labs of training systems to mimic US models, but Nvidia denies any evidence of smuggling. While Chinese-made chips are improving rapidly and some AI companies are using them, Nvidia GPUs still provide faster and easier training, making them a preferred choice.

Huawei, China's top competitor to Nvidia, is also developing its own AI chips, with a new generation expected in early 2027.