Nvidia’s China AI chip share nears zero as local rivals gain

Nvidia’s position in China’s AI chip market has collapsed as U.S. export controls cut off official access to its most important products. The company once controlled over 95% of China’s high-end AI chip market, with A100/H100 chips and the CUDA ecosystem widely used by internet platforms, supercomputing centers and AI clusters across finance, healthcare and autonomous driving.

Restrictions beginning in 2022 blocked exports of A100 and H100 chips, later covering downgraded A800 and H800 versions and tighter reviews of Asian clients. Nvidia’s China-specific H20 chip failed to regain momentum because of weaker training performance, uncertain supply and regulatory obstacles, leading Jensen Huang to plan for zero sales in China for the next several quarters.

Domestic suppliers are moving into the gap. Huawei’s Ascend 910B and 910C are described as outperforming Nvidia’s H20 in single-card computing performance, while Huawei’s CANN software stack is reducing migration costs. Hygon Information and Cambricon are also gaining traction in specialized uses. Bloomberg Intelligence found Chinese enterprises plan to allocate 46% of their AI accelerator budgets to local products over the next 12 months, up from the current 30%.

Huang still sees China as a critical market, with AI chip demand expected to grow from approximately $50 billion currently to hundreds of billions of dollars. But with domestic substitution strengthening and China historically accounting for more than 20% of Nvidia revenue, any return is likely to be gradual and centered on compliant products, partnerships and ecosystem integration rather than a restoration of its former dominance.