• Chinese vendors secured 41% of China’s domestic AI accelerator market in 2025.
  • Nvidia still leads with 55%, but share is declining as localisation accelerates.

What happened

Chinese GPU and AI chipmakers accounted for 41% of China’s AI accelerator server market in 2025. Domestic vendors shipped approximately 1.65 million accelerator units, out of a total 4 million AI accelerator cards shipped in China’s market. This confirms that the shift is within China’s domestic ecosystem, rather than the global market.

Nvidia remained the largest supplier, delivering around 2.2 million units, representing 55% market share, while AMD held roughly 4% with 160,000 units shipped.

Among Chinese firms, Huawei led with 812,000 units, followed by Alibaba’s T-Head with 265,000 units. Baidu’s Kunlunxin and Cambricon each shipped about 116,000 units. Other players, including Hygon, MetaX and Iluvatar CoreX, contributed smaller shares.

The growth reflects strong domestic demand for AI infrastructure and policy support encouraging adoption of locally developed chips, following US export restrictions on advanced semiconductors. China has also accelerated AI infrastructure investment, with local governments expanding computing centres that often favour domestic suppliers.

Why it’s important

The data highlights a critical shift: China’s AI chip market is becoming increasingly self-reliant. The rise to 41% share in China’s AI accelerator server market shows that local suppliers are no longer secondary players but core infrastructure providers.

For Nvidia, the impact is structural. China has historically been one of its most important overseas markets. A decline to 55% market share signals a gradual but persistent erosion of its dominance, particularly as procurement policies and ecosystem development favour domestic alternatives.

From a supply chain perspective, the shift reflects a broader decoupling trend. With 1.65 million locally supplied accelerator cards, domestic firms are filling gaps left by restricted access to advanced US chips. Performance gaps remain, but availability, policy alignment and system integration are becoming decisive factors.

Financially, this could weigh on Nvidia’s long-term market position in China, while strengthening growth visibility for domestic chipmakers supported by policy backing and expanding AI infrastructure investment.

More broadly, the development signals that AI computing is fragmenting into regional ecosystems. China is building a more self-reliant supply chain, reducing dependence on US technology while accelerating domestic innovation.