China adds homegrown AI chips to ‘secure and reliable’ procurement list for the first time

China has for the first time added nine domestically designed AI processors to its official "secure and reliable" procurement list, signaling an accelerated push to replace Nvidia hardware in state systems.

cnadmin
By
3 Min Read

China has for the first time added nine domestically designed AI processors to its official “secure and reliable” procurement list, signaling an accelerated push to replace Nvidia hardware in state systems.

New certification category

The China Information Technology Security Evaluation Centre and the National Secrecy Science and Technology Evaluation Centre jointly certified the chips under a new “AI training and inference” category within the Anke V3.0 security framework. Approved processors include Huawei’s Ascend 310 and 910, Alibaba’s T-Head Zhenwu M530 and M890, and offerings from Biren Technology, Hygon Information Technology, Iluvatar CoreX, MetaX, and Moore Threads. Certifications are valid for three years and effectively serve as a procurement mandate for government agencies, state-owned enterprises, and entities under Beijing’s Xinchuang initiative.

Notable absences and market dynamics

Cambricon Technologies and Baidu-backed Kunlunxin, two prominent domestic AI chip developers, did not appear on the list. Cambricon was included in a separate December procurement catalog, making its omission notable. An anonymous source indicated that companies may choose whether to submit products for testing, so exclusion does not necessarily imply a failed evaluation. Each chip must pass Anke V3.0 requirements to qualify.

Domestic gains at NVIDIA’s expense

Chinese semiconductor firms shipped 1.65 million AI GPUs in 2025 out of a total 4 million units, capturing 41% of local AI server shipments. Huawei alone shipped roughly 812,000 AI chips and projects $12 billion in AI processor revenue for 2026. Morgan Stanley estimates China’s AI chip market could reach $67 billion by 2030, with domestic supply covering about 76% of demand.

Manufacturing constraints remain

All certified chipmakers compete for limited production capacity at SMIC, whose most advanced stable node is the N+2 process, roughly equivalent to 7nm. SMIC reported utilization rates above 93% for 2025 and spent $8.1 billion in capital expenditures last year, with plans to maintain that level through 2026. Wafer fab capacity remains a binding constraint on domestic AI chip production.

The certification list marks a definitive step in China’s strategic decoupling from Western AI hardware, but sustained growth will depend on overcoming foundry bottlenecks. As domestic supply scales, Nvidia faces mounting headwinds in the world’s second-largest AI chip market.

Share This Article