China narrows Nvidia’s public-sector path with domestic AI chips
China’s AI chip procurement list now includes certified domestic options for the first time.📷 AI-generated image / TECH&SPACE
- ★China has added domestic AI chips to its secure government procurement list for the first time.
- ★Nine options received certifications that remain valid for three years.
- ★The move strengthens the domestic supply chain and continues the institutional shift away from Nvidia.
China has added domestic AI chips to its “secure and reliable” government procurement list for the first time, according to Tom’s Hardware. On the surface, this is a certification update. In practice, it is a clear signal that China’s public sector wants more of its artificial intelligence computing stack to rest on domestic hardware rather than on foreign accelerator suppliers.
Nine domestic options have been included, with certifications valid for three years. The approvals were issued by the China Information Technology Security Evaluation Centre and the National Secrecy Science and Technology Evaluation Centre, giving the move a security dimension as well as an industrial one. A procurement list like this does not merely describe what can be bought. It defines what the state considers acceptable for systems where confidentiality, supply-chain control and long-term availability matter alongside raw performance.
Nine domestic options received three-year certifications for government procurement, further narrowing Nvidia’s room in China’s public sector.
Three-year certificates give domestic accelerators a formal path into the public sector.📷 AI-generated image / TECH&SPACE
The most important detail is not the number nine, but the first-time inclusion of domestic AI chips in this category. China is sending public institutions and related buyers an operational message: for AI workloads that can remain inside the domestic ecosystem, certified alternatives now exist. That does not mean these chips automatically replace every system built around Nvidia data-center GPU platforms, but it does mean domestic suppliers now have a formal procurement pathway that can shape purchasing decisions.
For Nvidia, this is another kind of pressure: not only competitive, but institutional. China’s demand for AI compute remains large, yet government procurement is increasingly a tool for steering technological sovereignty. If these certifications become a practical filter in purchasing, domestic chips do not need to beat Nvidia in every benchmark table to gain ground. They need to be approved, available and aligned with rules that foreign suppliers do not control.
The three-year validity period also matters. It gives buyers predictability and gives chipmakers a window to plan deliveries, support and follow-on products. In public-sector procurement, that can be more decisive than one impressive specification, because institutions are not just buying silicon; they are buying a maintainable supply chain. The role of the China Information Technology Security Evaluation Centre underlines that security assessment is part of the same equation as industrial policy.
This should therefore be read as a controlled but firm shift in the architecture of China’s AI infrastructure. It is not the dramatic launch of a single chip. It is the creation of a list that can redirect budgets, priorities and trust. In semiconductors, such lists often work quietly, but their consequences last: companies on the list enter planning cycles; companies outside it must explain why they deserve an exception.

