China AI chip certification adds 9 domestic AI processors to state procurement

China AI chip certification has taken a meaningful step forward, and it could reshape which companies supply the country’s most sensitive public-sector systems. China has certified nine domestically designed AI processors for state procurement, opening a new route for homegrown chips to move deeper into government agencies and state-owned enterprises.

That matters because this is not just another product list. The approvals create a new “AI training and inference chips” category under the Anke security certification framework, effectively bringing AI accelerators into a procurement system tied to Beijing’s broader push to replace foreign technology in key IT environments.

The result is a clearer map of who is in, who is out, and where China’s AI hardware race is heading next. Huawei and Alibaba made the cut. So did several rising domestic chip developers. However, some notable names were missing, highlighting how policy momentum and manufacturing limits are shaping the market as much as raw chip performance.

China expands its AI chip procurement list

China certified nine domestically designed AI processors for state procurement, with the approvals issued by the China Information Technology Security Evaluation Centre and the National Secrecy Science and Technology Evaluation Centre.

The certifications are valid for three years. More importantly, they create a new “AI training and inference chips” category under the Anke security certification framework.

This China AI chip certification move gives government buyers and state-linked entities a more formal route to source domestic AI processors. In practice, the list functions as a procurement catalog for agencies, central state-owned enterprises, and other organizations covered by the Xinchuang initiative.

That is why this matters beyond the semiconductor industry itself. Xinchuang has long focused on replacing Western hardware and software in sensitive Chinese IT systems. By adding AI processors more directly into this system, China is expanding domestic substitution from traditional computing infrastructure into the AI stack.

Which chips and makers were approved

The approved list spans some of the country’s best-known AI hardware companies.

Included in the certification are:

• Huawei’s Ascend 310 and Ascend 910
• Alibaba’s T-Head Zhenwu M530 and T-Head Zhenwu M890
• Chips from Biren Technology, Hygon Information Technology, Iluvatar CoreX, MetaX, and Moore Threads

The breadth of the list stands out. It shows that China is not relying on a single national champion for AI compute in public-sector procurement. Instead, the China AI chip certification process is widening the field, giving multiple domestic vendors a state-recognized route into official purchasing channels.

That could have practical effects quickly. Procurement catalogs often shape real buying behavior, especially when agencies need products that fit compliance and security rules. A chip’s presence on this list may not guarantee orders, but it sharply improves its position in state-linked markets.

Why the Huawei Ascend chips matter here

Huawei’s Ascend 310 and Ascend 910 were among the approved products, which gives the company another foothold in China’s state procurement system. At the same time, Alibaba’s inclusion shows that the approved field extends beyond telecom and networking leaders into broader domestic AI hardware competition.

Who was left out and why it matters

Two prominent names were absent: Cambricon Technologies and Baidu-backed Kunlunxin.

Cambricon’s omission is especially noticeable because it had appeared on a December Xinchuang procurement list for AI processors, a separate state approval mechanism. This time, however, it was not included in the Anke certification group. Kunlunxin also did not appear.

The missing names matter because the certifications function as a de facto Xinchuang procurement list for a wide range of government and state-owned buyers. If a company is not on the approved list, it risks losing visibility in one of the most strategically important parts of China’s AI chip market.

At the same time, the absence does not answer every question. The available information does not establish why Cambricon or Kunlunxin were not listed. What is clear is that each chip must pass Anke V3.0 testing requirements, and the new category now sets another gate for vendors seeking state procurement access.

What China AI chip certification signals for the market

This is also a sign of how fast China’s domestic AI chip ecosystem is maturing.

Earlier, Beijing’s efforts in Xinchuang were more closely associated with replacing Intel and AMD CPUs or Oracle databases in government systems. AI accelerators are now being folded into that same substitution logic. That expands the strategic role of domestic AI chips from commercial alternatives to approved infrastructure for state computing needs.

There is also a competitive message here. Chinese chipmakers are gaining domestic market share against Nvidia, and policy support appears to be reinforcing that trend. Once official certifications and procurement channels align with domestic supply, local vendors gain more than symbolic support. They gain a route to sustained deployment.

What the move signals for China’s AI chip market

The market backdrop helps explain why this development is drawing attention.

Chinese semiconductor firms delivered 1.65 million AI GPUs in 2025, out of a total 4 million units, accounting for 41% of local AI server shipments. Huawei alone shipped roughly 812,000 AI chips and is projecting $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 roughly 76% of demand.

Those numbers point to a simple reality: state certification is arriving at a moment when domestic suppliers are no longer niche contenders. They are already taking a sizable share of the local market. The procurement framework now gives that momentum a stronger institutional base.

That is the second big reason this matters. For investors, suppliers, and enterprise buyers, the story is no longer just whether China can design AI chips. It is whether those chips can secure enough certified demand and enough manufacturing capacity to scale.

SMIC wafer capacity remains the bottleneck

Even with policy support and rising market share, the supply side remains tight.

SMIC is still a major production bottleneck, with certified chipmakers competing for limited advanced wafer capacity. Its most advanced stable node is the N+2 process. SMIC reported utilization rates above 93% for 2025 and spent $8.1 billion in capital expenditure last year, with plans to maintain that level through 2026.

That creates a tension at the heart of the China AI chip certification story. Demand signals are getting stronger. Procurement access is getting clearer. But fabrication capacity is still constrained.

In other words, certification can open doors, but it cannot by itself produce more wafers. If more approved chips are pushed into state procurement channels while SMIC wafer capacity stays tight, the biggest winners may be the companies best positioned to secure production slots rather than simply those with the strongest product portfolios.

For now, the list does more than identify approved processors. It shows how China is building a domestic AI hardware system through regulation, procurement, and industrial policy at the same time — and how that push may be limited less by demand than by the factories needed to fill it.