China’s quest for technological self-reliance has reached a new milestone as the central government formally integrates artificial intelligence chips into its "Safe and Reliable" certification framework. This move, orchestrated by the China Information Security Evaluation Center and the National Secrecy Technology Evaluation Center, signals a decisive shift in how the nation’s massive state-led technology sector will procure the compute power necessary for the generative AI era. By awarding the highest "Level I" rating to nine domestic processors, Beijing is effectively drawing a circle around its preferred "national team" of semiconductor champions.
The list includes high-profile entries from the country’s most resilient tech giants, notably Huawei’s Ascend series and Alibaba’s T-Head (Pingtouge) division. Emerging contenders like Biren Technology and Moore Threads also secured spots, despite facing varying degrees of pressure from U.S. export controls. This certification serves as a powerful "seal of approval," assuring state-owned enterprises and government agencies that these components are verified for security, supply chain integrity, and—most crucially—independency from foreign intellectual property.
Notably, the certification is not merely a badge of honor; it is a strategic roadmap for the "Xinchuang" (Information Technology Application Innovation) sector. As Chinese entities are increasingly pushed away from Nvidia’s industry-standard GPUs due to American sanctions, these nine chips represent the government’s sanctioned path forward for training and inference workloads. The exclusion of previously celebrated players like Cambricon and Baidu’s Kunlun suggests that the criteria for "reliability" are becoming more stringent, prioritizing hardware that can truly withstand the volatility of global geopolitics.
The economic stakes are immense, with market projections suggesting that domestic chips could capture over 75% of China’s internal AI market by the end of the decade. As China’s AI server market expands to meet the demand of the country's local Large Language Models (LLMs), the formalization of this list provides a clear signal to investors and engineers. The era of "best-of-breed" global sourcing is giving way to a "security-first" domestic ecosystem, where compliance is now as valuable as raw performance.