Loading article…
China has certified nine homegrown AI processors for state use, marking a major expansion of the Xinchuang initiative to reduce reliance on Western hardware.
China’s official technology security bodies have certified nine domestically designed AI processors for state procurement, establishing a new category for "AI training and inference chips" under the country's Anke security framework [1]. This move marks the first time that artificial intelligence hardware has been formally included in the nation's "secure and reliable" technology assessment system [4].
Key takeaways
The Xinchuang program, a long-standing campaign to replace foreign hardware and software in sensitive Chinese IT systems, has historically focused on CPUs and database software [1]. By creating a specific category for AI training and inference, Beijing is signaling a pivot toward securing the infrastructure required for advanced artificial intelligence [4]. While Huawei and Cambricon were previously included in a separate state approval mechanism in December, the current list represents a significant expansion, now covering nine distinct chips from seven different vendors [1].
Notably, some prominent developers were absent from this latest certification. Both Cambricon Technologies and the Baidu-backed firm Kunlunxin did not appear on the list [3]. An anonymous source indicated that companies may choose whether to submit their products for testing, meaning that exclusion from the list does not necessarily imply a failed evaluation [1]. Each chip must pass rigorous tests under the Anke V3.0 requirements to qualify for the three-year certification [1].
Despite the push for technological independence, domestic firms face ongoing hurdles in transitioning away from established Western hardware. While Chinese semiconductor firms are increasing their market share, they remain constrained by limited wafer fabrication capacity [1]. Manufacturers are competing for production slots at SMIC, which currently utilizes its N+2 process—roughly equivalent to 7nm—to produce these chips [1].
Furthermore, practical implementation remains a challenge for state-linked entities. Some institutions have reported difficulty adapting systems originally built on Nvidia hardware to domestic alternatives, with some executives noting that rewriting trading models for new architectures requires significant effort [2]. Despite these "growing pains," policymakers continue to emphasize that reducing reliance on foreign technology is a critical national security priority [2].
Coverage is mostly measured — 32 of 45 reports stay neutral.
Every Monday — the token unlocks, Fed dates & catalysts set to move crypto and markets this week. So you’re never blindsided.
Free · 3-min read · one-click unsubscribe
The RTX Spark is a system-on-chip (SoC) developed by Nvidia and MediaTek that combines a Blackwell GPU and an Arm-based CPU to run AI models locally on PCs.
Nvidia is partnering with MediaTek for chip design and with Microsoft, Dell, HP, ASUS, Lenovo, and MSI to integrate the chips into upcoming Windows PCs.
Nvidia is seeking to expand its AI footprint to the 'edge,' allowing advanced AI agents to run locally on consumer devices without needing constant cloud connectivity.
The formalization of this procurement list underscores Beijing's resolve to build a parallel technology stack capable of supporting its AI ambitions despite U.S. export controls [3]. As China continues to invest in its domestic semiconductor ecosystem, the government is effectively locking in a preference for local silicon in state-run data centers [3]. With Huawei projecting $12 billion in AI processor revenue for 2026 and domestic supply expected to cover a larger share of local demand, the move reflects a long-term strategy to insulate the nation's AI infrastructure from future international policy shifts [1].
AI-assisted synthesis by the TrendWatcher Editorial Desk · sourced from 4 outlets · May 31, 2026 · How we report
The chip uses unified memory, which allows the CPU and GPU to access the same memory pool, eliminating bottlenecks and enabling the execution of larger AI models.