Nine Domestic AI Chips Earn Top‑Tier Security Certification, Alibaba and Huawei Secure Two Slots Each

Why the Certification Is Called the "Coming‑of‑Age" for Computing Power

The certification, jointly issued by the China Information Security Evaluation Center and the National Confidential Technology Evaluation Center, marks the first time AI training and inference chips are included in the national "trust list". Previously, the list only covered CPUs, operating systems and databases, leaving AI chips as "outsiders". Adding AI chips provides two critical guarantees: data‑security compliance and sovereign control over the supply chain. Without the I‑level certificate, even high‑performance chips cannot be purchased for core national projects.

The Nine Certified Chips and Their Strengths

Huawei Ascend 310 : An edge‑inference processor marketed as the "cost‑performance king" for low‑power, high‑efficiency workloads such as smart cameras, industrial robots and vehicle‑mounted devices.

Huawei Ascend 910 : A flagship cloud‑training engine forming the backbone of Huawei’s AI ecosystem, designed for large‑model training. Huawei expects to ship 812,000 units in 2025 and generate over $12 billion in revenue in 2026.

Alibaba Pingtouge Zhenwu M530 / M890 : Alibaba’s self‑developed inference chips, optimized for Alibaba Cloud workloads. M530 targets lightweight inference, while M890 focuses on high‑performance scenarios like e‑commerce recommendation, content moderation and cloud gaming.

Biren Tech Walli™ 166 : A general‑purpose GPU that handles both AI training/inference and graphics rendering, breaking the overseas monopoly on universal GPUs.

HaiGuang DCU‑3G : A deep‑compute unit fully compatible with the CUDA ecosystem, offering minimal migration cost for data‑center AI and high‑performance computing workloads.

TianShu KCC‑V100X : A cloud‑training‑focused chip emphasizing high compute density for large‑model training.

Muxi MXC600 : An inference‑only processor delivering low latency and high throughput for massive AI inference clusters.

Moore Threads PH100 : A full‑function GPU that supports graphics rendering, AI computation and video codec acceleration, suitable for both consumer and data‑center applications.

Impact on the Domestic AI Chip Market

The certification instantly turns these chips into the "passport" for procurement in core sectors such as government, finance, telecom and energy. This shifts domestic AI compute from a "substitute" role to a "main‑force" position, directly challenging foreign giants.

Market data shows that in 2025 China will deliver 4 million AI GPUs, with domestic chips already accounting for 41 % of the shipments. Morgan Stanley forecasts the Chinese AI‑chip market to reach $67 billion by 2030, with domestic products expected to satisfy 76 % of the demand.

Broader Implications

Including AI chips in the trust list expands the "trusted" IT ecosystem from traditional hardware to core compute, creating a "trillion‑yuan" opportunity for domestic suppliers. Foreign chips not on the list will lose market share in key domestic projects, while domestic players not yet certified (e.g., Cambricon, Suiyuan, Baidu Kunlun) must accelerate development to stay competitive.

The certification also highlights three synergistic drivers for domestic AI compute: strong policy endorsement, exploding demand from AI‑agent applications, and a maturing ecosystem that reduces reliance on overseas technology.