China Accelerates Domestic AI Chip Development Amid US Restrictions

China is ramping up investment in home‑grown artificial‑intelligence chips as part of a broader push for technological sovereignty. Annual spending on AI accelerators is slated to rise from about 30% to 46% within the next year, and the government plans to allocate roughly $300 billion for new data centers, insisting that at least 80% of components—including AI chips—be domestically sourced. Major Chinese firms such as Alibaba, Tencent, Hygon, Cambricon and Huawei are racing to launch competitive products, with Huawei unveiling the Ascend 950/960 chips and the Atlas 950 supernode cluster aimed at training and inference workloads.

In parallel, Huawei published an official document outlining its "Tau Law" proposal, a new scaling approach that replaces the traditional transistor‑size metric with a time‑constant (τ) based target. The paper cites a 55% increase in transistor density and a 41% power‑efficiency improvement in a mobile SoC prototype, and projects a hundred‑fold hardware integration growth for AI systems by 2035. The initiative is presented as an alternative to the slowing returns of Moore’s Law and underscores China’s effort to build an independent semiconductor ecosystem.