Huawei Accelerates Domestic AI Push with New Supernode Clusters

Huawei has unveiled its Atlas 950 and 960 AI superclusters, marking a strategic shift toward self-reliance in artificial intelligence infrastructure. The systems, powered by Ascend NPUs, aim to deliver 1 million neural processing units by 2027—a direct response to U.S. restrictions on AI chip exports.

The company's SuperPoD Interconnect technology mirrors Nvidia's NVLink, enabling large-scale chip clustering for high-performance computing. This development positions Huawei as China's foremost alternative to Western AI hardware providers amid escalating geopolitical tensions.

Concurrent with the hardware launch, Huawei's cloud division restructuring signals a pivot toward AI-driven services. The MOVE comes after years of financial strain from U.S. sanctions, with the firm now betting heavily on domestic AI ecosystems to regain technological sovereignty.