Huawei Prioritizes Domestic AI Chip Demand Amid Surge, Scales Atlas Clusters to 120 EFLOPS
Huawei has paused plans for an international rollout of its AI chips, redirecting resources entirely to the Chinese domestic market where demand for its hardware is outpacing supply. The company is rapidly scaling its Atlas AI compute clusters, with current deployments reaching 15,488 chips per cluster. These clusters leverage advanced optical networking technology to enable high-bandwidth, low-latency communication between processors—positioning them as competitive alternatives to Nvidia's offerings in the Chinese market. Reports indicate that the total compute capacity across Huawei's Atlas infrastructure is scaling toward 120 EFLOPS. The strategic pivot underscores how U.S. export restrictions have accelerated China's development of indigenous AI silicon, with domestic champions like Huawei now fielding solutions that aim to meet the country's growing appetite for AI training and inference hardware.