In a flurry of announcements, China's domestic AI compute infrastructure is landing in quick succession. Sugon unveiled the Shuguang 8000 (Dengfeng), the country's first fully domestic 100,000-card AI supercomputing cluster, using immersion phase-change liquid cooling to solve high-density heat dissipation.
At HUAWEI CONNECT 2026, Huawei released the Ascend 960 supernode — the world's first to adopt NPO technology — aimed at accelerating training and inference for models at the ten-trillion-parameter scale, and proposed an AI DC data-infrastructure reference architecture for scaled AI deployment.
The China Academy of Information and Communications Technology (CAICT) said China has preliminarily met the conditions to study compute-power futures, with a market-based pricing mechanism yet to be built — signaling the maturing of the domestic compute ecosystem.
Analysts note the moves come as global competition over communications and compute infrastructure escalates, with European carriers mulling a coalition against Starlink and Nvidia pushing glass-substrate packaging — underscoring a widening, multi-front technology race.
