On the 9th a domestically built smart-compute cluster exceeding 10,000
accelerator cards went live in Shaoguan as part of the Guangdong–Hong Kong–Macao
Greater Bay Area hub. Total investment is 5.5 billion yuan; the facility deploys
11,520 Ascend domestic AI accelerator cards, delivering aggregate compute of
9,000P and performance described as internationally leading. The cluster is part
of the national integrated compute network’s GBA hub and is intended to serve
the region’s AI industry by linking domestic chips, operating systems and
algorithm frameworks to complete the compute-to-application ecosystem and
promote scale adoption of domestic chips and training frameworks, a step the
operator says marks big-model R&D entering a “compute autonomy” phase.
