China Telecom deployed a single training cluster with 11,520 Huawei Ascend 910C chips in Guangdong, achieving 9,000 PFLOPS of compute power. The deployment aims to break overseas high-end compute supply constraints, though no training efficiency benchmarks against Nvidia were announced.
China is building large-scale domestic AI compute infrastructure with Huawei chips, reducing reliance on foreign supply for critical training capacity.