They have nothing, it's not used for anything other than inference. The entire Chinese infrastructure relies on Nvidia GPUs, and while everyone else is advancing at breakneck speed, they're just trying to make it work.
At Huawei Connect 2025, the company introduced the Atlas 950 SuperCluster, delivering 524 FP8 ExaFLOPS for AI training and 1 FP4 ZettaFLOPS for inference. Built with over 500,000 Ascend 950DT NPUs, the system outpaces Oracle's and xAI's clusters while challenging Nvidia's upcoming Rubin platform.