Ascend 910C Vs NVIDIA H100

Jensen Huang recently acknowledged that Huawei has emerged as NVIDIA’s most significant competitor in the Chinese AI compute market — a space where NVIDIA has long dominated.

The tipping point? Huawei’s Ascend 910C, a next-gen AI accelerator that’s reportedly capable of rivaling — and in some workloads, even surpassing — NVIDIA’s H100 in performance. Combined with native software stack advancements like the CANN and MindSpore frameworks, Huawei is positioning itself as a viable alternative for China’s large-scale AI training and inference needs.

This comes at a critical moment. Due to ongoing U.S. export controls, NVIDIA is barred from selling its most advanced GPUs (like the A100, H100, and even the cut-down H800) to Chinese hyperscalers and research labs. The market vacuum is now being filled by domestic players like Huawei and Biren.

NVIDIA now faces a strategic dilemma:
1. Double down on non-restricted markets (e.g., Europe, MENA, India)
2. Lobby for regulatory flexibility to re-enter the Chinese market

But is there a third scenario?

Could NVIDIA develop custom silicon that complies with export regulations but still meets a competitive performance bar for Chinese customers? Or will China’s end-to-end AI stack independence (from silicon to software) accelerate beyond the point of no return?

The AI compute race is no longer just about FLOPS — it’s geopolitical.

What’s your take?

hashtag#NVIDIA hashtag#Huawei hashtag#Ascend910C hashtag#H100 hashtag#AIHardware hashtag#Geopolitics hashtag#Semiconductors hashtag#ExportControls hashtag#ChinaAI hashtag#TechStrategy