Huawei steps up AI chip race with Ascend 910D, targeting Nvidia’s high ground

Despite Washington’s efforts to curb Chinese access to leading-edge AI hardware, Huawei has made significant progress. “Huawei’s HiSilicon 910 series is a good outcome for China’s local AI server chip ambitions, which were accelerated by US export restrictions,” said Neil Shah, VP for research and partner at Counterpoint Research.

Faisal Kawoosa, founder and lead analyst at Techarc, noted that China’s longstanding belief in building its entire tech stack and ecosystem is now bearing fruit. “Huawei developing Ascend 910D is a step in this direction. It’s a significant move and will open up a dual AI front on the global arena,” he said. “Chinese tech companies have historically led mass adoption of technologies, and this could now accelerate global AI deployments as well.”

While countries like the US and India remain wary of Huawei, Kawoosa added that there is still “significant opportunity for Huawei to popularize its tech in markets where it remains welcome,” delivering fully localized solutions without reliance on foreign suppliers.

A powerful move — but how close to Nvidia?

Despite the fanfare, Huawei’s journey to rival Nvidia on technical grounds remains uphill. According to Shah, the Ascend 910D is a meaningful advancement but still lags behind Nvidia in critical areas.

“From an overall system-level design — compute, memory integration, networking scalability, and crucially, software orchestration — Nvidia remains three generations ahead,” Shah explained.

Huawei’s Ascend 910D will reportedly ship with two generations old high-bandwidth memory (HBM) compared to Nvidia’s latest offerings. Moreover, without an advanced CUDA-like ecosystem to optimize AI workloads across GPUs and networks, Huawei’s chips will face inherent disadvantages in scalability and efficiency.