Huawei boss claims homegrown AI chip sales top Nvidia in China

Huawei's rotating chairman, Eric Xu, says the company's Ascend neural processing units (NPUs) have overtaken Nvidia in Chinese market share, as US export controls and Beijing's push for self-sufficiency reshape the country's AI hardware market.

A claim without hard numbers

Xu made the remark in a Q&A-style release published this week. "It's pretty hard to collect data about the market share of Nvidia in China, but based on the data we have collected, Ascend has surpassed Nvidia," he said, adding that "Chinese people have a keen sense of urgency, and will not accept a future in which others decide whether or not we can have access to certain products."

US export policy has swung back and forth. Washington imposed licensing requirements on Nvidia's and AMD's China-spec accelerators in April 2025, reversed course that summer, and by late last year had let Nvidia sell its more powerful H200-series parts to approved Chinese customers for the first time.

Assured supply beats raw performance

The earlier blockade of H20 chips alarmed officials in Beijing, who pushed datacenter operators away from foreign accelerators. On its Q2 earnings call, Nvidia said H200 shipments to the region were under 1 percent of datacenter revenues.

Xu framed assured supply as Huawei's main advantage. "Even though our chips may be less advanced, at least their supply is assured, so that you don't have to worry about chip supply day in and day out," he said. "For Huawei, the path forward is undoubtedly to push for full self-sufficiency in chips and the entire semiconductor value chain."

Huawei's latest accelerators still trail Nvidia's or AMD's on paper, but the company makes up the deficit with scale. It is currently deploying a 256,000-card Atlas 950 SuperCluster, while a newer architecture can train multi-trillion-parameter models across as many as one million NPUs.

The ecosystem, not the hardware

Xu said Huawei is unlikely to challenge Nvidia outside China for a while: "We don't have enough capacity to satisfy the demand in China." Outside China, he said, the company supplies only a limited number of customers that have no access to alternatives.

Liao Heng, chief scientist of Huawei's HiSilicon chip unit, argued the hard part was never the silicon itself. "The first barrier encountered by Ascend is actually not hardware itself. It's primarily the ecosystem barrier, because two years ago there was still a tremendous software barrier for Ascend" - a reference to the so-called CUDA moat that every rival chipmaker faces.

"There have been drastic changes over the past 18 months," Liao said, noting that AI training once hinged on PyTorch running on CUDA. DeepSeek also released tools this week to make Huawei's Ascend accelerators more accessible to other AI labs.