Huawei Says AI Chip Demand Outstrips Supply as Ascend Takes on Nvidia

Huawei Technologies has said it is unable to produce enough artificial-intelligence computing equipment to satisfy demand within China, and is holding back on a full overseas push, as the company intensifies its challenge to Nvidia in the AI chip market.

Speaking at the company's Huawei Connect conference in Shanghai, rotating chairman Eric Xu said testing of the Ascend 950DT AI chip had yielded encouraging results, and that Huawei was in extensive talks with Chinese AI model developers. He expects many of them to begin training models on systems built around the chip next year.

"Since we don't have enough capacity to even satisfy the demand in China, we don't have a plan to expand into the international market in a fully-fledged way," Xu told reporters. Huawei does supply some countries where demand is particularly strong, he added, but volumes remain limited.

On the competitive landscape, Xu said reliable figures on Nvidia's share of China's AI chip market were hard to come by, but asserted that Ascend's market share should be larger than Nvidia's. Huawei did not provide data to back the estimate.

The remarks underline Huawei's expanding role in China's effort to build a domestic AI computing industry, as US export controls restrict access to Nvidia's most advanced chips and to semiconductor manufacturing equipment. Huawei has itself faced US trade restrictions since May 2019.

Xu tied the company's chip efforts directly to China's push to reduce reliance on foreign technology. "We cannot accept a destiny where we cannot control our fate being determined by others in terms of willingness to sell chips to China or not," he said, adding that pursuing full self-sufficiency in chips is the way forward for the Chinese government, Chinese industry and Huawei alike.

AI infrastructure has become central to Huawei's strategy. Guo Ping, chairman of the company's supervisory board, told employees in remarks released this week that Huawei sees AI as its "biggest opportunity" and wants its computing and connectivity infrastructure to play a role comparable to Nvidia's.

Next-generation chips pulled forward

Huawei also detailed its next generation of AI processors and computing systems. Rotating chairman David Wang said the Ascend 960DT chip is now planned for the first quarter of 2027, three quarters earlier than previously scheduled, while the 960PR is due in the third quarter, a quarter ahead of plan. Huawei operates a rotating-chair system in which senior executives lead the company in six-month turns.

The company intends to release a new generation of Ascend chips every year, with the 970 and 980 to follow in 2028 and 2029.

Compensating with scale

Huawei is increasingly seeking to offset limits on individual Chinese chips by linking large numbers of processors so they function as a single, larger computing system. Clusters of roughly 100,000 chips have become standard for training some of the largest AI models, the company said, while communication between machines can consume more than 40% of training time in conventional server systems. Faster links between processors can ease that bottleneck and make better use of available computing power.

Huawei unveiled a new computing architecture called Peerium, designed eventually to allow as many as 1 million processors to operate together as one system. Its UnifiedBus technology connects processors, memory, storage and networking equipment within those systems. The new Ascend 960 supernode is to connect as many as 4,096 AI processors, with multiple supernodes linkable into clusters of hundreds of thousands of processors and the largest planned systems supporting up to 1 million.

According to company materials, Huawei has already deployed more than 1,000 systems using its earlier Ascend 910C processors, while its Ascend 950 systems have entered commercial use. More than 5,200 developers are active each month on software for its Ascend chips, and more than 40 AI models have been trained directly on its computing platform.

Nvidia retains a significant advantage in software through its widely used CUDA platform, which lets developers build and run AI applications on Nvidia chips. The US chipmaker did not immediately respond to a request for comment.