By Che Pan, Casey Hall and Eduardo Baptista
SHANGHAI/BEIJING Sept 17 (Reuters) - Huawei Technologies cannot produce enough artificial-intelligence computing equipment to meet demand in China and is limiting overseas sales, rotating chairman Eric Xu said on Thursday, as the Chinese tech giant steps up its challenge to Nvidia.
Xu said testing of Huawei's Ascend 950DT AI chip had produced good results and the company was in extensive discussions with Chinese AI model developers. He expected many of them
to begin training models on systems using 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 at the company's Huawei Connect conference in Shanghai.
Huawei supplies some countries where demand is particularly strong, Xu said, but volumes remain limited.
Xu also said reliable data on Nvidia's share of China's AI chip market was difficult to obtain but added: "I think Ascend market share should be bigger than Nvidia's."
Huawei did not provide data to support the estimate.
The comments underscore Huawei's growing role at the centre of China's drive to build a domestic AI computing industry, as US export controls restrict access to Nvidia's most advanced chips and semiconductor manufacturing equipment. Huawei has faced separate US trade restrictions since May 2019.
Xu explicitly linked Huawei's chip push to China's effort to reduce its 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.
"No matter if it's for the Chinese government, industry in China, or for Huawei, it is certainly the way forward to try to push for full self-sufficiency for chips."
Huawei has made AI infrastructure central to its strategy. Guo Ping, chairman of Huawei's supervisory board, told employees in remarks released this week that the company viewed AI as its "biggest opportunity" and wanted its computing and connectivity infrastructure to play a role comparable to Nvidia's.
NEXT-GENERATION AI CHIP TO BE RELEASED EARLIER THAN PLANNED
Huawei also detailed its next generation of AI processors and computing systems on Thursday.
The company plans to have its Ascend 960DT chip ready in the first quarter of 2027, three quarters earlier than previously planned, and the 960PR ready in the third quarter, one quarter ahead of schedule, rotating chairman David Wang said.
Huawei has a rotating-chair system in which senior executives take turns leading the company for six-month periods.
Huawei intends to release a new generation of Ascend chips every year, with the 970 and 980 to follow in 2028 and 2029.
WEAKER INDIVIDUAL AI CHIPS, POWERFUL COMPUTING SYSTEMS
The company is increasingly trying to compensate for limits on individual Chinese chips by connecting large numbers of processors so they can work together as one larger computing system.
Huawei said clusters of about 100,000 chips have become standard for training some of the largest AI models, while communication between machines can consume more than 40% of training time in conventional server systems. Faster links between processors can reduce that bottleneck and make better use of available computing power, the company said.
Huawei unveiled a new computing architecture, called Peerium, designed eventually to allow as many as 1 million processors to operate together as a single system.
Its UnifiedBus technology connects processors, memory, storage and networking equipment inside those systems.
Huawei said its new Ascend 960 supernode would connect as many as 4,096 AI processors. Multiple supernodes could then be linked into clusters containing hundreds of thousands of processors, with the largest planned systems supporting as many as 1 million.
Huawei has already deployed more than 1,000 systems using its earlier Ascend 910C processors, while its Ascend 950 systems have entered commercial use, according to company materials.
The company said more than 5,200 developers were active each month on software for its Ascend chips and that more than 40 AI models had been trained directly on its computing platform.
Nvidia retains a major advantage in software through its widely used CUDA platform that allows developers to build and run AI applications on Nvidia chips. The US chipmaker did not immediately respond to a request for comment.
(Reporting by Casey Hall and Pan Che; Editing by Muralikumar Anantharaman and Jacqueline Wong)
