Huawei Technologies Co. unveiled new technology from memory chips to AI accelerators Thursday, outlining publicly for the first time its multiyear plan to challenge Nvidia Corp.’s dominance in a growing market.
The highlight of the company’s presentation on Thursday were new SuperPod cluster designs that will allow Huawei to link as many as 15,488 of its Ascend neural processing units for artificial intelligence and operate them as a coherent system, rotating chairman Eric Xu said at the event. Those SuperPod products will be built with new generations of Ascend chips from next year.
The next-generation Ascend 950 series will be accompanied by new high-bandwidth memory designed by Huawei itself, Xu said, without elaborating on who will fabricate the semiconductors. Huawei also plans to roll out an Ascend 960 in late 2027, to be succeeded by a 970 model in late 2028.
“This is a significant milestone in China’s long march of the AI chipset industry,” said Charlie Dai, vice president at Forrester Research. “This achievement reflects breakthroughs in system design, interconnect technologies, and local fabrication capabilities. It signals a stronger push toward self-reliance and resilience in the face of export restrictions.”
Shenzhen-based Huawei is China’s most advanced chip designer and strongest contender to build alternatives to Nvidia’s industry-leading AI hardware. Washington has for years blocked the export of Nvidia’s top products to China, and Beijing recently told its biggest tech companies not to use Nvidia’s RTX Pro 6000D, a graphics card for workstations that can be repurposed for AI applications. The move marked a clear attempt by Chinese leaders to wean the country off Nvidia hardware and boost domestic alternatives.
Like its US competitor, Huawei is developing systems that combine several of its AI chips into more capable clusters. The newly announced Atlas 950 SuperPod will deliver 6.7 times more computing power than Nvidia’s upcoming NVL144 systems, Xu said. Huawei is also planning a super cluster with about 1 million graphic cards, based on the new SuperPod technology.
Earlier this year, company founder Ren Zhengfei told state newspaper People’s Daily that Huawei still lags behind the US in terms of output from a single chip, but “can still get the results we want by compensating with cluster-based computing.”
The aggressive approach helps China’s AI chip leader draw more performance from its semiconductors, which face a manufacturing constraint as the company cannot readily advance to more sophisticated fabrication due to trade restrictions on leading-edge machinery. While short of a big breakthrough in chip technology, Huawei’s solution marks the latest development by Chinese firms trying to develop homegrown alternatives and lessen the impact of US sanctions.
Chinese tech stocks have surged in past weeks, driven by a perception that the nation’s industry leaders are making steady progress in developing homegrown AI and chips. Alibaba Group Holding Ltd. and Baidu Inc. are among the companies that have secured important clients for their in-house designs, while Cambricon Technologies Corp. — seen as a proxy for the country’s AI chip sector — has surged in market value this year.
Among other measures, Chinese authorities have also discouraged companies from using Nvidia’s H20 chip that’s designed for AI workloads. While the guidance stops short of an outright ban, it has nonetheless had a chilling effect. Nvidia, despite having gotten Washington’s official green light for some H20 exports, has not executed those shipments, the company said last week. That creates a bigger gap in the Chinese market for domestic chipmakers like Huawei and Cambricon to fill.
—————-
FROM www.bloomberg.com
EduTimes Africa, a product of Education Times Africa, is a magazine publication that aims to lend its support to close the yawning gap in Africa's educational development.