Full Breakdown
Huawei Unveils Advanced AI Computing Systems Amidst US Competition
9/20/2025, 12:54:37 PM
Major Announcements at Huawei Connect Conference
On September 18, 2025, Huawei Technologies unveiled its latest advancements in artificial intelligence (AI) computing at the Huawei Connect Conference in Shanghai. The company introduced the Atlas 950 SuperPoD and Atlas 960 SuperPoD, which it claims are the world's most powerful AI computing clusters. These systems are designed to integrate thousands of Huawei's Ascend processors, achieving unprecedented levels of computing power, memory capacity, and interconnect bandwidth. The Atlas 950 SuperPoD will support 8,192 Ascend cards, while the Atlas 960 SuperPoD will accommodate up to 15,488 cards, with both systems expected to launch in 2026 and 2027, respectively.
Eric Xu, Huawei's rotating chairman, emphasized the importance of computing power for AI development, particularly in China, stating, "Computing power has been and will continue to be the key to artificial intelligence." He noted that despite US sanctions limiting access to advanced chip manufacturing, Huawei has made significant strides in developing large-scale AI systems through innovations in interconnect technology.
Competitive Landscape with Nvidia
Huawei's announcements come amid heightened competition with US chipmaker Nvidia, which has faced restrictions on exporting its most powerful chips to China. The US government has imposed bans on Nvidia's Blackwell chip and other high-performance products, citing national security concerns. In contrast, Huawei's new systems are positioned as direct competitors to Nvidia's offerings, with Xu claiming that the Atlas 950 SuperPoD will have "6.7 times more computing power and 15 times more memory capacity" than Nvidia's upcoming NVL144 system.
The competitive landscape is further complicated by China's regulatory actions, including a recent ban on local companies purchasing Nvidia's RTX Pro 6000 chips, aimed at reducing reliance on foreign semiconductors. This move aligns with China's broader strategy to bolster its domestic tech industry and enhance self-sufficiency in AI technologies.
Technological Innovations and Future Roadmap
Huawei's Ascend chip series, which includes the upcoming 950, 960, and 970 models, is central to its strategy for competing with Nvidia. The Ascend 950 series, set to launch in early 2026, will feature two types of chips: the 950PR, optimized for inference, and the 950DT, designed for model training. The company has also developed proprietary high-bandwidth memory technology, which it claims is more cost-effective than existing industry standards.
Despite the challenges posed by US sanctions, Huawei has demonstrated resilience in the face of adversity. Analysts have noted that the company's experience in connectivity technology positions it well to achieve breakthroughs in AI hardware. Paul Triolo, a partner at Albright Stonebridge Group, remarked that "if any company can accomplish the miracle of producing AI hardware that is at least close to Western leaders, it is Huawei."
Official Statements and Responses
In his remarks, Xu acknowledged the limitations imposed by US sanctions, stating, "Because of US sanctions, we cannot manufacture at TSMC, so the computing power of a single Huawei chip still lags behind Nvidia’s." However, he expressed confidence in Huawei's ability to innovate and compete effectively in the AI space.
Conflicting Reports and Gaps
While Huawei claims significant advancements in its AI capabilities, skepticism remains regarding the company's ability to match the performance of Nvidia's chips. Some experts caution that Huawei's success hinges on the manufacturing capabilities of Semiconductor Manufacturing International Corporation (SMIC), which may not yet be on par with leading foundries like TSMC.
Verbatim Quotes
- “Computing power has been and will continue to be the key to artificial intelligence—especially for China’s AI development.” — Eric Xu, Rotating Chairman, Huawei
- “If any company can accomplish the miracle of producing AI hardware that is at least close to Western leaders, it is Huawei,” — Paul Triolo, Partner, Albright Stonebridge Group
- “Because of US sanctions, we cannot manufacture at TSMC, so the computing power of a single Huawei chip still lags behind Nvidia’s,” — Eric Xu, Rotating Chairman, Huawei
Huawei's ambitious plans for AI computing systems reflect its determination to establish a leading position in the global AI race, despite the challenges posed by international trade tensions and technological restrictions.
