Huawei has unveiled an ambitious three-year plan aimed at strengthening its position in the field of artificial intelligence and enhancing competition with Nvidia in the AI chip market. During the presentation, the company’s chairman, Eric Xu, outlined a roadmap that includes the launch of an updated line of Ascend chips and the deployment of a large-scale SuperPod platform.
This is reported by Business • Media
Focus on Scalability and Government Support
The foundation of Huawei’s new strategy is based on leveraging mass production, advanced networking technologies, and government support. The company notes that these factors are intended to compensate for the current lag behind Nvidia in the performance of individual processors, enabling the creation of competitive solutions through chip clustering.
According to Huawei, the SuperPod platform will allow the integration of up to 15,488 Ascend chips into a single system. This is expected to provide a level of performance comparable to Nvidia’s flagship products. At the same time, the company’s developed UnifiedBus protocol accelerates data exchange between chips by several times compared to similar products from other manufacturers. Thus, Huawei aims to establish a flexible architecture for the development of artificial intelligence in China.
“The public demonstration of such plans indicates the company’s confidence in future domestic supply within China. The manufacturer had previously avoided openly showcasing technologies due to concerns about U.S. sanctions, but is now demonstrating confidence in the development of a local manufacturing base,” analysts at Bernstein note.
Challenges for Mass Production and Technological Race
The company’s new direction reflects China’s national strategy for developing its own semiconductor ecosystem. In addition to Huawei, other tech giants such as Alibaba and Baidu are also actively promoting their own AI chip projects. This aligns with Beijing’s policy of reducing dependence on foreign technologies, particularly American ones, and coincides with restrictions on Nvidia component purchases for government entities.
At the same time, experts point out several limitations for Huawei. In particular, a next-generation Ascend chip falls short in performance compared to Nvidia’s latest chips, and mass production is complicated by limited access to modern lithography technologies and restricted collaboration with leading suppliers like TSMC. In some cases, such as with the Ascend 910D chip, the company has already faced difficulties at the production stage.
Despite these challenges, Huawei’s management is confident that through the development of networking solutions, processor clustering, and improvements in inter-chip connections, the company will be able to narrow the gap with industry leaders. The corporation is focusing on creating a resilient technological base for the development of artificial intelligence in China.
It is worth noting that earlier, the Chinese authorities recommended that companies refrain from using Nvidia’s H20 chips, emphasizing the push for the localization of strategic technologies.