Huawei has introduced a new AI infrastructure called SuperPoD (Super Device/Super Performance) that aims to challenge Nvidia's dominance in the AI chip market. This innovative technology enables the connection of up to 15,488 Ascend AI chips, allowing them to function as a unified, massively powerful computing cluster.
The SuperPoD technology is designed to address the growing demand for high-performance computing in AI applications. Huawei's proprietary UnifiedBus protocol enables seamless, high-speed communication between chips, addressing bottlenecks in large model training. This allows for massive scalability, making it suitable for complex AI applications and large-scale data processing.
Huawei claims that its SuperPoD technology delivers unprecedented computing power, outperforming competitors in key metrics like total number of NPUs, aggregate computing power, memory bandwidth, and interconnect performance. The company's roadmap includes the Ascend 910C chip and Atlas 900A3 SuperPod, expected to arrive in 2025, capable of linking up to 384 processors.
The Ascend 950, planned for release in 2026, will have the ability to scale clusters to 8,192 chips. Furthermore, the Ascend 960, expected in 2027, will have the potential to tie together up to 15,488 units in a single system.
Huawei's development of SuperPoD technology is a strategic response to US sanctions and Nvidia's dominance in the AI chip market. By developing its own full-stack AI infrastructure, Huawei aims to reduce reliance on foreign technology and align with China's broader goals of technological sovereignty.
The introduction of SuperPoD technology is expected to have significant implications for the AI industry, potentially altering the competitive landscape and driving innovation in AI computing. As the demand for high-performance AI computing continues to grow, Huawei's SuperPoD technology is poised to play a major role in shaping the future of AI infrastructure .
