Huawei Technologies Co has unveiled its Atlas 950 SuperPoD – the industry's largest-scale AI supernode – at the 2026 World Artificial Intelligence Conference in Shanghai.
Huawei unveiled its advanced Atlas 950 SuperPoD at the ongoing WAIC 2026 in Shanghai. (Photo via China Daily)
According to the company, the technology is a major leap forward in computing infrastructure for the Agentic AI era.
Huawei said that, by integrating unprecedented scale with breakthrough interconnect technology, the Atlas 950 SuperPoD is set to redefine the performance ceiling for trillion‑parameter model training, large-scale data center deployments, and high-concurrency inference workloads.
The Atlas 950 SuperPoD, housing 1,024 Ascend chips, which had previously been discussed only in technical briefings, was displayed on the floor of the Shanghai conference as Huawei said it was ready to deliver commercial-grade supercomputing power at an industrial scale.，
Huawei also released data showing impressive commercial traction for its Ascend ecosystem. The previous-generation Ascend 384 SuperPoD has been deployed in more than 750 commercial projects in sectors including internet services, telecommunications, finance, education, healthcare, transportation and manufacturing.
To date, Huawei has partnered with over 3,000 industry collaborators to co‑develop more than 7,000 tailored solutions, serving over 2,000 core government and enterprise clients.
Beyond hardware, Huawei is placing strong emphasis on its open‑source software ecosystem, which is critical to lowering the barrier for AI development. The company had fully open‑sourced its CANN heterogeneous computing architecture and MindSeries foundational software suite by the end of last year. To date, the CANN open‑source community has spawned 67 projects, amassing over 12.44 million lines of code and attracting more than 3,500 monthly active developers.
At this year's WAIC, Huawei showcased more than 20 benchmark case studies spanning over 60 real‑world business scenarios, as developers make the case that the combination of hardware and software excellence is driving AI from technological breakthroughs to large‑scale, profitable commercial adoption across every industry vertical.