Huawei Technologies anticipates a significant transition in 2027 where much of China's AI model training will rely on its Ascend-powered SuperPoD and SuperCluster computing infrastructure, signaling a challenge to Nvidia's dominance despite ongoing hardware supply limits.
- Huawei’s Ascend chips surpass Nvidia in China’s domestic AI market.
- SuperPoD and SuperCluster clusters expected to power most AI training in 2027.
- Global AI computing supply-demand balance projected near 2029-2030.
What happened
At the Huawei Connect 2026 conference in Shanghai, Huawei's rotating chairman Eric Xu Zhijun announced a major domestic shift in AI model training to Huawei’s proprietary computing systems using Ascend processors starting from 2027. Xu noted that these AI model training systems, specifically the SuperPoD and SuperCluster clusters powered by Ascend 950DT chips, have gained significant ground in China, overtaking Nvidia’s market share despite limited detailed figures. Huawei has already deployed over 1,000 Atlas SuperPoD units across hundreds of Chinese enterprises.
The Shenzhen-based company also unveiled the Atlas 960 SuperPoD cluster equipped with Ascend 960 chips, offering double the performance of its predecessor. However, due to production capacity constraints, these advanced systems will primarily remain within China. Huawei has shipped limited units abroad for testing in select high-demand regions but has no plans for broad international rollout at this stage. Additionally, their new Peerium Computing Architecture aims to link up to 1 million processors into a cohesive computing resource and is currently tested on new hardware.
Why it matters
Huawei’s ramp-up in AI computing infrastructure reflects China's broader geopolitical strategy to reduce dependence on US and foreign technology amid tightening export controls, which limit China’s access to cutting-edge chips like Nvidia's. While Nvidia continues to provide the backbone for the most advanced models in China, Huawei's Ascend platform offers stronger domestic support and addresses concerns over geopolitical risks and supply reliability.
This shift exemplifies China’s accelerating push towards semiconductor self-sufficiency, aiming to build a more resilient AI ecosystem. Huawei’s growing footprint in AI computing infrastructure not only challenges Nvidia’s dominance in the massive Chinese market but also highlights the strategic importance of indigenous technology development in AI hardware as the country prepares for soaring AI demands.
What to watch next
Industry observers will monitor Huawei’s ability to scale production of its Ascend chip-based AI clusters to meet domestic demand fully and whether it can expand international sales beyond limited testing. The evolution and deployment of Huawei’s Peerium architecture will also be a critical factor, as enabling massive processor interconnection is key to handling increasingly complex AI workloads.
Additionally, the timeline for the global balance of AI computing supply and demand, projected around 2029, and a slightly later balance anticipated in China by 2030, will be significant for evaluating Huawei’s long-term impact. How Nvidia and other international competitors respond to Huawei's growing domestic market share amid US-China tech tensions will also shape future developments in AI infrastructure.