Huawei Technologies is poised for a major shift in China's AI model training market in 2027, with its Ascend-based SuperPoD and SuperCluster systems projected to become the primary platforms, according to the company's rotating chairman Eric Xu Zhijun.
- Huawei Ascend SuperPoD to lead AI model training in China from 2027
- Ascend chips surpass Nvidia in domestic market share amid geopolitical pressures
- Huawei unveils Atlas 960 cluster and Peerium Architecture to boost AI capacity
What happened
During the Huawei Connect 2026 conference in Shanghai, rotating chairman Eric Xu Zhijun announced that starting in 2027, a significant portion of AI model training in China will shift to Huawei’s proprietary computing infrastructure, specifically the SuperPoD and SuperCluster systems based on the Ascend 950DT chip. Huawei has reportedly overtaken Nvidia in China’s domestic AI chip market share, despite ongoing supply constraints and US export controls that have restricted the availability of foreign high-performance semiconductor products.
Huawei recently unveiled the Atlas 960 SuperPoD computing cluster powered by the Ascend 960 chip, which doubles the training and inference performance compared to its predecessor, the Atlas 950. The company also introduced its Peerium Computing Architecture designed to connect up to 1 million processors to operate as one unit. While Huawei currently has limited international shipment plans due to production capacity focused largely on addressing domestic demand, it has shipped small quantities overseas for specific testing projects.
Why it matters
Huawei's growing dominance in China's AI chip landscape highlights the broader strategic push towards technological self-reliance amid US-imposed export restrictions. The company's Ascend chips and associated AI infrastructure provide a homegrown alternative supporting domestic enterprises wary of geopolitical risks involved with relying on foreign suppliers such as Nvidia. Huawei’s ability to scale AI computing capacity could accelerate China’s AI development by reducing reliance on foreign technology and securing supply chains.
The widening gap between China’s soaring AI demand and available global computing capacity underscores the importance of Huawei’s advancements. While Nvidia still powers China’s most advanced AI models, Huawei’s solutions have been rapidly adopted across enterprises, with over 1,000 Atlas SuperPoD units deployed among 370-plus customers, signaling confidence in Huawei’s ecosystem as a viable long-term foundation for AI innovation in China.
What to watch next
Key developments to monitor include Huawei’s ability to ramp up production capacity to meet growing domestic demand and its progress integrating Peerium Architecture into its newer Atlas 960 clusters for enhanced multi-processor computing power. Observers should also watch how Huawei navigates international expansion amid ongoing US export controls and whether new licensing arrangements will enable broader overseas deployments.
Additionally, the timeline for achieving computing supply-demand equilibrium—projected globally by 2029 and slightly later in China by 2030—will be critical for assessing the overall maturation of AI infrastructure. Industry stakeholders will closely follow competitive dynamics between Huawei and Nvidia in China’s AI training market, especially as geopolitical factors continue to influence technology access and investment strategies.