- Huawei unveils its SuperPoDs, SuperClusters and UnifiedBus protocol as it lays out long-term chip plans at Huawei Connect 2025.
- The company aims to reduce reliance on foreign AI hardware suppliers like Nvidia by expanding its Ascend and Kunpeng chip offerings.
What happened: Huawei reveals chip expansion roadmap
Huawei has introduced powerful “SuperPoDs” and “SuperClusters” at its Huawei Connect 2025 event, signalling a major push into the AI semiconductor market. Eric Xu, Deputy Chairman and rotating Chairman, said that computing power “is – and will continue to be – key to AI. This is especially true in China.”
The new portfolio includes the Atlas 950 SuperPoD with 8,192 Ascend NPUs and the Atlas 960 SuperPoD with 15,488 Ascend NPUs. Huawei also showed off the Atlas 950 SuperCluster (500,000+ NPUs) and Atlas 960 SuperCluster (over one million Ascend NPUs).
Alongside these, Huawei launches the UnifiedBus 2.0 interconnect protocol, designed to improve chip-to-chip communication over long distances with low latency. The company also introduced its TaiShan 950 SuperPoD, a general-purpose model, and promised integration with its GaussDB distributed database system. These announcements accompany Taipei and Kunpeng server-chip timelines now made public.
Also read:Knetco and Huawei drive Kuwait 5G advanced rollout
Also read:Huawei challenges Nvidia in AI chip race
Why it’s important
Huawei’s move is a significant escalation in the global AI infrastructure competition. For years, companies like Nvidia, AMD and Intel have dominated high-end AI chips. By deploying clusters with hundreds of thousands of Ascend NPUs and promoting its own protocols like UnifiedBus, Huawei is signalling that it intends to be a serious alternative. This development may encourage other companies to invest more aggressively in domestic chip research and development.
For AI-driven industries in China and beyond, these innovations promise more locally sourced hardware, which can reduce supply chain risks tied to geopolitical pressures and sanctions. Huawei’s strategy directly responds to those challenges. Moreover, by releasing product roadmaps for Ascend and Kunpeng chips publicly, the company increases transparency—a welcome move in an industry often criticised for secrecy.
From a positive stance, Huawei’s announcements suggest it is not merely aiming to catch up, but to lead in computing power at scale. The launch of large-scale SuperClusters, plus the push to open UnifiedBus 2.0 for ecosystem partners, shows intent to foster an AI hardware ecosystem rather than maintaining proprietary lock-in. This could benefit developers, researchers and customers who seek more competition and innovation in AI hardware. For the global market, Huawei’s growth represents both challenge and opportunity: others will need to respond with innovation, price competitiveness, and reliability to avoid being left behind.
