Huawei unveiled the Atlas 960E SuperPoD at HUAWEI CONNECT 2026 in Shanghai, a computing cluster built to compete directly with Nvidia in AI infrastructure. The system packs 4,096 Ascend NPUs into a single memory pool and claims performance and efficiency gains over its predecessor.
Huawei took direct aim at Nvidia's AI infrastructure dominance on September 17, unveiling the Atlas 960E SuperPoD at HUAWEI CONNECT 2026 in Shanghai. The cluster fits up to 4,096 of Huawei's Ascend Neural Processing Units into a single unified memory-addressing scheme, letting every chip draw from the same memory pool instead of shuttling data across slower interconnects.
Performance claims target trillion-parameter models
Huawei says the system delivers 8 EFLOPS for FP8 computations and 16 EFLOPS for FP4, with up to one petabyte of High Bandwidth Memory across the full configuration. Compared with the earlier Atlas 950 SuperPoD, the company claims the 960E delivers between 2.3 and four times better performance on training and inference tasks for models approaching ten trillion parameters.
Efficiency gains accompany the performance jump. The Atlas 960E replaces roughly 48,000 conventional 800G optical modules with approximately 5,500 Hi-ONE Near-Packaged Optics units. That swap cuts power consumption by more than 550 kilowatts per pod. System availability reaches 99.8%, and mean time between failures doubles compared with predecessor models.
Huawei signals larger ambitions ahead
Huawei is also targeting a single AI computing framework scalable to 256,000 nodes, with SuperCluster configurations eventually exceeding one million NPUs. Larger versions of the Atlas 960 platform are expected in Q4 2027, and the Ascend 960 development program is reportedly running ahead of its original schedule.
As AI training workloads grow and data center operators face energy constraints, cutting 550 kilowatts of consumption per pod is not a minor footnote. Efficiency improvements like this translate directly into total cost of ownership advantages that can shift procurement decisions even when raw performance stays close between rivals.
Source: Crypto Briefing
Trading involves risk.
Risk Disclaimer
