Huawei unveils latest tech to boost AI power in push to break China’s Nvidia reliance
Huawei unveiled its Atlas 960 SuperPoD computing cluster and an upgraded UnifiedBus interconnect technology. These components are central to Huawei's strategy for advanced AI systems. The new SuperPoD integrates near-packaged optics for higher speeds and lower energy use. It can connect up to 4,000 processors. The Atlas 960 SuperPoD, powered by Ascend 960 chips, shows a 2.3-fold improvement in training and a 2.5-fold improvement in inference for 10-trillion-parameter models compared to its predecessor.
Huawei's Atlas 960 SuperPoD is an incremental, not revolutionary, step. The 2.3-fold training improvement and 2.5-fold inference improvement for 10-trillion-parameter models are solid, but they do not close the gap with global leaders. Near-packaged optics and the ability to connect 4,000 processors extend optical connections deeper into the system. This improves internal efficiency, but it does not address the fundamental chip manufacturing constraints Huawei faces.
This development reflects China's continued push for self-sufficiency in AI infrastructure. Huawei's efforts, while constrained, strengthen domestic alternatives to Nvidia. This benefits Chinese cloud providers and AI developers by offering a viable, if not cutting-edge, local option. The investment reinforces China's technological independence, even if it means sacrificing some performance relative to international benchmarks.
The thing to watch is the adoption rate among major Chinese AI players. If Baidu or Alibaba commit significant workloads to the Atlas 960 SuperPoD, it validates Huawei's approach. Without substantial domestic uptake, Huawei's advancements remain primarily a national ambition rather than a market-shifting product.
