Towards the end of May, Huawei is expected to begin sampling its Ascend 910D AI processor, according to reports from the Wall Street Journal.
The 910D is specifically designed to compete against Nvidia’s Hopper-generation H100, first introduced in 2022.
Mass production of the Ascend 910D is reportedly scheduled for the first quarter of 2026, utilizing SMIC’s 7nm fabrication process.
While the 910D may not fully match the H100’s performance in AI training workloads, it is considered competitive in AI inference tasks — a critical operational segment for datacenters and enterprise applications.
At a rack-scale configuration, Huawei’s predecessor model, the Ascend 910C, has already demonstrated comparable performance to Nvidia’s hardware, albeit with a disadvantage in efficiency metrics.
The new 910D appears to be engineered to improve operational efficiency, targeting better energy-to-performance ratios in high-density datacenter environments.
Huawei’s 910C chip, meanwhile, is reported to be entering volume shipments this May, marking a significant step in Huawei’s strategic efforts to establish a stronger foothold in the AI hardware market.
