- Huawei is preparing to test its most advanced AI -chip to date
- Reports say Ascend 910D aims to challenge the NVIDIA H100 in AI Performance
- Analysts are skeptical due to software holes and the disadvantages of the ecosystem
Huawei has increased his ambitions to be considered the Chinese rival to Nvidia in recent weeks -first launched a new AI infrastructure architecture set, Cloudmatrix 384 Supernode, to compete with US Chip Giant’s NVL72 system.
Then, a day after the United States announced plans to further tighten restrictions on AI chip exports to China -including NVIDIA’s H20 processor -revealed Huawei Ascend 920, its next generation AI chip, which was to go into mass production in the second half of 2025.
Now a report in The Wall Street Journal Huawei claims are preparing to test its most advanced AI chip to date, the Ascend 910D, and have approached several Chinese tech companies to begin technical evaluations. Example batches are expected to be available as soon as the end of May 2025.
Sights set on Nvidia’s H100
The Ascend 910D is the latest iteration of Huaweis AI processors based on the existing 910B and 910C models. While the chip is not yet commercially available, Huawei reportedly hopes it can match – or even exceed – the performance of Nvidia’s H100.
Independent analysts are, of course, skeptical that Huawei will be able to close the benefit and the ecosystem’s gap between themselves and nvidia anytime soon.
HPCWIRE Report Neil Shah, VP by Counterpoint Research, like saying: “From a unified system level design – calculating, memory integration, network scalibility and crucial software orchestration – nvidia remains three generations ahead.”
This hole is partly due to Nvidia’s established software stacks, especially Cuda, which plays an important role in speeding up and managing AI workloads. Huawei, on the other hand, lacks a similar mature software platform, making it more difficult to optimize over GPUs or scale over complex AI infrastructures.
Despite these obstacles, Huawei continues to send its 910C chips in volume to Chinese data centers and research laboratories, and the move to test 910D suggests that the company is doubling in the effort to build a self-sufficient AI-Hardware ecosystem.
Whether 910D becomes a true rival to Nvidia’s H100, or remains a “good enough” opportunity for domestic use, it signalizes a continued shift in Global AI hardware competition.
