Table of Contents
Last week, Huawei unveiled its AI CloudMatrix 384 data center supercomputer, positioning it as a direct competitor to Nvidia’s upcoming GB300 NVL72. The Chinese tech giant claims the CloudMatrix 384 delivers 300 PFLOPS of computing power, 269 TB/s of network bandwidth, and a massive 1,229 TB/s of total memory bandwidth. As a result, reviewers and AI enthusiasts are already calling Huawei’s CloudMatrix 384 “better than the GB300 NVL72 on all metrics.”
However, this AI momentum from China might be short-lived, as the US government has recently imposed an indefinite ban on Nvidia’s H20 chip sales to the country. Still, Huawei’s rapid progress suggests it’s not far behind Nvidia’s H20 in terms of performance. Instead, the restrictions might end up hurting Nvidia more, with a report from ForexLive estimating the ban could cost the company a staggering US$5.5 billion this quarter.
Of course, there’s trouble for China, too
While the US ban on H20 chips might push China into even higher gear, Chinese AI giants like Tencent and Alibaba, who rely on H20 chips to train their AI models like the latest Qwen 2.5 Max, which already surpasses DeepSeek’s V3 will now have to face delays or come up with their own alternatives.
Latest tech deals.
- AMD Ryzen™ 9 9900X 12-Core - was $499 now $328
- CORSAIR 4000D PC Case - was $124.99 now $75.99
- Hisense 65" Class E6SR - was $549 now $377.99
- Apple 2026 MacBook Air - was $1,999 now $1,799
- Skytech Gaming PC Desktop (9850X3d) - was $4,299 now $3,799
- Apple Magic Mouse - was $79 now $59
- MSI Prestige 16 Flip - was $2,149 now $1,899
- Kotin 9700X Gaming PC - was $2,399 now $2,199
Prices correct as of October 9th, 2026.
Fortunately, Huawei already has a major rival to challenge Nvidia’s AI chips in the form of its upcoming Ascend 910C chips, which not only seem to take a simpler approach than Nvidia’s but have also managed to offer 60% of the H100’s performance already. So it’s only a matter of time before the Chinese company develops its own H20-level AI GPUs, potentially closing the door on Nvidia’s access to the Chinese AI market.
Some believe that, due to Huawei’s rapid progress, banning the H20 GPUs wouldn’t have much impact. Instead, restricting tools and sub-components, such as semiconductor equipment and design software, could be far more effective. This means that if the US government doesn’t slow down Huawei’s fast-paced ramp, the H20 ban could end up doing more harm to Nvidia than to Huawei itself.
Written by
