![]() |
Huawei's Ascend chip, central to the company’s AI computing efforts, is at the heart of the new CloudMatrix 384 system designed to rival Nvidia’s hardware. |
Huawei has intensified its push into the artificial intelligence infrastructure market by unveiling the CloudMatrix 384, a powerful AI computing system built to rival Nvidia’s dominant hardware. The Chinese tech giant also announced the upcoming Ascend 910D processor, designed to challenge Nvidia’s H100 chip in performance and efficiency. These developments come as Huawei works around restrictions imposed by ongoing U.S. sanctions, which have limited access to cutting-edge semiconductors from American suppliers.
The CloudMatrix 384 integrates 384 of Huawei’s Ascend 910C processors and delivers a staggering 300 petaflops of BF16 compute power. This performance figure surpasses that of Nvidia’s recently announced GB200 NVL72, which offers 180 petaflops. Huawei’s system also features a larger memory capacity, more than tripling what Nvidia provides in its comparable platform. The increased power comes at a cost, however, as the system consumes nearly four times more energy than Nvidia’s setup, raising questions about long-term efficiency and operating costs.
![]() |
Huawei’s CloudMatrix 384 AI computing system integrates 384 Ascend 910C chips and delivers 300 petaflops of compute power, positioning it as a domestic alternative to Nvidia’s top-tier AI platforms. |
Despite the energy concerns, CloudMatrix 384 marks a major milestone for Huawei and for China’s broader push for technological self-reliance. With U.S. export controls restricting access to Nvidia’s top-tier AI chips, Huawei’s new system provides an alternative path for domestic firms building large language models and other AI-driven services. The launch aligns with Beijing’s strategic objectives to develop indigenous high-performance computing capabilities.
In parallel, Huawei is preparing to release its next-generation Ascend 910D chip, which targets Nvidia’s high-end H100 GPU. Sample units are expected to be delivered to key partners later this month for evaluation. While the final performance benchmarks are yet to be disclosed, early indicators suggest Huawei aims to close the gap with Western rivals by leveraging vertical integration and aggressive investment in chip development.
These moves reflect a broader trend of Chinese tech firms accelerating efforts to build domestic alternatives in critical technology sectors. If Huawei’s hardware delivers competitive results, it could shift momentum in the global AI arms race, especially in markets where access to Nvidia’s chips is limited. The company’s strategy appears focused not only on technical parity but also on reshaping the supply chain of AI infrastructure in China and beyond.