Huawei’s CloudMatrix 384 Supernode Could Disrupt Global AI Chip Landscape
Chinese tech giant Huawei has unveiled a high-performance computing system, the CloudMatrix 384 Supernode, that could reshape the global AI hardware race as reported by AINews. Early reports from Chinese media claim the system outperforms Nvidia’s flagship NVL72, positioning Huawei as a serious contender in the AI accelerator market.
The CloudMatrix 384 Supernode reportedly achieves 300 petaflops of processing power, significantly surpassing Nvidia’s NVL72 system, which delivers 180 petaflops. If verified, this performance milestone makes the CloudMatrix 384 the most powerful AI infrastructure built with domestically developed Chinese components.
Developed to tackle computing bottlenecks in large-scale AI model training and inference, the system supports 1,920 tokens per second throughput and achieves accuracy levels comparable to Nvidia’s H100 chips.
Building Under Sanctions
This advancement is especially notable given the US export restrictions imposed on Huawei. Barred from accessing advanced American semiconductor tools and design software, Huawei has pivoted to domestic innovation.
The CloudMatrix 384’s success reflects the company’s ability to innovate despite constraints, particularly in designing a proprietary interconnect technology that rivals Nvidia’s NVLink.
AINews reports that the Huawei is working with Chinese AI infrastructure startup SiliconFlow to deploy the CloudMatrix 384 in support of DeepSeek-R1, an advanced reasoning model developed by Hangzhou-based DeepSeek. This signals the system’s readiness for real-world, high-performance AI workloads.
Supernodes like the 384 model are designed with enhanced compute, memory, and bandwidth, making them ideal for training trillion-parameter foundation models and supporting large-scale inference.
Also read: OpenAI Introduces Budget-Friendly 'Flex Processing' API for Lower-Priority AI Tasks
Part of a Larger AI Infrastructure Push
According to AINews, Huawei’s breakthrough forms part of a broader effort by Chinese firms to reduce reliance on foreign technology in the AI space. In February 2025, Alibaba committed over $52 billion to building domestic computing and AI infrastructure, the largest such investment by a Chinese company to date.
These efforts underscore China’s strategy to achieve technological self-sufficiency, especially in AI and semiconductor manufacturing, amid ongoing trade and technology tensions with the United States.
Implications for Global AI Development
If verified, Huawei’s CloudMatrix 384 Supernode could offer a cost-effective alternative to Nvidia AI hardware and marks a major step in China’s tech self-sufficiency, showing that US sanctions may have fueled, rather than stalled, domestic innovation.
