Huawei Unveils AI Supernode to Rival Nvidia’s NVL72 in Computing Power Race
- tech360.tv
- 3 days ago
- 2 min read
Huawei Technologies has launched a new artificial intelligence infrastructure architecture that reportedly rivals Nvidia’s top-tier offerings, marking a significant step in China’s push for tech self-sufficiency.
The CloudMatrix 384 Supernode, introduced last week, is said to match Nvidia’s NVL72 system in addressing computing bottlenecks in AI data centres, according to STAR Market Daily.
Huawei’s supernode, currently deployed in its data centres in Wuhu, Anhui province, reportedly delivers 300 petaflops of computing power, surpassing the 180 petaflops offered by Nvidia’s NVL72.
Nvidia’s NVL72, launched in March 2023, features a 72-GPU NVLink domain that functions as a single GPU, enabling real-time inference for trillion-parameter large language models at speeds 30 times faster than previous generations.
Huawei’s CloudMatrix 384 Supernode is part of its broader CloudMatrix infrastructure, launched in September to meet growing demand for computing capacity amid the global AI boom.
Supernodes are AI infrastructure systems equipped with enhanced resources, including central and neural processing units, network bandwidth, storage and memory, allowing them to boost cluster performance and accelerate foundational model training.
Huawei is reportedly working with Chinese AI infrastructure start-up SiliconFlow to support DeepSeek-R1, a reasoning model developed by the Hangzhou-based firm and launched in January.
The collaboration suggests Huawei’s system, powered by Chinese-made chips, has achieved performance on par with Nvidia’s H100 chips, reaching a throughput of 1,920 tokens per second with high accuracy.
The development underscores Huawei’s progress in overcoming US sanctions and advancing domestic computing capabilities amid ongoing US-China tech tensions.
Other Chinese firms are also ramping up AI infrastructure investments. In Feb., Alibaba Group Holding announced a capital expenditure of 380 billion yuan (USD 52.4 billion) over the next three years for computing and AI infrastructure.
This marks the largest-ever investment by a private Chinese company in a computing project.
Huawei launched the CloudMatrix 384 Supernode to rival Nvidia’s NVL72
The system delivers 300 petaflops, surpassing Nvidia’s 180 petaflops
Huawei is collaborating with SiliconFlow to support DeepSeek-R1
Source: SCMP