Huawei Technologies has launched a new artificial intelligence (AI) infrastructure architecture intended to rival US chip maker Nvidia’s offerings. The new architecture, named the CloudMatrix 384 Supernode, was introduced last week and is reported to significantly enhance computing power for AI data centers.

According to sources cited by STAR Market Daily, the CloudMatrix 384 Supernode is described as a ‘nuclear-level product’, matching the performance of Nvidia’s NVL72 system in mitigating computing bottlenecks.

The NVL72 system, which Nvidia launched in March 2022, features a configuration of 72 graphics processing units (GPUs) linked through Nvidia’s NVLink technology. This setup enables real-time inference for large language models (LLMs) at speeds up to 30 times faster than earlier systems.

Nvidia’s NVLink facilitates high-speed communication between multiple GPUs, allowing for efficient data sharing and enhanced computational capabilities.

Reportedly, Huawei’s new supernode has achieved an impressive 300 petaflops of computing power, surpassing Nvidia’s NVL72, which offers 180 petaflops, according to Huawei’s data. The advancement signals Huawei’s strategic positioning in AI technology and its efforts to compete with major players in the market.

As of now, Huawei has not provided any official comments on the launch as they did not respond to media inquiries.