AMD To Build Massive AI Cluster with 1.2M GPUs to Compete with NVIDIA

On June 26 according to TapTechNews, the number of GPUs equipped in the world's strongest supercomputer is less than 50,000. However, AMD aims to launch a data center AI cluster that supports 1.2 million GPUs in an effort to enhance its competitiveness against NVIDIA.

In an interview with TheNextPlatform, Forrest Norrod, executive vice president and general manager of AMD's Data Center Solutions Group, said that the company is seriously considering building the largest-scale AI GPU cluster.

TapTechNews note: An AI training cluster is usually built with thousands of GPUs and connects server racks through high-speed interconnections. AMD's 1.2 million being studied is undoubtedly exponentially increased in magnitude, which will inevitably bring challenges in many aspects such as latency, power consumption, and hardware failures.

The world's strongest supercomputer, Frontier, has only 37,888 GPUs, so the AI cluster sketched by AMD is 30 times the number of the current strongest supercomputer.