By Max A. Cherney
SAN FRANCISCO (Reuters) – Microsoft (NASDAQ:) said on Thursday it plans to offer its cloud computing customers a platform of AMD artificial intelligence chips that will compete with components made by Nvidia (NASDAQ:), details of which will be given at the Build developer conference next week.
It will also launch a preview of new Cobalt 100 custom processors at the conference.
Microsoft’s clusters of Advanced Micro Devices (NASDAQ:)’ flagship MI300X AI chips will be sold through the Azure cloud computing service. They will offer their customers an alternative to Nvidia’s H100 family of high-performance graphics processing units (GPUs) that dominate the AI data center chip market but can be difficult to obtain due to high demand.
To build AI models or run applications, companies typically need to string (or cluster) multiple GPUs together, because the data and calculations can’t fit on a single processor.
AMD, which expects $4 billion in revenue from AI chips this year, has said the chips are powerful enough to train and run large AI models.
In addition to Nvidia’s high-end AI chips, Microsoft’s cloud computing unit sells access to its own internal AI chips called Maia.
Separately, the Cobalt 100 processors that Microsoft plans to preview next week offer 40% better performance than other processors based on Arm Holdings (NASDAQ:)” technology, the company said. Snowflake (NYSE:) and others have started using them.
The Cobalt chips, announced in November, are being tested to power Teams, Microsoft’s messaging tool for businesses, and positioned to compete with Amazon.com’s (NASDAQ:) in-house Graviton CPUs.
