Microsoft (MSFT.O) announced on Thursday its intention to introduce AMD (AMD.O) artificial intelligence chips to its cloud computing customers, providing an alternative to Nvidia (NVDA.O) components. Further details will be disclosed at its upcoming Build developer conference.
At the conference, Microsoft will also unveil a preview of its new Cobalt 100 custom processors.
These clusters of Advanced Micro Devices’ flagship MI300X AI chips will be available through Microsoft’s Azure cloud computing service, offering customers an alternative to Nvidia’s H100 family of GPUs, which currently dominate the data center chip market for AI.
Typically, companies need to cluster multiple GPUs to build AI models or run applications due to data and computation limitations.
AMD, projecting $4 billion in AI chip revenue this year, asserts that its chips are capable of training and running large AI models.
In addition toNvidia’s AI chips, Microsoft’s cloud computing unit offers access to its own in-house AI chips known as Maia.
The Cobalt 100 processors, set to be previewed next week, promise a 40% performance improvement over other processors based on Arm Holdings’ technology, according to Microsoft. Snowflake (SNOW.N) and other entities have already begun utilizing them.
Announced in November, the Cobalt chips are undergoing testing to support Teams, Microsoft’s business messaging tool, and are positioned to rival Amazon.com’s in-house Graviton CPUs.
