Is Microsoft teaming up with AMD to develop AI chips?

[harlan4096]

A report by Bloomberg suggests that Microsoft is working with AMD to develop chips specifically for AI applications. The report bolsters an The Information report from April 2023. The magazine stated back then that Microsoft has been working on an AI chip, internally code named as Athena, with 300 employees.

Early prototypes of the chip are already being tested at Microsoft and OpenAI. Microsoft is cooperating with OpenAI on several fronts, including by using OpenAI's ChatGPT technology to power Bing Chat.

Microsoft is using chips by NVIDIA currently for its AI services. The switch to AI chips created inhouse would remove the dependency on NVIDIA and likely also result in production cost reductions and also optimizations.

Bloomberg's report brings AMD into the picture. AMD, which is battling it out with NVIDIA on several fronts, has been a Microsoft partner for a long time. AMD has created custom chips for Microsoft Xbox gaming systems. NVIDIA is dominating the AI chip market currently.

The new custom AI chip, produced by AMD and used by Microsoft in its data centers to power its AI services, could help AMD conquer a bit of the expanding market.

Bloomberg confirms the Athena code name of the AI chip and that the two companies are working on the AI chip together. Microsoft is providing AMD with support, which includes engineering resources according to the article.

Neither Microsoft nor AMD have commented on the rumor at the time of writing. First Athena-based chips are expected in 2024.

While there is no official confirmation of the claims, switching to AI chips designed inhouse would improve Microsoft's position in the market. It would reduce the dependency on NVIDIA and available stock, allow Microsoft to design custom chips for its applications and services, and may also bring down the price of the product.

Microsoft is not the only major tech company that has decided to create custom chips for some of its services and products. Apple is already using some of its inhouse chips in some of its products, and Google has been designing and deploying Tensor Processing Units already as well.
...

Continue Reading