The new AI arms race among technology titans such as Google and Microsoft appears to be compelling businesses to go to new extremes. The latest, per a report from The Information (paywall), as reported by The Verge, is as follows: In an effort to reduce its dependence on Nvidia’s server processors, Microsoft is currently developing its own AI circuits that can be used to educate its products.
This Microsoft AI processor reportedly codenamed “Athena,” has been in development beginning as early as 2019 and is currently in use by specific Microsoft and OpenAI employees, with a wider internal deployment possible in 2020.
It is unclear, however, whether the processors will remain for internal use only or will also be distributed to consumers who use Azure cloud products. Several future variants of the processor are also in the works, although The Information and The Verge note that the semiconductors will not immediately supplant Nvidia’s H100 GPUs.
A report from a few years ago in 2020 indicated that Microsoft was contemplating its own custom ARM-based server processors, but it is unclear whether this has any relation to Athena. Surface devices already feature Microsoft’s SQ series of Arm-based processors, developed in collaboration with Qualcomm.
Microsoft also collaborated with AMD on custom Ryzen processors for the Surface Laptop line of devices. As Microsoft continues to incorporate Artificial Intelligence into its primary products, it is evident that it has experience in this chipset area. Those who monitor the industry should not be too surprised by this move. Google employs its own AI supercomputers with over 4,000 TPUs and custom components capable of running AI models. Recently, the complete documentation was published online, and it has been in use since 2020.