Microsoft Builds Its Own Artificial Intelligence Chip on TSMC’s 5nm Process

Microsoft is internally working by itself synthetic intelligence chip, dubbed “Athena,” and has been engaged on the challenge since 2019, in line with a report. Info (opens in new tab). The primary model of the chip is deliberate to make use of Taiwan Semiconductor’s (TSMC) 5-nanometer course of, however there’ll possible be a number of chip generations as a part of the challenge.

Info He argues that Microsoft’s aim is no less than partly to economize when shopping for chips from suppliers like Nvidia.

Microsoft has poured billions into OpenAI, which makes ChatGPT, to advance in a fast-growing race to make merchandise utilizing synthetic intelligence and broad language fashions (LLMs). Athena chips are reportedly designed to coach LLMs and related software program, together with inferring from knowledge the fashions acquired throughout coaching.

chips accordingly Infosources are presently utilized by a small subset of staff at Microsoft and OpenAI. Microsoft might probably make the chips extra widespread “early subsequent yr”, nevertheless it’s unclear whether or not the chip will roll out to Azure prospects.

Many individuals who educate LLM regionally or within the cloud have defaulted to Nvidia’s choices, together with highly effective graphics playing cards which have confirmed appropriate for any firm that does not have their very own chips. Dylan Patel of analysis agency SemiAnalysis mentioned: Info “Athena, if aggressive, can lower the associated fee per chip by a 3rd in comparison with Nvidia’s choices.” If Microsoft provides AI capabilities past Bing Chat to its Workplace merchandise and GitHub, it might save a ton.

After all, Microsoft is not the primary to make its personal chips. Google, Amazon and Meta are among the many tech giants which might be already investing in in-house silicon.

That does not imply that Microsoft considers Athena to be an entire alternative for Nvidia. Two firms in March Partnered with Nvidia’s DGX Cloud (opens in new tab) help for each Azure and Microsoft 365, particularly for AI supercomputing and the Nvidia Omniverse cloud.

If Microsoft produces its Athena chips extra extensively subsequent yr, it should now not be in TSMC’s high-end nodes. However with Microsoft’s multi-billion funding in OpenAI and its early (typically reckless) foray into chatbots from Bing Search and past, it isn’t stunning to see Microsoft searching for methods to enhance it additional.