Microsoft is Making its Own AI Chips, Called Athena

Microsoft is developing its own AI chips to power its large language models.

The Information reports that Microsoft has been working on the chips, known as Athena, since 2019.

The hardware is currently being used by some Microsoft employees as well as staff at OpenAI, the research lab behind ChatGPT in which Microsoft struck a close working relationship after investing in the startup.

Some 300 staff are reportedly working on the chips, with sources saying Athena could see wider use by both Microsoft and OpenAI sometime next year. The company is unsure as to whether to make Athena accessible to Azure cloud computing customers, however.

The Information reports that Microsoft’s interest in AI post-ChatGPT launch has accelerated Athena’s development.

OpenAI has used Microsoft’s supercomputing systems to train language models that power its ChatGPT application. However, those computers run on Nvidia chips.

Microsoft is reportedly planning on using contract foundry TSMC to manufacture the Athena chips. They will be based on a 5-nanometer process, which is one generation behind the 3nm process introduced late last year.

AI Business has contacted Microsoft for comment.

Microsoft joins other big names in tech that are working on their own AI chips. Just a couple of weeks prior, Google teased its TPU v4 chips that it uses to train around 90% of the company's AI models.

Meta has been working on its own custom-built machine learning chips but has since picked Qualcomm to power its VR headsets in a deal unveiled last September.