Microsoft has been working on a secret project codenamed Athena since 2019.
The project entails the creation of an inside artificial intelligence (AI) chip, which is being tested by a select group of Microsoft and OpenAI personnel. The Athena chip is intended for training software such as large-language models (LLMs), which are critical to the current progress in generative AI.
The chip, which is constructed on a 5nm production node, is projected to outperform what Microsoft presently buys from other suppliers, possibly saving the corporation time and money on its pricey AI projects. Microsoft also hopes to overcome GPU constraints from Nvidia, the industry leader in AI training processors, which have stifled LLM growth. Microsoft is also apparently evaluating whether to expand Athena’s use within Microsoft and OpenAI next year, as well as whether to make it more freely available to Azure customers.
The sources for this piece include an article in DataCenterDynamics.