Microsoft to Launch Athena, Its First AI Chip

Microsoft, the tech giant behind Windows, Office, and Azure, is reportedly planning to unveil its first artificial intelligence (AI) chip next month at its Ignite conference. The chip, codenamed Athena, is designed to power Microsoft’s data center servers that are used to train and run large language models (LLMs) and other AI applications. By developing its own AI chip, Microsoft aims to reduce its dependency on Nvidia, the leading supplier of graphics processing units (GPUs) for AI workloads.

Nvidia GPUs are widely used by companies and researchers who need high-performance computing power to train and deploy LLMs, such as ChatGPT, the popular conversational AI system developed by OpenAI. Microsoft has invested billions of dollars in OpenAI and hosts its LLMs on Azure, its cloud computing platform. However, Nvidia GPUs are expensive and in short supply, which limits the scalability and affordability of AI services.

According to a report, Microsoft has been secretly working on Athena since 2019, using Taiwan Semiconductor’s (TSMC) 5-nanometer process. TSMC is the world’s largest contract chipmaker and also produces chips for Apple, AMD, and Qualcomm. The report claims that Athena could reduce the cost per chip by a third when compared with Nvidia’s offerings. For example, Nvidia’s most powerful AI chip, the H100, costs $40,000.

The report also suggests that Athena is currently being tested by a small group of employees at Microsoft and OpenAI, and could be made available to Azure customers as early as next year. However, Microsoft does not intend to replace Nvidia’s products entirely, as the two companies have recently partnered on several AI initiatives, such as Nvidia’s DGX Cloud and Omniverse cloud.

Microsoft is not the only tech company that is developing its own AI chips. Google, Amazon, and Meta (formerly Facebook) have also invested in in-house silicon to power their AI services and products. Google’s Tensor Processing Units (TPUs) are used for its search engine, YouTube, Gmail, and Google Photos. Amazon’s Inferentia chips are used for its Alexa voice assistant and AWS cloud. Meta’s GROK chips are used for its social media platforms and virtual reality devices.

By creating its own AI chip, Microsoft hopes to gain a competitive edge in the fast-growing AI market and offer more value to its customers and partners. Athena could also enable Microsoft to enhance its own AI features in its productivity applications, such as Office 365 and GitHub. With Athena, Microsoft could potentially create a mini-sun of its own in the AI space.