Microsoft to Launch Its First AI Chip Next Month, Reducing Reliance on Nvidia
-
On October 7, foreign media cited sources to reveal that Microsoft plans to unveil its first AI-dedicated chip at its annual developer conference next month. This move, the result of years of effort, could help Microsoft reduce its dependence on Nvidia's AI chips, which have been in short supply amid surging demand.
Microsoft's chip, similar to Nvidia's GPUs, is designed for data center servers that train and run large language models—the software behind conversational AI features like OpenAI's ChatGPT. Currently, Microsoft's data center servers rely on Nvidia's GPUs to power AI functionalities for cloud customers, including OpenAI and Intuit, as well as Microsoft's productivity applications.
The chip, codenamed "Athena," is expected to be unveiled at Microsoft's Ignite conference in Seattle on November 14. Athena is anticipated to compete with Nvidia's flagship H100 GPU, accelerating AI workloads in data centers. This custom chip has been secretly tested by Microsoft and its partner OpenAI.
Microsoft began developing Athena around 2019, aiming to cut costs and gain leverage in negotiations with Nvidia. Azure currently depends on Nvidia's GPUs for AI capabilities used by Microsoft, OpenAI, and cloud customers. With Athena, Microsoft could follow competitors like AWS and Google in offering self-developed AI chips to cloud users.
Details about Athena's performance remain unclear, but Microsoft hopes it will rival Nvidia's H100. Despite many companies promoting superior hardware and cost efficiency, Nvidia's GPUs remain the top choice for AI developers due to its CUDA platform. Attracting users to new hardware and software will be key for Microsoft.
Developing its own AI chip could also reduce Microsoft's reliance on Nvidia amid GPU shortages. Reports indicate Microsoft ordered hundreds of thousands of Nvidia chips to support OpenAI's needs after their close collaboration began. Using its own chips could yield significant cost savings.
OpenAI may also be looking to reduce dependence on Microsoft and Nvidia. Recent reports suggest the AI research lab is considering developing its own AI chips, with job postings indicating plans to evaluate and co-design AI hardware.
While Microsoft and other cloud providers have no immediate plans to stop buying Nvidia GPUs, long-term economic benefits may come from persuading cloud customers to adopt in-house chips over Nvidia's GPU servers. Microsoft is also collaborating with AMD on its upcoming AI chip, the MI300X. As AI workloads surge, this diversified approach offers multiple options, mirroring competitors' strategies to avoid vendor lock-in.
Amazon and Google have strategically integrated their AI chips into cloud promotions. Amazon funded OpenAI rival Anthropic on the condition it uses Amazon's AI chips, Trainium and Inferentia. Meanwhile, Google Cloud announced clients like Midjourney and Character AI use its Tensor Processing Units.
As AI chips become critical to data centers, betting on this field could yield high returns. With this development, Microsoft joins rivals in competing for market share in the fast-growing AI chip sector. Athena could offer cloud customers more choices while charting a more independent path for next-gen AI infrastructure.