Anthropic is in early-stage discussions with Microsoft to utilize Microsoft’s custom-built AI chips for powering its advanced language models, according to reports from The Information. This potential move signals a strategic shift for the AI lab, which has historically relied on computing infrastructure provided by its primary investors, Amazon and Google.
The negotiations reportedly focus on Microsoft's Maia series chips. Microsoft unveiled its first homegrown AI accelerator, the Maia 100, last year to offer a cost-effective alternative to Nvidia’s dominant H100 GPUs within its Azure cloud environment. For Anthropic, integrating Maia chips could provide greater flexibility in managing massive inference costs and ensuring a steady supply of specialized hardware as model scales continue to grow.
If a deal is finalized, it would place Microsoft in a unique position as a cloud provider for both OpenAI and its chief rival, Anthropic. This development underscores the pragmatic approach top-tier AI labs are taking toward hardware diversification. By spreading workloads across AWS, Google Cloud, and potentially Azure, Anthropic aims to mitigate the risks of hardware shortages and vendor lock-in. As of now, neither Anthropic nor Microsoft has officially commented on the ongoing talks.
