Anthropic is negotiating to use Microsoft's in-house AI chips, a move that could validate the software giant's multibillion-dollar silicon ambitions and reshape the AI supply chain.
Anthropic is negotiating to use Microsoft's in-house AI chips, a move that could validate the software giant's multibillion-dollar silicon ambitions and reshape the AI supply chain.

Anthropic is in talks to rent servers powered by Microsoft’s custom Maia AI chips, a landmark deal that threatens to divert millions in revenue from Nvidia and validates the costly in-house silicon strategies of major cloud providers. The potential partnership, first reported by The Information, would mark a significant win for Microsoft in its race against Alphabet and Amazon to develop viable alternatives to the expensive Nvidia chips that dominate the AI industry.
Microsoft shares rose about 1.5 percent in pre-market trading after the report. A deal would represent a major external customer win for Microsoft's long-running and costly custom chip program, validating its strategy to lower the hardware costs of its Azure cloud services and protect profit margins.
The negotiations are still in early stages and may not result in an agreement. Microsoft's Maia 200 chips, now being deployed in Azure data centers, are designed to run existing models like Anthropic's Claude more efficiently, targeting inference workloads rather than the large-scale training dominated by Nvidia's GPUs.
For investors, the deal is a crucial test for Microsoft's estimated $50 billion in annual capital expenditure on data centers. A major customer win would justify the investment and pressure Nvidia's 80 percent-plus market share, while a failure would raise questions about the viability of non-Nvidia hardware for top-tier AI models.
The move is a natural extension of Anthropic's deliberate strategy to diversify its hardware suppliers. Unlike competitors who are heavily reliant on Nvidia, Anthropic has built a multi-cloud, multi-chip foundation, using hardware from Google, Amazon Web Services, and Nvidia. The company has committed to spending a combined $33 billion on cloud services from the three largest US providers.
With revenue projected to hit $11 billion this quarter, a more than 100 percent increase from the previous, Anthropic has the financial firepower to influence its partners. By working with Microsoft on Maia, Anthropic could not only secure more computing power for its popular Claude model but also help shape the design of future chip generations to better suit its needs.
For Microsoft, landing Anthropic as an external customer for its Maia chips is a critical proof point. The company, along with Google with its TPUs and Amazon with its Trainium and Inferentia chips, is pouring billions into developing alternatives to Nvidia's hardware to control soaring costs and improve cloud service margins.
The Maia chip program reportedly faced delays last year, putting Microsoft behind its key rivals. However, the company announced in January that its Maia 200 chips were being used internally to power its own Copilot AI tools, which are supported by models from both OpenAI and Anthropic. Securing a flagship external customer is the next logical step to achieving scale and recouping development costs.
The deepening relationship comes as Microsoft's primary AI partner, OpenAI, reportedly explores diversifying its own chip suppliers and developing its own silicon. Microsoft has already invested up to $5 billion in Anthropic, which in turn has committed to a $30 billion cloud spend on Azure, though the current scale of its usage is still smaller than OpenAI's.
This article is for informational purposes only and does not constitute investment advice.