Microsoft confirms Meta won’t buy new chips from it to scale Muse

Meta is not buying new chips from Microsoft to scale Muse, its newly launched personal AI agent, and Microsoft has confirmed as much. Muse is a full autonomous agent that books travel, manages emails, and fills out forms on your behalf, all inside secure, persistent cloud virtual machines. That kind of workload requires a lot of hardware, and the question of who supplies it is now a very live one.

Muse launched on September 8, 2026, and within days it had climbed to the top of app store charts, amassing hundreds of thousands of daily active users.

Why this chip decision is bigger than it looks

Because Muse operates as an autonomous agent inside persistent cloud VMs, it needs continuous CPU resources for orchestration and VM maintenance alongside GPU support for model inference. That is a fundamentally different infrastructure demand from anything Meta has deployed for consumer AI before.

The market noticed immediately. In the days following Muse’s debut, AMD’s market cap briefly touched $1 trillion. Intel shares rose 12% between September 22 and 23, 2026. Arm shares climbed 17% over the same period.

Meta’s answer to that demand is a multi-vendor approach. The company has partnerships with AMD, Nvidia, and others, and it is not adding Microsoft to that list, at least not for new chip procurement tied to Muse.

Muse itself runs on Meta’s Muse Spark model, operating inside those persistent cloud VMs that give it the ability to carry out multi-step tasks without losing context between sessions.

What Meta’s multi-vendor bet signals for the industry

For Microsoft, the confirmation that Meta is not coming with a new chip order is notable given that both companies are effectively competing in the autonomous AI agent space. Microsoft has its own agentic AI ambitions built around Copilot and its broader Azure infrastructure.

AMD reaching a $1 trillion market cap milestone, even briefly, is the kind of number that reframes a narrative. For years AMD was the scrappy challenger to both Intel and Nvidia. The catalyst being an AI agent that specifically needs CPU resources for VM orchestration is exactly the kind of detail that makes semiconductor analysts update their models.

For Intel, a 12% move in a two-day window is not nothing. Arm’s 17% rise reflects similar logic, with Arm architecture present throughout cloud infrastructure, and persistent VM workloads playing to its strengths in power-efficient computing.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.