Mount Thor raises $45 million to build Mac data centers for AI agents

Mount Thor, a Bay Area startup that builds cloud infrastructure on Apple silicon, has raised $45 million across Seed and Series A rounds to build Mac data centers for AI agents, according to 9to5Mac's report. Its pitch to investors was that AI agents run very well on Apple silicon, and nobody rents those machines at data-center scale.

This piece covers who backed the company, how its managed and customer-managed Mac compute works, the inference angle tied to the M5 Ultra Mac Studio, and how Mount Thor positions itself against existing Mac cloud providers. It is written for teams building computer-use agents, Apple software, or open-weight model inference.

01CRV and Conviction lead the Series A

CRV and Conviction co-led the Series A, while Chad Byers and Offline Ventures co-led the Seed. CSC Leasing, Enea Capital, Long Journey Ventures and Upper90 also participated.

Angel investors from OpenAI, Meta, Cognition, Together AI and Vercel joined the rounds as well. Mount Thor also raised separate non-dilutive financing to pay for new hardware.

The company was founded this year and deployed its first data center in under six months. Its team comes from Cerebras, CoreWeave, Fluidstack, Nebius, OpenAI and xAI.

02macOS as elastic compute for agents

Mount Thor runs macOS on Apple silicon as elastic compute instances, the same way a cloud provider rents Linux VMs. Customers get on-demand fleets of Macs for work that only happens on macOS.

The company lists four workloads: training computer-use models in reinforcement learning environments, running secure macOS sandboxes for agents, and build and test farms for Apple software. It also offers long-running instances that let an agent work across a user's devices under one Apple ID.

The underlying argument is that agents are becoming macOS users. In that view, demand for Apple's platform scales with the number of agents rather than the number of people.

03Managed and customer-managed Mac fleets

Mount Thor offers two ways to use its service. With managed compute, a customer picks a machine class and an image and gets either a full bare-metal Mac or an isolated macOS VM with administrator access.

Managed machines can be provisioned through a web portal, a command-line tool, Kubernetes or the API. Billing is per Mac on a 24-hour basis.

With customer-managed compute, an organization runs its own Macs at a Mount Thor site using its own MDM. Mount Thor operates the facility and provides inventory by serial number, problem reporting and event notifications.

04M5 Ultra Mac Studio and open-model inference

The other half of the pitch is inference. The M5 Ultra Mac Studio shipped in August with up to 512GB of unified memory, which puts large open-weight models within reach of a single machine at a fraction of the power draw of a traditional GPU-based server.

Mount Thor says a focus of the new funding is building what it calls the best inference lab for open-source models on Apple silicon. Founder and CEO Peter Voell said that "with chips like the M5 Ultra, Apple silicon is becoming a high performance, more power- and cost- efficient platform for production inference."

Voell also called datacenter-scale access to Apple's operating system and silicon "a massive untapped opportunity in AI infrastructure." He argued that decades of investment in accessibility and human-computer interaction make macOS "the best place for agents to work and for humans to hold them accountable."

05Mac cloud rivals AWS and MacStadium

Macs in the cloud are not new. AWS has rented Mac instances for a few years, and MacStadium has offered Mac hosting for longer than AWS.

Mount Thor's distinction is the customer it targets: an AI agent that needs a Mac to do its job, with demand that grows with the number of agents rather than the number of developers. Brittany Walker, General Partner at CRV, said the company "sits at the intersection of two important shifts: the rapidly expanding capabilities of autonomous agents and the increasing demand for compute that can support both local and cloud workloads."

The company says the money will go toward expanding its compute footprint and building out the team. It plans to add capacity quickly to keep up with customer demand.