Mount Thor Raises $45M for Apple-Based AI Infrastructure

Mount Thor asks a user to type a machine's name before releasing it. The confirmation has a practical purpose: the company's command-line tool will wipe that Mac when the user gives it back. For an AI team renting computers, the end of an allocation can also be the end of the working environment built on it.

That small operating detail sits inside a much larger financing announcement. Mount Thor, a San Francisco startup founded in 2026, announced $45M in equity across Seed and Series A financings on November 12, 2026. CRV and Conviction co-led the Series A, while Chad Byers and Offline Ventures co-led the Seed. Mount Thor also disclosed separate non-dilutive financing for hardware purchases, without specifying its amount.

Founder and CEO Peter Voell is building cloud infrastructure for AI teams that need Apple computers for research and deployment. The company also targets Apple software testing and open-source inference. Its financing brings an ordinary physical obligation into the agent economy: more software doing work can mean more computers to buy, configure, allocate and eventually prepare for someone else.

What Mount Thor's $45M finances

The announced equity total covers 2 funding stages. Mount Thor did not disclose the amount allocated to either stage, a valuation, or the terms of its additional hardware financing. CSC Leasing, Enea Capital, Long Journey Ventures and Upper90 participated in the financings. The announcement also names angels affiliated with several technology companies, a different relationship from those companies making institutional investments.

Mount Thor says the proceeds will support a larger compute footprint, a growing team and an inference lab focused on open-source models running on Apple silicon. The company reports deploying its first data center in less than 6 months. That is a company-reported operating milestone; the announcement supplies no customer count, revenue figure or independently measured fleet capacity.

Hardware financing is particularly relevant to this business model because capacity has to exist somewhere before a customer can rent it. The commercial work spans both acquiring equipment and making it useful repeatedly. A fleet operator must keep track of what is available, which environment belongs to whom and what can safely be reused. Buying another machine begins that responsibility.

Offline Ventures lists Mount Thor in its portfolio as a macOS cloud platform for AI workloads and agent training. The investment places the firm alongside backers funding the operational layer around agents, where a useful application also depends on having a suitable computer available when the work starts.

Why agents need different kinds of Mac environments

Mount Thor offers dedicated bare-metal Macs and isolated macOS virtual machines. Its official quickstart describes access through a web portal, the mthr command-line tool and Kubernetes. Customers can connect through SSH or a remote desktop, allowing the rented environment to support work that reaches beyond a simple text exchange with a model.

The company's use cases include training computer-use models, testing agents in macOS sandboxes and keeping agent instances running for weeks. Those are related workloads with different ideas of a useful machine. A researcher may need to repeat an experiment across many clean environments. A persistent agent may need to return to an environment that still contains the context of an ongoing assignment.

The lifecycle details help explain the service being sold. Mount Thor's documentation states that releasing a bare-metal machine wipes it, while deleting a virtual machine removes both the instance and its disk. A customer deciding when to release capacity is therefore also deciding what has finished. Convenience in acquiring a computer makes that second decision easier to overlook.

For an AI product team, this creates a concrete design question: which parts of the working environment should remain available, and which should be reset before the next task? Mount Thor supplies infrastructure that lets teams make those choices. The choices themselves remain tied to the product and to the people whose work the agent is carrying forward.

Cloud Macs already have competition

Amazon EC2 Mac instances already provide on-demand macOS environments for developing, building and testing Apple applications. Cloud access to Macs is an established service category. Mount Thor enters that market with an emphasis on the additional workloads created by computer-use agents and AI research.

That emphasis gives prospective buyers a specific basis for comparison. An Apple development team, a computer-use research lab and a provider of persistent agents may all need macOS, yet ask different questions about images, allocation, continuity and management. The value of a specialized provider depends on how well its service fits those operating requirements. An agent-focused description has to become a practical experience for the team paying the bill.

Brittany Walker, a General Partner at CRV, describes Mount Thor's opportunity through the relationship between autonomous agents and local and cloud compute. The observation matters because an agent's usefulness can depend on the environment in which it acts. A model capable of operating software still needs access to that software and a computer configured to run it.

Voell's company manifesto expands the ambition into Apple-silicon inference. That remains a company thesis about suitable workloads and economics. The financing announcement provides no comparative benchmark that would establish a general cost or performance advantage, so buyers will need evidence tied to their own workloads.

The fleet becomes part of the customer relationship

Mount Thor's expansion connects research demand with the quieter obligations of operating a service. Someone must ensure that the machine a customer requests is available and that its environment can be reached and managed. When an agent depends on the same environment over a longer period, those obligations follow the work beyond its first successful demonstration.

This is a useful commercial distinction for infrastructure buyers. They are paying for access to equipment, but their engineers experience that purchase through allocation, connectivity, configuration and the handling of a machine's eventual release. A provider can earn a place in that workflow by making the routine parts dependable enough that a team can concentrate on its experiments and products.

Mount Thor plans to add both capacity and people, with its careers page providing the public route into that team. As the fleet grows, the person choosing to end a machine allocation will still face the same small confirmation. The system has to make clear which computer is being returned, and the customer has to know which work is ready to leave it.