Torus Launches Modular AI Compute Line With $500 Million Magnetar-Led Funding in Utah, US

Torus, a Salt Lake City-based manufacturer of distributed power and compute infrastructure, announced that it has raised USD 500 million in growth capital led by Magnetar.

The financing coincided with the debut of the company's newest product, Foundry, a modular AI infrastructure system built for inference workloads.

12-Week AI Inference Deployment Target

The company positions the launch around a simple premise: enterprises need inference capacity faster than traditional construction can deliver it.

Each Foundry unit provides 1 to 20 megawatts of inference capacity per site and moves from signed contract to running AI workloads in 12 to 16 weeks, a timeline Torus contrasts with the years required to build a conventional hyperscale campus.

Why Inference Is Reshaping Where Compute Goes

Torus frames Foundry as a response to a structural change in AI demand.

As workloads shift from training models to serving inference for enterprises, compute is being pulled toward on-premises and edge sites closer to where users are. Inference runs continuously, responds in real time, and lends itself to distributed deployment, according to the company.

Market projections cited by Torus support that trajectory. McKinsey data referenced in the launch forecasts annual growth of 35 percent in data center demand for AI inference through 2030, outpacing the 22 percent projected for training. Enterprise requirements add another dimension.

Torus notes that its customers want security, protection of intellectual property, and control over where their workloads execute, which, the company says, a distributed architecture can satisfy more readily than centralized capacity.

Unlocking Headroom Behind the Meter

The core of Torus's argument is that power, not chips, is the binding constraint on AI infrastructure today, and that the constraint is one of available capacity rather than insufficient generation. Foundry addresses this by using existing interconnections to tap unused grid capacity sitting between substations and existing commercial and industrial buildings.

Founder and CEO Nate Walkingshaw distilled the approach in a statement: "Inference doesn't have to wait. We've spent years learning how to make the grid more affordable, flexible, resilient, and intelligent, and Foundry applies that to compute capacity in weeks, with power that's already available." He added, "Most providers are bringing power to the chips. Foundry brings the chips to power."

Walkingshaw said the industry has largely tried to solve AI's power problem by building net-new infrastructure, an approach he described as costly and slow to bring compute online. Torus instead treats the situation as a capacity problem before it is a generation problem, contending that many sites already have unused grid capacity that its systems can make usable.

Because Foundry installs behind the meter, drawing on available headroom at individual transformers, the company says the units add steady, high-load-factor demand without triggering the infrastructure upgrades that often produce rate increases and without compromising grid reliability.

Prefabrication From a Utah Factory Floor

Foundry units are prefabricated at GigaOne, Torus's manufacturing facility in Salt Lake City, a step the company says removes construction time from deployment sites and shortens time to operation. The smallest unit measures roughly the size of a small cargo van, and the modular systems require no building construction at all.

On-site timelines are correspondingly compressed: site preparation takes about a week, while installation and commissioning take one to two days. The product is designed, engineered, and manufactured in Utah by Torus's team of nearly 400 employees, with key components built in-house.

Storage and Load-Shaping Built Into Every Unit

Every Foundry deployment ships paired with Torus Reserve, the company's energy storage and management product. Reserve delivers grid-enhancing technologies that Torus groups into three capabilities: inertial-based firm power, grid flexibility with load curtailment, and ramp-rate support.

These functions are executed through what Torus calls load-shaping recipes, predefined controls that govern how power is dispatched. According to the company, the recipes shape loads to manage demand, support grid stability, and ramp power in milliseconds.

The integrated inertial flywheel and battery storage and management serve a dual purpose. They cover demand peaks so that more compute can fit on an existing interconnection, and they supply backup power during grid outages.

A Single-Provider, Full-Stack Architecture

Torus describes Foundry as a vertically integrated offering spanning fiber to site, chiller to coolant distribution unit, and grid to rack.

The stack comprises three products: Torus Reserve for load shaping and redundancy, Torus Foundry for inference compute, and Torus Protect for integrated networking and security. Every deployment is turnkey, with ongoing operations and maintenance handled by Torus itself rather than left to the customer.

Three Commercial Models and a Launch Cohort

Foundry debuts with first customers Hydra Host and another AI cloud provider and is offered in three commercial models.

The Strategic model is a powered shell, supplying power and enclosure without GPUs; the Partner model is a turnkey bare metal unit that bundles power, enclosure, and GPUs; and the Direct model delivers sovereign inference capacity straight to enterprise customers, with single-tenant hardware allocated exclusively to one company's workloads.

Taken together, the launch and the USD 500 million Magnetar-led round position Torus to scale a distributed model of AI capacity delivery, one that leans on existing grid interconnections, factory-built modular hardware, and compressed deployment timelines rather than new construction.