At a glance
- The company announced the raise and the product on Nov. 12 and said the capital will support scaling Foundry.
- Foundry units ship in shipping-sized containers and are installed behind the meter, drawing on headroom at existing transformers rather than a new grid connection.
- For allocators underwriting AI compute, the pitch moves the scarce input from interconnection queue position to spare transformer capacity at a host site and the terms the local utility sets for using it.
Torus, a Utah-based energy storage developer, raised $500 million in Series C funding led by Magnetar and launched Torus Foundry, a prefabricated modular data center product for AI inference at 1MW to 20MW per site. The company announced the raise and the product on Nov. 12 and said the capital will support scaling Foundry.
Foundry units ship in shipping-sized containers and are installed behind the meter, drawing on headroom at existing transformers rather than a new grid connection. Each is paired with Torus Reserve, the company's flywheel-and-battery storage system, which Torus says smooths AI inference loads so more capacity fits on an existing connection and provides backup power during outages. Torus puts site preparation at roughly a week, installation and commissioning at one to two days, and the full run from signed contract to operations at 12 to 16 weeks.
For allocators underwriting AI compute, the pitch moves the scarce input from interconnection queue position to spare transformer capacity at a host site and the terms the local utility sets for using it. If it holds at scale, the diligence question changes from grid access to headroom at a specific transformer and who controls it. The product splits three ways — a powered shell without GPUs, a turnkey bare-metal unit with GPUs, or dedicated single-tenant inference capacity — and each carries a different counterparty and revenue model.
Torus prefabricates the units at a 545,000 sq ft factory in Salt Lake City, combining power, storage, cooling, networking and compute into one unit with closed-loop cooling. The facility is built to produce 500MW of IT load, and Torus says it plans to expand that to 1GW. The first announced customers are Hydra Host, a GPU cloud provider, and a second, unnamed AI cloud provider. Torus says thousands of GPUs are already deployed in Foundry units and that it plans more than 47,000 by the end of 2027.
Torus comes to compute from storage, and Foundry's case to a customer rests on fitting inside power a site can already get. "Most providers are bringing power to the chips," said Nate Walkingshaw, the company's chief executive and founder. "Foundry brings the chips to power." Walkingshaw said building net-new infrastructure is slower and can raise costs, with delivery measured in years rather than Torus's 12-to-16-week window.
A $500M raise for data centers that fit on existing power
Series C raise, unit size and factory IT-load capacity, announced Nov. 12, 2026
Series C
$500M
Capacity per site
1-20MW
Factory IT load today
500MW
Factory IT load planned
1GW
GPUs targeted by end-2027
47,000
Foundry brings the chips to power.Nate Walkingshaw, chief executive and founder of Torus
Reported by the Data centers & fiber desk from Data Center Dynamics. Published . Corrections