Torus Launches Foundry Modular AI Inference Infrastructure with $500M Funding from Magnetar
Salt Lake City-based Torus launches Foundry, a prefabricated modular system that brings AI inference compute to existing grid capacity in 12-16 weeks, backed by $500M led by Magnetar Capital.

By TechWhoKnows
4 min read
Salt Lake City-based Torus has launched Foundry, a modular AI inference infrastructure platform designed to deploy compute capacity in weeks rather than years by using power that already exists on the grid.
The company also announced $500 million in growth capital led by Magnetar Capital, with participation from ICONIQ, Origin Ventures, Ridgeline Partners and Trinity Capital. The funding will expand Torus's GigaOne manufacturing facility and nationwide Foundry deployments.
From contract to tokens in 12-16 weeks
Foundry packages power, storage, cooling, networking and compute into prefabricated units built at Torus's 545,000-square-foot GigaOne facility in Utah. Each unit delivers 1-20 megawatts of inference capacity and installs behind the meter at existing commercial and industrial sites, using available headroom on transformers without requiring new grid infrastructure.
"The industry has been trying to solve AI's power problem by building net-new infrastructure, which causes costs to potentially rise and takes years to bring compute online," said Nate Walkingshaw, CEO and Founder of Torus. "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."
Site preparation takes about a week, with installation and commissioning completed in one to two days. The smallest system is roughly the size of a small cargo van.
Energy storage smooths inference load
Every Foundry unit pairs with Torus Reserve, the company's energy storage and management product. Reserve uses inertial flywheel and battery storage to cover demand peaks, allowing more compute to fit on an existing interconnection while providing backup power during grid outages. Load-shaping recipes govern how power is dispatched, presenting steady, flexible load to the grid instead of the spiky EKG-like pattern of raw AI inference workloads.
First customers and deployment targets
Hydra Host and another unnamed AI cloud provider are the first Foundry customers, with thousands of GPUs already running. Torus plans to deploy more than 47,000 GPUs in Foundry units by the end of 2027. The GigaOne production line is online and capable of producing 500 MW of IT load, with plans to expand to 1 GW.
Three deployment models
Foundry is available in three configurations: Strategic (powered shell without GPUs), Partner (turnkey bare metal with GPUs), and Direct (sovereign single-tenant inference capacity for enterprise customers).
Manufacturing and sustainability
Foundry units are designed, engineered and manufactured in Utah by Torus's team of nearly 400 employees. The closed-loop cooling system uses 95 percent less water than evaporative coolers. Each system also donates GPU time to medical research through Folding@home.
Torus, founded in 2021, has raised $767 million to date and holds utility partnerships under MOU with PacifiCorp covering 500 MW. The company's energy storage systems have achieved 99.9 percent fleet uptime with a 25,000-cycle lifespan on its Torus Spin flywheel technology.
Sources
- 1Torus press releasetorus.co
- 2Torus blog post by CEO Nate Walkingshawtorus.co