Torus, a Salt Lake City-based manufacturer of distributed power and compute infrastructure, launched Foundry on November 11, 2026. The modular AI infrastructure is designed to meet growing demand for inference compute by delivering capacity to existing sites in 12 to 16 weeks, according to a company statement.
Foundry provides 1 to 20 megawatts of inference capacity per site. The prefabricated system integrates power, storage, cooling, networking, and compute, allowing it to be installed behind the meter using available headroom at individual transformers. This approach aims to utilize unused grid capacity between substations and existing commercial or industrial buildings without requiring new infrastructure upgrades.
The company's Salt Lake City manufacturing facility, GigaOne, produces the Foundry units. The 545,000-square-foot facility opened its first production line in early 2026 and has opened a new line every 21 days throughout 2026. The current production line is capable of producing 500 MW of IT load, with future plans to expand to 1GW.
To support the launch, Torus secured $500 million in growth capital led by Magnetar, with participation from ICONIQ, Origin Ventures, Ridgeline Partners, and Trinity Capital. The funds are intended for the expansion of GigaOne and the deployment of Torus Reserve and Torus Foundry across the United States.
Nate Walkingshaw, founder and CEO of Torus, stated that most providers bring power to the chips, whereas Foundry brings the chips to power. He noted that the technology applies years of expertise in making the grid flexible and resilient to compute infrastructure, providing capacity in weeks using power that is already available.
The shift toward AI inference is driving demand for distributed edge compute closer to where demand exists. McKinsey projects that data center demand for AI inference will grow 35 percent annually through 2030, compared to 22 percent for training.
Torus plans to have more than 47,000 GPUs deployed in Foundry units by the end of 2027. The Foundry system features a waterless, closed-loop cooling system that uses 95 percent less water than evaporative coolers and includes a system for donating GPU time to medical research through Folding@home.
