Torus raises US$500 million for AI computing units using spare grid capacity
Magnetar led the financing, which will support factory expansion in Salt Lake City and Foundry deployments across the United States.
Energy storage company Torus has secured US$500 million in financing and introduced Foundry, a modular computing system designed to run AI workloads at businesses and factories with unused electrical capacity. The company said Magnetar led the investment, with participation from ICONIQ, Origin Ventures, Ridgeline Partners and Trinity Capital.
Torus plans to use the funding to expand its GigaOne manufacturing facility in Salt Lake City and deploy Foundry systems across the United States. Bloomberg identified the financing as a Series C round. No valuation was disclosed.
Foundry combines computing equipment with cooling, networking, storage and power infrastructure in prefabricated systems ranging from 1 to 20 megawatts. According to Torus, the units use available capacity at existing transformers without requiring another grid connection. Its Reserve battery and flywheel system manages peak demand and provides backup electricity.
The systems target inference, which involves running AI models after training. Torus said thousands of GPUs are already operating in Foundry units, with plans for more than 47,000 by the end of 2027. Initial customers include Hydra Host and an unnamed AI cloud provider.
Torus estimates that deployment takes 12 to 16 weeks from signing a contract to running AI workloads. It also says the cooling system operates without water. At GigaOne, which employs nearly 400 people, the Foundry production line can produce 500 megawatts of IT load; Torus plans to increase that capacity to 1 gigawatt.
