Torus launched Torus Foundry on November 9 to provide modular AI inference compute that utilizes power already available at existing sites. The company announced it has raised $500 million in funding, led by Magnetar, to take the technology nationwide.
The Foundry units are prefabricated modules containing power, storage, cooling, networking, and compute. These units are built at GigaOne in Salt Lake City. According to Torus, site preparation takes approximately one week and installation takes one to two days, allowing a transition from contract to token in 12 to 16 weeks.
Each Foundry unit pairs with Reserve, a proprietary energy storage and management system. This system is designed to smooth out the spiked electricity loads typical of AI inference, providing a steady load to the grid. Torus installs these systems at commercial and industrial sites to function as distributed energy resources or virtual power plants.
The company's initial customers include Hydra Host and another AI Cloud provider, which currently run thousands of GPUs. Torus aims to reach 47,000 GPUs by the end of 2027. Funding for the venture was provided by Magnetar, ICONIQ, Origin Ventures, Ridgeline Partners, and Trinity Capital.
Torus previously announced the renaming of its flagship products, including Torus Reserve, Torus Protect, and Torus Arc, on November 9.
