Absorb compute load swings

Your questions, answered.

• What is Torus Foundry?

  Foundry is a prefabricated compute unit deployed with energy storage on existing grid interconnections. We handle every step of manufacturing and deployment, speeding up time to token. 

• How does Foundry manage AI workload swings?

  AI workloads spike and drop in seconds. Every Foundry is paired with Reserve, a hybrid energy storage system with a flywheel and an LFP battery. Reserve supplies or absorbs the difference as your compute load changes, so Foundry runs your compute without adding strain to the grid.

• How does Reserve respond faster than traditional batteries or generators?

  Reserve’s flywheel has a 10C power rating, 20x higher than traditional battery storage, and it delivers full power in milliseconds. A generator has a 5–10 second transfer gap before it picks up load, and AI workloads can spike and drop in that time. Reserve covers the swing as it happens, and the flywheel is built for the fast cycling that wears down traditional batteries.

• Do I need a new grid connection to run AI workloads with Foundry?

  No. Foundry installs on developed land with your existing grid connection. Reserve balances the load swings on that connection, so you can run compute without waiting on a new one.

• How can you run compute without load curtailment?

  Available headroom is spread across many feeders and substations. Foundry modules fit those pockets. Reserve’s batteries discharge at peak, lowering a site’s peak load and freeing grid capacity without new utility infrastructure. That freed capacity can serve new load, like compute.