The new prefabricated infrastructure leverages idle grid power to deliver 1–20 MW of AI inference compute within 12–16 weeks, speeding up deployment for heavy‑weight workloads.

Torus has unveiled its new “Foundry” system, a modular infrastructure solution that can deliver up to 20 MW of AI inference capacity in as little as four weeks, dramatically accelerating the rollout of high‑performance compute for demanding workloads.

What is the Foundry system?

The Foundry is a prefabricated, plug‑and‑play platform that taps into idle grid power to provide scalable AI inference compute ranging from 1 MW to 20 MW. Each module is designed for rapid deployment, with a typical build time of 12–16 weeks from order to operational status.

Key benefits for AI workloads

By leveraging existing power infrastructure, the system reduces the need for costly on‑site power upgrades. It also offers flexible scaling, allowing operators to add capacity in incremental steps as demand grows, without the long lead times traditionally associated with data‑center construction.

  • Accelerated deployment—up to 20 MW in under a month
  • Utilizes idle grid capacity, lowering energy costs
  • Modular design supports incremental scaling
  • Prefabricated units simplify logistics and installation

Potential impact on the AI ecosystem

The rapid availability of high‑density inference power could help AI developers meet the surging demand for real‑time model serving, especially in sectors like autonomous vehicles, video analytics, and large‑scale language model inference.

For readers interested in broader trends, see the Infrastructure coverage on uxc.news for related developments in modular data‑center technologies.

Torus launches Foundry to meet AI inference power demand

Track coverage on Momentum · More in Infrastructure.