Three-Party Long-Duration Storage Model

Diving deeper into

Form Energy

Company Report
The 300 MW / 30 GWh Xcel Energy-Google project and Crusoe's 12 GWh capacity reservation establish a three-party model
Analyzed 5 sources

This model turns power procurement into a joint go to market motion, not a simple battery sale. Google and Crusoe are not just buyers of electricity, they are anchor customers that create enough demand to justify dedicated capacity, while Xcel handles grid integration and Form supplies the long duration storage block. That matters because a 100 hour battery is easiest to finance and deploy when a large load customer, a utility, and a storage vendor each solve a different piece of the project.

  • The Google project shows how this works on the grid. Google’s Minnesota data center commitment is paired with 1,400 MW of wind, 200 MW of solar, and a 300 MW, 30 GWh Form battery inside Xcel’s resource plan, so the new load helps pull new supply onto the system instead of just joining the queue.
  • The Crusoe deal pushes the same structure one step further. Crusoe reserved 12 GWh of Form capacity starting in 2027 for AI data centers, effectively treating multi day storage manufacturing slots like scarce infrastructure inventory that can be locked up in advance to support future campuses.
  • This is also a competitive wedge against other ways to serve AI load. Crusoe already builds energy intensive data center infrastructure and has spent heavily on owned capacity, while Form’s system is built for roughly 100 hours of discharge. That positions iron air against firm power options like small modular nuclear or fusion backed campuses, and against overbuilding shorter duration batteries.

The next step is a repeatable template where hyperscalers, utilities, and industrial operators pre contract dedicated multi day capacity before the site is even energized. If that happens, Form stops looking like a niche utility storage supplier and starts looking like a core infrastructure vendor for speed to power constrained AI and industrial buildouts.