Crusoe, the AI factory company, and ON.energy have announced a partnership to deploy 5 gigawatts of ON.energy's patented AI UPS technology across multiple hyperscale data center campuses, with commissioning set to begin in 2026 and extend into 2027.

The agreement positions the technology as a core component of Crusoe's expanding AI infrastructure buildout, addressing what both companies describe as a growing mismatch between the power demands of modern AI training facilities and the capabilities of the existing electrical grid.

A New Kind of Power Problem

AI data centers create load patterns that the grid was not designed to handle. During training runs, large clusters of GPUs ramp up and down in synchronized bursts, causing power demand to fluctuate hundreds to thousands of times a day.

From the perspective of a grid operator, a large AI training facility can swing between high and low consumption in ways that traditional industrial or commercial facilities simply do not. When those fluctuations are left unmanaged, they propagate back through the point of interconnection into the transmission system.

That places additional stress on grid equipment and limits grid operators' ability to manage other variables across the system at the same time.

The scale of the problem grows in proportion to the size of the data center campus, meaning that hyperscale AI infrastructure presents a qualitatively different challenge compared with earlier generations of large load customers.

What the AI UPS Technology Does

ON.energy's AI UPS is a medium-voltage uninterruptible power system that sits in line between power sources and AI data centers.

The company holds patents on the technology.

By positioning itself at that junction, the system effectively decouples the data center facility from the grid, creating a buffer that operates in both directions simultaneously.

On one side, the AI UPS protects data center equipment by riding through voltage faults without disconnecting from the grid during disturbances.

On the other side, it prevents the internal GPU load swings generated inside the facility from propagating outward and affecting the grid.

The system stores enough energy to power an entire campus for extended durations, which the companies say significantly reduces variability on the grid during peak hours.

The technology has been validated against the Large Load Interconnection requirements set by ERCOT, the grid operator for most of Texas. Specifically, ON.energy states the AI UPS meets the grid stability and voltage ride-through standards that ERCOT has proposed as the baseline for responsible large-load interconnection in Texas, including compliance with NOGRR 282 and ERCOT's Low Voltage Ride-Through protocols.

Why Crusoe Selected the Technology

Chris Dolan, Chief Data Center Officer of Crusoe, said the company evaluated the full range of available power solutions before settling on ON.energy's approach.

According to Dolan, the deciding factor was the technology's ability to address multiple problems at the system level rather than in isolation.

"This technology protects our infrastructure, it protects our generation assets, and it aligns with where grid requirements are heading," Dolan said in the announcement.

Dolan framed the selection as consistent with Crusoe's broader energy-first positioning, describing the result as turning data centers into "part of the solution to support long-term grid stability and flexibility" rather than simply a source of demand stress on the grid.

Alan Cooper, Co-Founder and CEO of ON.energy, described the current moment as one the technology was specifically built for. "AI UPS is built for this moment," Cooper said.

"It keeps data centers online through disturbances while ensuring the grid sees stable, predictable behavior." Cooper also characterized Crusoe's energy-first approach as a genuine competitive advantage rather than a marketing position.

Scale and Timeline

The 5-gigawatt deployment will span multiple hyperscale campuses.

The companies did not identify specific locations beyond indicating that the ERCOT grid requirements validation is relevant to Texas operations. Commissioning is scheduled to begin in 2026 and continue through 2027.

The scale of the deployment places it among the larger announced power infrastructure commitments in the AI data center sector. Crusoe separately announced in July 2026 a 1.0 gigawatt AI data center campus in Childress, Texas, developed in partnership with Lancium, indicating the company is pursuing multiple large-scale infrastructure projects in parallel.

Company Backgrounds

Crusoe describes itself as an AI factory company focused on providing AI infrastructure through large-scale energy sourcing, AI-optimized data center construction, and an AI cloud platform.

The company is headquartered in Denver.

ON.energy describes its business as building power infrastructure for grid-safe data centers and mission-critical facilities.

The company supplies and operates hyperscale power systems and states its track record spans industrial, manufacturing, infrastructure, transportation, and grid-scale storage applications.

ON.energy develops projects worldwide and is headquartered in Miami. The AI UPS product line is supported by patented technology and proprietary software.