Every conversation about AI infrastructure eventually gets to power. Far fewer get to water—but water is becoming just as real a constraint, and in some regions, a bigger one. As AI deployments grow, so do concerns around water use, energy consumption, and cooling capacity.

Atlas One provides a fully waterless, single-phase immersion cooling architecture with a targeted PUE of 1.0-1.1. This post looks at why that matters—not just as an efficiency number, but as a real operational and siting advantage as water becomes a harder constraint to build around.

Water Is Becoming an AI Infrastructure Bottleneck

Power gets most of the attention in AI infrastructure planning, but water is quickly becoming a real constraint. Global data-center water use is roughly 560 billion liters a year, and that figure could rise to 1.2 trillion liters by 2030—enough to supply more than four million U.S. households. A meaningful share of new and under-construction data centers already sit in regions facing rising water scarcity, which means water availability is no longer just an environmental talking point. It’s an operational one.

That matters directly for anyone planning AI infrastructure: a facility that depends on water for cooling is a facility whose long-term viability is tied to a resource that’s getting harder to count on in exactly the regions where infrastructure is expanding fastest.

The Water-Energy Tradeoff Most Cooling Approaches Can’t Escape

Conventional data-center cooling often forces a choice. Water-based and evaporative cooling systems are typically more energy-efficient, but they consume real, ongoing volumes of water.

Air-based systems conserve water, but they require more electricity and hit a hard density ceiling well before modern AI accelerator racks need them to. Most infrastructure decisions end up picking one tradeoff or the other—efficient and thirsty, or dry and power-hungry.

Atlas One is built to avoid that tradeoff entirely rather than pick a side of it. The platform runs on single-phase immersion cooling: no cooling towers, no evaporative water consumption, and no need to compromise compute density to accommodate conventional cooling infrastructure.

Efficiency doesn’t come at the cost of water, because there’s no water in the loop to begin with.

What “Waterless” Actually Means

“Waterless” isn’t a marketing simplification here—it describes the actual cooling loop. Atlas One’s single-phase immersion architecture circulates a dielectric fluid in a closed system, in direct contact with the hardware it’s cooling.

There’s no cooling tower drawing in outside air and evaporating water to reject heat, and no ongoing water makeup requirement to replace what’s lost to evaporation. The cooling loop simply doesn’t consume water as part of how it operates.

That’s a structurally different approach from evaporative or water-assisted cooling systems, which are effective specifically because they trade water for cooling capacity. Atlas One’s architecture gets its cooling capacity from direct fluid contact instead, which is what allows it to reach a targeted PUE of 1.0-1.1 without that tradeoff.

Why This Matters Beyond the Efficiency Number

A low PUE achieved through evaporative cooling and a low PUE achieved through a waterless architecture look identical on a spec sheet. They are not identical in what they expose an organization to.

Facilities that depend on water face a growing list of operational risks that have nothing to do with compute performance: local water availability, drought-driven restrictions, permitting friction in water-stressed regions, and increasing regulatory and public scrutiny over industrial water use tied to AI expansion. None of that shows up in a PUE number, but all of it shows up in how easily a facility can be sited, approved, and kept running through the life of a deployment. Removing water from the equation removes that entire risk category, not just a utility bill line item.

What Makes Atlas One Different 

Waterless operation isn’t a separate feature bolted onto Atlas One—it’s a direct consequence of the same single-phase immersion architecture behind everything else in this series.

It’s what lets the platform deploy quickly without site-specific water infrastructure, what keeps PUE low enough to maximize token output per square foot, and it uses the exact same cooling mechanism that removes the thermal ceiling limiting air-cooled AI hardware.

One architecture, four different advantages.