AI's heat wave sparks a 10x jump in liquid cooling demand

The boom in artificial intelligence is creating an urgent, secondary demand for advanced liquid cooling systems, as the immense heat from new AI servers overwhelms traditional cooling methods and threatens to cap growth.

The International Energy Agency estimates that electricity consumption from data centers could roughly double by 2030, a surge driven almost entirely by AI. This power draw translates directly into heat, creating a critical bottleneck that companies are now spending billions to solve.

This power consumption challenge is forcing a rapid shift in how data centers are designed and built. China recently brought a $228 million, 24-megawatt underwater data center online, which uses seawater for cooling and reduces electricity use by nearly 23 percent. In another sign of the buildout, Applied Digital recently announced it had reached 1 gigawatt of contracted capacity for its data centers.

For investors, the thermal limits of silicon are creating a new "picks and shovels" market. The focus is shifting from just the AI chips themselves to the essential infrastructure that allows them to operate, potentially driving significant returns for companies that solve the heat problem.

The End of Air Cooling

For decades, data centers have relied on air conditioning to manage heat, a method that is becoming untenable in the face of AI's power density. Chips used for AI training and inference operate at much higher temperatures than general-purpose CPUs, rendering air cooling inefficient and costly. This has led hyperscalers and enterprise data centers to explore and implement direct-to-chip or immersion liquid cooling solutions.

Companies like Eaton Corporation (NYSE:ETN) are seeing surging demand for their power management and electrical infrastructure systems, which are fundamental to these new cooling architectures. Eaton provides critical components like switchgear and power distribution units that can handle the higher energy loads required by both the AI servers and their associated liquid cooling systems. While the company's stock saw some pullback in late 2025, its role in the AI infrastructure buildout remains central, as noted by the Janus Henderson Forty Fund in its Q4 letter. The shift represents a fundamental change in data center design, where thermal management is now a primary consideration alongside computing power.

New Solutions From Sea and Land

The industry is responding with a range of solutions, from the conventional to the creative. The underwater data center off the coast of Shanghai, powered by an adjacent offshore wind farm, is a prime example of this new thinking. By placing the servers beneath the sea, the project developers eliminated the need for fresh water and reduced the land footprint by 90 percent compared to a traditional facility.

On land, companies are retrofitting existing facilities and designing new ones specifically for liquid cooling. This involves significant capital expenditure but is increasingly seen as a non-negotiable cost of participating in the AI race. The buildout is not just about cooling the chips themselves but also about upgrading the entire power chain to support racks that can draw over 100kW, a tenfold increase from just a few years ago. This trend is a direct tailwind for specialized electrical and cooling component suppliers who are now integral to the AI value chain.

This article is for informational purposes only and does not constitute investment advice.