Data Center Cooling Stocks

Why cooling became the constraint on AI data centers

A conventional data center rack drew perhaps five to ten kilowatts. A rack filled with AI accelerators can draw more than a hundred. Every watt consumed becomes heat that must be removed, and air can no longer remove it fast enough. That single physical fact turned thermal management from a background utility into one of the constraints deciding how quickly AI capacity can be built.

Liquid cooling is the response, and it comes in two forms. Direct to chip circulates coolant through cold plates mounted on the processors, removing heat at the source while leaving the rest of the room air cooled. Immersion submerges entire servers in a dielectric fluid, which is more efficient still and more disruptive to how a facility is built and serviced. Both require the surrounding infrastructure to change: coolant distribution units, piping, heat exchangers and the rear door units that capture what escapes.

Thermal infrastructure is the layer that rejects that heat to the outside world: chillers, cooling towers, air handlers and the industrial heating and ventilation equipment that has served commercial buildings for decades and is now being specified at densities it was never designed for. Water consumption has become a siting constraint in its own right, which is pushing designs toward closed loop and air cooled heat rejection despite the efficiency penalty.

Power distribution is the twin problem, and it is why power and cooling companies appear in the same theme. Getting a hundred kilowatts into a rack safely requires busways, switchgear, uninterruptible supplies and backup generation sized for the load. The efficiency of that chain determines how much of a facility's power reaches the processors rather than being lost or spent on cooling.

Server and rack integration sits at the intersection. The companies assembling complete AI systems ship them as an integrated thermal and mechanical design rather than as boxes, because at these densities the cooling architecture and the server architecture cannot be separated.

Demand comes from a single concentrated source: capital expenditure by cloud providers, model developers and the operators building facilities for them. That spending is disclosed quarterly, is revised, and is the reason this group can reprice sharply on a single earnings call from a customer rather than from any company in the theme itself.

Stocks are sorted by today's price change, strongest first. Click any column header to sort by another measure, or switch to Charts for a daily candlestick of every name. This is not a recommendation to buy or sell any security. All information is provided for educational purposes only, and past performance does not guarantee future results.