Ecolab just paid $4.75 billion in cash for a Calgary-based company most investors have never heard of, and that single transaction tells you more about where AI infrastructure spending is headed than almost anything happening in chips right now.
The March 2026 acquisition of CoolIT Systems came at a price of 29 times next-twelve-months adjusted EBITDA. That is not the kind of multiple a serious industrial buyer pays on a whim. It’s a signal that liquid cooling has moved from a nice-to-have into a required piece of the AI data center stack, and Ecolab is not the only one who has figured that out.
The math behind the move is simple physics more than anything else. Modern AI training runs on GPU servers packed with far more processing density than traditional data centers were ever built to handle, and that density generates heat that conventional air cooling simply cannot dissipate fast enough. Nvidia’s newest Blackwell racks push thermal loads that force operators into direct-to-chip cooling, rear-door heat exchangers, or full immersion systems just to keep the hardware running. ASHRAE’s 2026 AI Data Center Energy Performance Framework now names these approaches explicitly as requirements, not options, for high-density AI deployments. MarketsandMarkets pegs the global liquid cooling market at $4.07 billion this year, growing to $27.65 billion by 2033, a 31.5% annual growth rate that would make most industries jealous.
Ecolab’s purchase of CoolIT is just the most recent entry in a run of consolidation that has been building for over a year. Trane Technologies has announced plans to acquire LiquidStack. Schneider Electric bought Motivair in 2025 specifically to build out its liquid cooling capabilities. Vertiv closed its acquisition of ThermoKey on June 12 and opened a new manufacturing facility in Malaysia on July 1 just to keep up with AI infrastructure orders. Every major player in industrial thermal management has either bought a specialist in this space or announced plans to. When that many strategic acquirers are willing to pay near 30 times forward earnings for private cooling companies, the public small and micro-cap names sitting in the same value chain tend to get repriced whether or not they’re involved in a deal themselves.
A handful of small-cap names sit right in the middle of this shift. Modine Manufacturing has spent the last three years transforming itself from a legacy automotive parts supplier into a company focused on data center thermal management, divesting older auto businesses along the way to sharpen that story. Limbach Holdings has leaned into the build-out from a different angle, highlighting its modular construction and prefabrication platform for data centers in a June announcement that positions the company squarely inside the high-density projects hyperscalers are commissioning right now. And nVent Electric, which has spent more than a decade building liquid cooling distribution and high-density power systems, saw organic orders jump roughly 65% in a recent quarter driven almost entirely by large cooling orders tied to hyperscaler programs. The company has already deployed more than a gigawatt of cooling capacity across its installed base.
The AI investment story so far has mostly been about chips, and understandably so. But chips are useless if you can’t keep them cool enough to run at full capacity, and that second half of the equation is where the next wave of investment dollars appears to be heading. Power, water, and industrial thermal management are becoming just as important to the AI buildout as the silicon itself, and the M&A activity happening right now is the clearest evidence yet that the biggest names in industrial equipment already see it that way. Modine, Limbach, and nVent aren’t household names, and that’s exactly the point. When a $77 billion company pays nearly $5 billion for a private cooling specialist, the small-cap names doing similar work for the same customers are the ones worth watching next.
