Josh Magnuson does not care about your AI bubble thesis. The general manager of Ecolab's global water solutions business said as much to Bloomberg this week, days after his company agreed to pay $4.75 billion for CoolIT Systems. "If there's an AI bubble or not, I don't think it matters," Magnuson said. "Whether it's a AI bubble or a lot of little bubbles that pop, the overall demand curve is going to continue to move forward regardless."
That is a bold claim from someone whose employer just wrote a very large check. It also happens to be the most important strategic statement anyone in the cooling industry has made this year. Because the question of whether AI infrastructure spending is a bubble, and whether cooling vendors will get caught holding the bag when it pops, has hung over this entire sector like a fog since late 2025. Magnuson's answer, backed by nearly five billion dollars in conviction, is that the question itself is wrong.
Start with what KKR just pulled off. The private equity firm took a majority stake in CoolIT in 2023 when the Calgary-based company was valued at roughly $270 million. Three years later, Ecolab is paying $4.75 billion. That is approximately a 15x return on KKR's original equity investment, inclusive of distributions. In private equity, that kind of multiple on a three-year hold makes careers. It also makes a statement about what the smartest money on earth thinks about the trajectory of liquid cooling demand.
CoolIT is expected to generate around $550 million in sales over the next twelve months. Ecolab is paying roughly 29x and 24x estimated next-twelve-month and 2027 adjusted EBITDA, respectively. Those are steep multiples for a company that, until recently, made cooling gear for gaming PCs. But Ecolab CEO Christophe Beck sees CoolIT as the missing piece that doubles the company's Global High-Tech addressable market from $5 billion to $10 billion. (We covered the deal mechanics in detail here.)
This piece is about the thesis underneath.
Magnuson's argument rests on a structural observation that bubble skeptics tend to overlook. Data center cooling demand is not purely a function of new AI model training runs. It is driven by three overlapping forces: new capacity buildouts for AI workloads, the retrofit cycle for existing facilities upgrading to higher-density racks, and the ongoing migration from air cooling to liquid cooling as chip thermal design power keeps climbing. Even if venture-backed AI startups crater tomorrow, the hyperscalers have already committed tens of billions in capital expenditure to facilities that will need cooling for the next decade. Microsoft, Google, Amazon, and Meta are not turning those buildings into warehouses.
Magnuson told Bloomberg that cooling demand will grow 20% or more per year "for the foreseeable future." The reasoning is straightforward. Every new generation of GPU and accelerator chip runs hotter than the last. NVIDIA's GB200 NVL72 racks push north of 120kW per rack. At those densities, air cooling is physically insufficient. You cannot move enough air through a cabinet to dissipate that much heat without turning the data hall into a wind tunnel. Liquid cooling, whether through direct-to-chip cold plates, rear-door heat exchangers, or full immersion, is the only path forward. And once a facility converts, the cooling infrastructure becomes a recurring revenue stream: maintenance, coolant chemistry, monitoring, replacement parts.
CoolIT's core product line centers on direct liquid cooling, or DLC. The company designs and manufactures cold plates, coolant distribution units (CDUs), liquid loops, and rack manifolds that bolt onto server hardware from the major OEMs. Their cold plates sit directly on GPU and CPU packages, circulating a liquid coolant that absorbs heat at the die level and transfers it to a facility-level rejection loop. This is fundamentally more efficient than air cooling because liquid has roughly 1,000 times the volumetric heat capacity of air. CoolIT has custom-designed solutions validated by NVIDIA and AMD, which means their hardware is already embedded in the reference architectures that hyperscalers and colocation providers spec into new builds. That validation process takes 12 to 18 months per chip generation. It is a real barrier to entry.
Ecolab, for its part, already operates across more than 1,000 data centers globally, managing water treatment, chemical programs, and facility-level thermal management. The CoolIT deal gives them the rack-level hardware piece they were missing. The combined entity can now offer an end-to-end fluid management and cooling platform, from the chip to the cooling tower. That vertical integration is what Beck means when he talks about "a complete cooling solution that improves performance and reliability while reducing water and energy use."
Private equity exits tell you more about market conviction than press releases ever will. KKR acquired CoolIT through its Global Impact Fund II, and the 15x return in roughly three years makes this one of the most profitable PE deals in the data center infrastructure space. Period. When KKR bought in, liquid cooling was a niche technology serving maybe 5% of deployed racks. Today, every major hyperscaler has committed to liquid cooling for their next-generation AI clusters. The market moved beneath KKR's feet in exactly the direction they bet it would.
Mubadala, the Abu Dhabi sovereign wealth fund that held a minority stake in CoolIT, is also cashing out in this deal. When both a top-tier PE shop and a sovereign wealth fund are willing to sell at $4.75 billion, it usually means one of two things: they think the asset is fairly priced, or they think the buyer is overpaying. Given that Magnuson is projecting 20%+ annual growth, Ecolab clearly believes the $10 billion addressable market will make those 29x multiples look cheap in hindsight.
There is another wrinkle to this story that has received almost no attention. On March 12, Ecolab announced a global 10 to 14% energy surcharge on all products and services, effective April 1. The company cited oil prices rising close to 60% and European natural gas surging nearly 80% since late 2025, driven by Middle East supply disruptions. Every Ecolab customer, across every business line and every country, will pay more starting next week.
For data center operators already budgeting for a $4.75 billion acquisition premium to flow through CoolIT's pricing, this adds another layer of cost pressure. Ecolab is simultaneously telling the market that cooling demand is unstoppable and that the cost of delivering cooling services is going up. Both of those things can be true at the same time. But it means the data center industry's shift to liquid cooling will be more expensive than anyone modeled six months ago.
What Magnuson is really saying, stripped of corporate-speak, is that the cooling industry has decoupled from the AI hype cycle. The physical infrastructure is being built. The chips are getting hotter. The racks are getting denser. Whether the AI applications running on those chips generate the returns their backers hope for is a separate question entirely, one that affects software valuations, not plumbing. Ecolab is in the plumbing business. And the plumbing business just got a $4.75 billion vote of confidence.
The cooling industry spent the last two years wondering if it was riding a real wave or a sugar high. Ecolab's answer is to buy the biggest liquid cooling company on the market at 29x EBITDA, layer on a double-digit energy surcharge, and tell Bloomberg the bubble question is irrelevant. That is not hedging. That is a company putting nearly five billion dollars on the table and daring the skeptics to call their bluff.