$4.75 billion in cash, at 29 times next-twelve-month earnings, for a company that makes cold plates. Ecolab announced the agreement to acquire CoolIT Systems, a Calgary-founded liquid cooling manufacturer, from private equity firm KKR on March 20, 2026 (Ecolab, 2026). Ecolab had guided to a third-quarter close. It closed the deal on July 2, ahead of that schedule (Ecolab, 2026). CoolIT's year-to-date sales had grown more than 100 percent by the time the deal closed (Ecolab, 2026).

That price tag is the story. Twenty-nine times next-twelve-month EBITDA and twenty-four times 2027 estimated EBITDA is a multiple you pay for a company riding a hyperscaler growth curve, not a multiple anyone pays for a plumbing supplier (Chemical Engineering, 2026). Ecolab has sold water treatment and hygiene services into data centers for years. Those contracts kept facilities compliant and running. None of them put Ecolab inside a conversation about whether a rack could handle the next generation of NVIDIA silicon.

Air stopped working before anyone budgeted for it

An NVIDIA GB200 NVL72 rack draws 120 to 130 kilowatts. Air handling systems run out of economic headroom well before that number, and most data center floor space was engineered for racks pulling 15 to 30 kilowatts (EBC Financial Group, 2026). Goldman Sachs estimates liquid-cooled AI servers went from 15 percent of deployments in 2024 to 54 percent in 2025, and projects 76 percent by the end of 2026 (Goldman Sachs, 2026).

The prevailing read on AI infrastructure spending has been about chips and power. NVIDIA allocation and grid capacity dominate the boardroom conversation, and both constraints are real. Fewer buyers have priced in that heat removal became the ceiling on how much of that chip supply a facility can run at density, without a retrofit that takes twelve to eighteen months of civil work.

Ecolab's acquisition is a bet that this ceiling is where the durable revenue sits, not in the chips themselves.

CoolIT was not a generic vendor to acquire

CoolIT has built its product roadmap around NVIDIA's own release cycle for more than one generation. Its cold plate technology is optimized for the GB200 GPU, and the company has shipped a coolant distribution unit sized to support twelve NVIDIA GB300 NVL72 racks (Sustainability Magazine, 2026). That is not a company Ecolab could have built from scratch on a two-year timeline. Buying it was faster than competing with it.

Ecolab chief executive Christophe Beck said the acquisition lets AI scale faster while respecting the communities and resources around it, extending Ecolab's reach from chip fabrication plants to the data centers running the finished chips (Smart Water Magazine, 2026). Strip the framing down and the mechanism is simpler. Ecolab bought the piece of the stack that determines whether its existing water contracts stay relevant.

Ecolab's own numbers make the stakes explicit. Global High-Tech, the division housing CoolIT, generated about $150 million in annual sales in 2021. Following this acquisition and an earlier one of water treatment firm Ovivo, that division is approaching $1.5 billion in 2026 annualized sales, with a target of $4 billion by 2030 (Smart Water Magazine, 2026). A ten-fold jump in five years does not happen through organic water treatment contract growth. It happens through acquisition, and CoolIT is the largest piece of it.

A water treatment contract that cannot speak to rack density is a contract a hyperscaler renegotiates without much friction.

The chemistry-hardware convergence has a shelf life question

Direct-to-chip liquid cooling is not the only architecture competing for the next generation of AI racks. Immersion cooling offers a higher thermal ceiling but requires a full rack redesign. Rear-door heat exchangers offer a lower-disruption retrofit path for operators not ready to rebuild (GigeNET, 2026). CoolIT's core business sits inside the direct-to-chip category. If the market consolidates around a different architecture, or if NVIDIA and its rivals push more of the cooling function onto the chip package itself, the specific technology Ecolab paid a hyperscaler multiple for could lose ground to a competing approach it does not own.

Ecolab is also not the only incumbent making this move. Water, chemical, and industrial services companies watching data center capital expenditure figures have the same math available to them. Whether Ecolab's head start on the NVIDIA relationship holds up depends on how fast competitors can either acquire their own cooling hardware maker or build the integration in house.

Beck is selling Wall Street on more than thermal engineering

A full-page advertisement running in the Wall Street Journal this week, weeks after the deal closed, makes the same case to a different audience: capacity alone will not be enough, and efficiency will matter just as much. That is investor messaging aimed at a specific worry, not deal advocacy. Data center water and power consumption is now the subject of state disclosure law in California, Michigan, and Iowa (Adam Silva Consulting, 2026), and hyperscalers building in water-stressed regions face community pushback that can stall a project longer than any supply chain issue.

Ecolab is positioning the CoolIT acquisition as the answer to a compliance and reputation problem, not only a thermal one. That framing extends the shelf life of the deal beyond the current cooling architecture debate. Regulation does not disappear if immersion cooling replaces direct-to-chip. It attaches to whichever company owns the water and energy reporting relationship with the operator.

About Shashi Bellamkonda

Loading bio…

LinkedIn Profile

Get new posts by email:

Subscribe

Powered by