Ecolab Water-Neutral Data Centers Face a Bigger Test Than Cooling Chips

Ecolab says its $4.75 billion CoolIT Systems acquisition gives AI data centers a path toward water-neutral operations, despite growing pressure on local supplies. The deal adds direct-to-chip liquid cooling to Ecolab’s water treatment, monitoring, and semiconductor businesses. It also turns a century-old industrial supplier into a more central player in the AI infrastructure race.

The phrase “water-neutral” carries the tension. Closed-loop cooling can sharply reduce water consumed inside a facility, but a data center’s total footprint extends beyond its server room. Electricity generation, chip manufacturing, local climate, and water replenishment all affect the final calculation.

Microsoft, Google, Amazon, and other hyperscalers are pursuing their own low-water designs. Ecolab must therefore prove more than technical compatibility. It must show that an integrated cooling platform can reduce real water consumption across varied sites without shifting the burden to electricity systems or nearby communities.

Ecolab’s CoolIT Deal Moves It Directly Into the Server Rack

Ecolab is no longer positioning itself only as the company that treats water after infrastructure decisions have been made.

The company completed its acquisition of Calgary-based CoolIT Systems in July 2026. The approximately $4.75 billion transaction followed an agreement announced in March and closed earlier than Ecolab had initially expected.

CoolIT specializes in direct liquid cooling for high-density computing. Its cold plates sit on heat-producing components such as GPUs and CPUs. Coolant passes through the plates, absorbs heat, and carries that heat toward a coolant distribution unit.

That architecture brings fluid much closer to the source than conventional room-level air cooling. It is increasingly relevant because AI accelerators concentrate far more heat within each rack than many earlier enterprise servers.

Ecolab already supplied cooling-water chemistry, monitoring systems, and operational services. Its 3D TRASAR platform monitors factors such as temperature, coolant concentration, pH, and flow. CoolIT adds the physical equipment needed to move coolant through high-density computing systems.

The combination is the basis for Ecolab’s water-neutral data centers claim. CEO Christophe Beck told Fortune that the company now has a solution capable of making facilities water-neutral. He connected that argument to closed-loop cooling, which recirculates coolant instead of continually evaporating water.

The company’s broader spending also matters. Beck said Ecolab had committed about $7 billion over six months, including the CoolIT transaction and related investments. The strategy covers several parts of AI infrastructure rather than one isolated cooling product.

Ecolab bought Ovivo Electronics in December 2025 for approximately $1.6 billion. Ovivo supplies ultrapure water systems used in semiconductor manufacturing. With CoolIT, Ecolab now participates in chip production, facility cooling, coolant management, and some water-intensive parts of power generation.

The company says its Global High Tech operation can reach $4 billion in annual sales by 2030. That target gives the acquisition a clear commercial purpose. AI infrastructure is becoming both a growth market and a water-management problem.

Ecolab plans to introduce an integrated cooling platform at the Supercomputing conference in Chicago in November 2026. According to the company’s CoolIT acquisition announcement, the system will combine cold plates, coolant distribution units, cooling fluids, and digital optimization.

That launch will provide the first major opportunity to examine whether the acquisition has produced a unified system rather than a collection of related products.

Why AI Data Centers Have Turned Water Into a Constraint

The immediate pressure comes from computing density, but the political pressure comes from where water is consumed.

Training and serving AI models require large clusters of accelerators. Each chip converts much of the electricity it receives into heat, and tightly packed racks concentrate that heat within a small physical area.

Operators must remove the heat continuously. A cooling failure can reduce computing performance, shorten equipment life, or force servers offline. That makes cooling part of the computing system rather than a secondary building service.

The growth curve is steep. A congressionally requested federal energy report estimated that U.S. data centers consumed 176 terawatt-hours of electricity in 2023. Its scenarios place 2028 demand between 325 and 580 terawatt-hours.

The same research estimated that U.S. facilities directly consumed roughly 17 billion gallons of water during 2023. That direct figure excludes much of the water associated with electricity generation, equipment manufacturing, and construction.

National totals can also hide local stress. Water is drawn from particular utilities, watersheds, and aquifers. A facility that appears modest within national statistics can still compete with households, farms, and existing businesses during a dry season.

This local concentration explains growing resistance to data center projects. Developers often describe annual water demand, while utilities must plan around peak daily withdrawals. Communities also want to know whether facilities will use drinking water, treated wastewater, or another source.

The accounting vocabulary complicates the debate. Water withdrawal measures the amount taken from a source. Water consumption generally describes water that does not return promptly to the same watershed, often because it evaporates or becomes incorporated into another process.

A cooling system can withdraw water, return part of it, and consume another part. It can also eliminate on-site evaporation while increasing electricity demand. If the extra power comes from a water-intensive thermal plant, some consumption moves beyond the data center boundary.

That distinction matters for Ecolab. Its technology can reduce losses inside a cooling loop, but no cooling product controls every part of the facility’s footprint.

Ecolab argues that water-efficient design must consider the entire site. That includes the server loop, heat-rejection equipment, water chemistry, operating temperature, climate, and power source. This system-level framing is stronger than treating one water-use metric as the whole answer.

It also creates a higher standard of proof. Ecolab’s water-neutral data centers cannot be evaluated only by checking whether a closed pipe leaks. Operators need consistent boundaries, site-level measurements, and credible comparisons against the systems being replaced.

How Ecolab Liquid Cooling Changes the Water Equation

The key mechanism is recirculation: coolant captures heat at the chip and remains inside a managed loop.

In direct-to-chip cooling, a cold plate makes thermal contact with a processor or another hot component. Liquid flowing through internal channels absorbs the heat. Pumps then carry the warmed coolant away from the server.

A coolant distribution unit separates the server-side loop from the facility-side cooling system. It controls flow, pressure, temperature, and heat exchange while protecting sensitive computing hardware from unsuitable facility water.

This arrangement explains how water-neutral data centers work at the server level. The liquid does not need to be discarded after each pass. Once the loop is filled, the same coolant can circulate repeatedly, apart from maintenance losses or component replacement.

Liquid also transfers heat more effectively than air. That efficiency can support denser racks and allow heat to leave the chip at warmer fluid temperatures. Higher operating temperatures can widen the range of climates where operators reject heat without evaporating water.

However, direct liquid cooling does not determine the final destination of the heat. A facility might transfer it to a dry cooler, which uses fans and outside air. It might use a cooling tower that consumes water through evaporation. A hybrid system might switch between those methods as weather and computing demand change.

The tradeoff is familiar. Dry cooling reduces direct water consumption but can require more electricity, particularly during hot conditions. Evaporative cooling uses water to remove heat efficiently and can lower electrical demand.

Ecolab’s proposed advantage is integration. CoolIT supplies cold plates and distribution hardware. Ecolab contributes coolant chemistry, monitoring, water treatment, and field service across the rest of the cooling chain.

Its 3D TRASAR technology monitors fluid conditions that can affect corrosion, biological growth, scaling, and thermal performance. Those measurements matter because small changes inside a cooling system can threaten expensive servers or reduce heat-transfer efficiency.

Monitoring also helps distinguish a laboratory design from a maintainable commercial system. Large data centers operate continuously, often with thousands of fluid connections. Operators need leak detection, water-quality controls, maintenance procedures, replacement fluids, and clear responsibility when performance moves outside specifications.

Ecolab began building this position before buying CoolIT. In May 2025, it introduced monitoring specifically for direct-to-chip cooling. The company said the system tracks coolant concentration, temperature, pH, flow rates, and other indicators in real time.

The acquisition moves Ecolab deeper into the equipment layer. Instead of optimizing water around hardware selected by another vendor, it can help specify the cold plates, coolant units, fluids, and controls together.

That tighter integration can reduce design conflicts. It can also give Ecolab access to operational data across the entire cooling path. The value of that data will depend on whether customers permit useful aggregation and whether reported savings use consistent methods.

Ecolab liquid cooling therefore offers a credible mechanism for cutting on-site water use. It does not, by itself, establish that an entire campus has become neutral.

The Main Conflict Is a Clear Promise Against a Complicated Footprint

“Water-neutral” sounds like a physical outcome, but it remains an accounting claim until Ecolab defines its boundaries.

Ecolab’s public language has varied. Beck told Fortune that the company can make data centers water-neutral. A July article from the company described movement toward a “near-zero water footprint.” Those phrases are related, but they do not necessarily mean the same thing.

Near-zero water use usually suggests that a facility consumes very little water. Water neutrality can mean that remaining consumption is balanced through replenishment, restoration, or conservation elsewhere in the same watershed.

A facility might therefore be called neutral while continuing to consume water locally. The credibility of that label depends on timing, location, additionality, and measurement.

Timing asks whether replenishment occurs during the same period as consumption. Location asks whether benefits reach the affected watershed. Additionality asks whether the project created savings that would not otherwise have happened.

The physical boundary creates another question. Does Ecolab’s claim cover only cooling water within the building? Does it include water consumed by power generation? Does it include ultrapure water used to manufacture the chips?

Fortune’s original water-neutral claim focused on closed-loop technology that captures chip heat without spraying or evaporating water. That supports a narrow claim about direct cooling losses.

It does not automatically cover a full AI workload. A model depends on processors, electricity, backup generation, network equipment, and buildings. Each stage carries a different water footprint.

Power is the largest complication. A facility using dry cooling might report little direct consumption while drawing electricity from plants that use water for steam generation or heat rejection. Another site might consume more water directly but use less electricity overall.

Water quality also matters. Drinking-quality water carries a different community impact from treated wastewater that has few competing uses. A single gallon figure cannot express that difference.

Climate further limits universal comparisons. Dry cooling performs differently in Minnesota, Arizona, and coastal Virginia. The best balance can change by season and even by hour.

These caveats do not make the technology unimportant. They show why the claim needs site-level evidence. Ecolab should report direct withdrawal, direct consumption, water source, energy changes, and any indirect estimates separately.

The company should also identify the baseline. Savings against an older evaporative system will look different from savings against a newer air-cooled design.

Without those disclosures, “water-neutral” risks becoming a broad label attached to a narrower cooling improvement. With them, the phrase can become a measurable operating target.

The distinction is particularly important for communities evaluating new projects. Residents do not experience a global average. They experience changes in their own utility capacity, summer demand, rates, and watershed.

Microsoft and Other Operators Are Chasing the Same Outcome

Ecolab’s strongest competition is not another water-treatment company; it is the industry’s shift toward low-water cooling as a standard design requirement.

Microsoft says its newest AI-focused facilities do not consume water for cooling during normal operations. Those sites circulate liquid at the chip and reject heat through air-cooled equipment outside the building.

The loop requires an initial fill. Microsoft says it then operates without routine cooling-water consumption. Existing facilities still use several designs, and some apply evaporative cooling during hotter periods.

Microsoft’s zero-water designs show that closed-loop cooling is not unique to Ecolab. They also show how quickly the competitive baseline is moving.

Google has pursued reclaimed water, climate-specific cooling methods, and watershed replenishment projects. Amazon has expanded the use of recycled water and adjusted server operating conditions. Meta has also reported facility water-efficiency metrics and replenishment commitments.

These companies control large portions of their infrastructure designs. They can develop cooling specifications internally, negotiate directly with equipment suppliers, and apply operational changes across extensive fleets.

Ecolab’s opportunity lies between those capabilities. Many colocation providers, enterprise operators, and infrastructure developers lack the same water-engineering resources. They may prefer an integrated supplier that combines equipment, fluids, controls, and service.

Digital Realty offers an early example. In October 2024, it announced an Ecolab-led water optimization pilot across 35 U.S. data centers. The system uses machine learning to identify inefficient cooling operations and recommend changes.

The partners projected water-use reductions of up to 15 percent and avoidance of up to 126 million gallons of potable-water withdrawal each year. They explicitly cautioned that these were company-specific projections and that actual results could vary.

That qualification is important. The 35-site pilot evaluates operational optimization, not the complete CoolIT platform Ecolab later acquired.

Still, the project gives Ecolab relationships, operating data, and experience across multiple facilities. It also creates a potential route from incremental water savings to larger cooling-system changes.

The company must now demonstrate that its combined platform beats separate components on outcomes that operators care about. Those include uptime, energy demand, maintenance, deployment speed, water consumption, and compatibility with next-generation accelerators.

Price will matter, even if technical reporting avoids specific commercial terms. Operators will compare a full-service platform with in-house engineering, alternative cold-plate suppliers, immersion cooling, and conventional hybrid designs.

This competitive setting makes Ecolab’s claim more consequential. If the system works across many climates and facility types, it can turn water performance into a purchasable infrastructure capability. If results remain limited to favorable sites, the market will treat it as one option among several.

What Must Be Proven Before Water-Neutral Becomes a Standard

The next stage is about measured operations, not another promise about cooling architecture.

The first signal will arrive at Supercomputing in November 2026. Ecolab says it will present its integrated 3D TRASAR platform with CoolIT cold plates, coolant distribution units, advanced fluids, and digital controls.

That presentation should reveal how tightly the pieces work together. Buyers need details about supported rack densities, heat-rejection options, monitoring boundaries, maintenance responsibilities, and compatibility with server platforms.

A detailed product launch would strengthen Ecolab’s argument that the acquisition created an end-to-end platform. A general demonstration without operating data would leave the central claim unresolved.

The second signal will be customer evidence. Ecolab needs measured results from operational sites using the combined equipment, not only modeled savings or pilots of separate monitoring software.

Useful reporting would include annual and peak water withdrawal, consumption, source type, cooling energy, equipment utilization, weather conditions, and downtime. Results should compare like-for-like workloads or explain adjustments clearly.

Evidence from several climates would matter more than one favorable installation. It would show whether Ecolab liquid cooling can maintain low water use during summer heat, when utilities face their greatest stress.

The third signal will be the definition of neutrality. Ecolab should publish a repeatable methodology that separates avoided consumption from replenishment and distinguishes direct cooling water from indirect use.

That framework should address electricity generation, even if those numbers remain estimates. It should also explain whether semiconductor manufacturing belongs inside the claim when Ecolab describes support across the AI value chain.

Clear accounting would strengthen the company’s position even when totals are not zero. A narrow but measurable statement is more useful than a broad promise with unclear boundaries.

The pressure to provide that evidence will only grow. AI infrastructure development increasingly depends on permits, utility agreements, and community support. Water performance can influence whether a site advances, how quickly it connects, and what operating restrictions it faces.

Ecolab has assembled a plausible technical stack for reducing cooling-water consumption. It has also attached that stack to one of the hardest environmental promises in the data center market.

The outcome will affect more than one supplier. Verified results could give operators a practical model for combining high-density computing with local water limits. Weak disclosure would reinforce skepticism that neutrality language hides transferred or offset consumption.

For developers, enterprise buyers, and AI users, the relevant question is no longer whether liquid can cool a chip. That has been established. The question is whether Ecolab water-neutral data centers can report lower total water pressure without trading it for excessive energy demand.

Watch the November platform launch, the first multi-site operating results, and the accounting rules behind the word “neutral.” Those three signals will show whether Ecolab has built a scalable answer or simply given a narrower efficiency gain a larger name.