St. Paul-based Ecolab sees gold in surge of new AI data centers

As artificial intelligence fuels a global surge in computing power, the sprawling data centers behind it are scaling up faster, running hotter and packing in more machines than ever.

The explosive growth, in Minnesota and around the world, is providing big opportunities for companies including 3M, nVent and Xcel Energy. Ecolab is reorganizing its business units as it makes a huge bet on the industry’s growth.

In Minnesota alone, at least 11 large-scale projects are in various stages of planning or development, including Meta’s $800 million facility in Rosemount, set to be completed next year.

These aren’t traditional server farms; they are mega centers designed specifically for AI workloads, which demand far more electricity and produce far more heat than conventional computing.

“The data center industry is such a gold rush right now,” said Thom Jackson, a mechanical engineer for Dunham Associates who designs cooling systems for data centers.

Amid that rush, Ecolab sees an opening. The St. Paul-based Fortune 500 company, known for industrial cleaning and water treatment, is betting that its expertise in sustainable water use and advanced cooling technologies will become essential infrastructure for AI.

This year, Ecolab renamed its Global Industrial segment “Global Water,” reflecting the company’s growing focus on high-tech water solutions for industries such as chipmaking. The segment brought in $1.8 billion in sales in just the first quarter of 2025.

Ecolab Chairman and CEO Christophe Beck told McKinsey in May that the company is investing in AI data centers because it estimates that, by 2030, AI will consume as much power as the entire country of India and as much water as the U.S. population drinks.

Ecolab has since shifted much of its research and development toward data centers and semiconductor manufacturing, Beck said, “because we will do it in a way that uses less energy and water.”

Ecolab is rolling out technologies designed specifically for AI workloads. One of its newest offerings is the 3D Trasar technology for direct-to-chip liquid cooling, which uses metal plates mounted directly to computer processors to capture heat and transfer it to a closed-loop liquid cooling system.

The liquid is then cooled and recirculated, and the system reduces water use by 15% and significantly cuts energy consumption, Magnuson said. It also uses AI to monitor the liquid coolant’s temperature, pH and flow rate in real time.

“AI can be key to solving some of the very challenges that it creates,” Magnuson said in a press release.

But Ecolab’s pitch isn’t just environmental. It’s economic. Cooling accounts for a significant portion of both water and electricity use in data centers. Reducing those inputs means cutting costs.

“Nobody wants to pay a high energy bill,” said Jackson, who designs cooling systems for data centers.

In Minnesota, these economic and environmental concerns are now reflected in a new state law requiring companies proposing data centers that will use more than 100 million gallons of water per year to obtain approval from the state, demonstrating they have both an adequate water supply and have considered ways to conserve it.

“When I think about a data center, and if somebody says, ‘Well, it’s sustainable’ — what’s the definition there? What’s the metric?” said Jason Moeckel, assistant division director for the Minnesota Department of Natural Resources (DNR).

Direct-to-chip liquid cooling, like Ecolab’s tech, has become the mainstream choice for AI-focused data centers, Jackson said, because it pulls heat straight from the processor efficiently.

Some, such as the Meta site in Rosemount, plan to use closed-loop dry cooling systems that minimize water use. A Meta spokesperson declined to say whether Ecolab technology would be used at the site.

In Pine Island, near Rochester, the construction firm Ryan Cos. has proposed a large data center that would use air-cooling rather than water, specifically to minimize water use.

Ecolab’s cooling tools are already embedded in the industry. Its 3D Trasar technology has been used for years in traditional data center cooling, helping customers save more than 830 million gallons of water in 2023 alone, said Katie Ericsson Wanek, a senior communications manager at Ecolab.

But what’s really driving the spike in water and energy consumption among AI data centers isn’t necessarily the cooling method, Jackson said. It comes down to size.

“The perception is that all data centers use a lot of water,” he said. ”But the scale of these buildings is getting so much larger that total water usage is increasing.”

Moeckel said the Minnesota DNR will notify developers about local water constraints before they apply for water-use permits.