Data centers are among the fastest-growing water consumers in the world. AI workloads and high-density chips are pushing cooling systems far beyond their original design limits, and most facilities still rely on evaporative cooling, which can use millions of gallons of water a year and create serious environmental pressure in water-stressed regions.
A new generation of startups is changing that. Using liquid cooling, microfluidics, modular design and closed-loop thermal systems, they remove heat without evaporating water while meeting the performance needs of modern compute. Investors have noticed: the five companies below have raised more than $18 billion between them, including project finance, according to Net Zero Insights data.
Sources: Net Zero Insights, Net0 platform; Axios; Crusoe.
Here are five startups reducing data center water consumption with next-generation cooling, from chip-level microfluidics to full-scale AI campuses.
Corintis (Switzerland)
Corintis advances chip-level cooling for data centers and high-performance computing. Founded by Remco van Erp and Sam Harrison, the company develops microfluidic systems that route liquid coolant through tiny channels engineered directly into or onto the chip, removing heat right where it is generated.
This tackles a core constraint in semiconductor progress: rising heat density now limits chip performance and drives excess energy and water use across data centers. By integrating cooling into the chip architecture, Corintis lets hardware run reliably at lower temperatures without water-hungry cooling towers.

Corintis also builds cold plates through additive manufacturing, with channels as narrow as 70 micrometres that target hot spots precisely, using bio-inspired designs from its own simulation tools. In trials with Microsoft, the approach removed heat up to three times more effectively than existing solutions, and Corintis reports chip temperatures more than 80 percent lower than with traditional air cooling.
Corintis raised a $24 million Series A in September 2025, then a further $25 million led by data center developer Applied Digital in December 2025. In 2026 it joined the first Project Switzerland scale-up cohort.
Flexnode (USA)
Flexnode builds prefabricated, liquid-cooled modular data centers for high-density workloads and rapid deployment. It minimises water use with a manifold microchannel heatsink, a hybrid immersion architecture and an additively manufactured dry-cooling heat exchanger, so its units can run in extreme environments without water-intensive cooling.
Flexnode combines power, cooling, compute and enclosure in one vertically integrated hardware stack, which cuts cost and deployment time. Digital-twin and analytics tools support real-time monitoring, predictive maintenance and performance optimisation.
The company is collaborating with the US Department of Energy’s ARPA-E COOLERCHIPS programme on next-generation cooling, backed by around $3.5 million in DOE grants. It raised a $9 million Seed round in February 2024 from investors including Yes VC, Zacua Ventures, JE Dunn Construction and Arup.
Submer (Spain)
Submer, founded by Daniel Pope and Pol Valls, develops immersion-cooling systems that submerge servers in a non-conductive liquid. This removes the need for traditional air-cooling infrastructure, cuts energy use, supports higher-density compute and reduces water consumption, with compact modular deployments for AI, high-performance computing and edge sites.

Submer has become one of the leading players in immersion cooling through rapid scaling and continuous R&D. Its $55.5 million Series C in October 2024, backed by Planet First Partners, M&G and Norrsken VC, valued the company at $500 million, and Banco Santander added $20 million in growth debt in December 2024.
Crusoe (USA)
Crusoe builds large-scale AI data centers with an energy-first strategy, pairing high-density compute with efficient liquid cooling and low-carbon power. It uses closed-loop, direct-to-chip cooling to keep GPUs at peak efficiency through extreme temperature swings, avoiding the water lost in evaporative cooling towers.

At its 1.2 GW Abilene campus in West Texas, cooling is fully closed-loop and non-evaporative: heat is rejected through air-cooled chillers and the same water recirculates rather than being consumed. Annual water use is expected to be just 12,625 gallons per building, roughly a tenth of a typical US household’s. The site integrates wind power to reduce emissions.
Crusoe chose direct-to-chip over immersion for easier servicing, a lower water footprint and modular maintenance at hyperscale, which supports fast hardware refreshes for large AI training clusters. It raised a $1.4 billion Series E in October 2025 at a valuation above $10 billion, followed by a $3 billion Series F in September 2026 with Atreides Management and Mubadala.
Firmus Technologies (Australia)
Firmus Technologies is a vertically integrated developer and operator of AI infrastructure, with a mission to build the most energy-efficient data centers. It began designing systems for 1,000+ GPUs in 2019 and developed its AI Factory platform for the energy and cooling demands of large-scale AI.
Its edge is liquid-cooling efficiency. Standard 19-inch rack servers are immersed in a synthetic, non-conductive fluid inside purpose-built racks, removing heat far more efficiently than air and without evaporative water loss. Newer sites combine immersion with direct-to-chip cooling and density-optimised facility design, all run on the company’s FactoryOS software.

Firmus raised a $327 million Series B in November 2025 to scale Project Southgate, developed with CDC Data Centres and NVIDIA, to 1.6 GW by 2028. In February 2026 it secured a $10 billion financing facility led by Blackstone and Coatue, and in April 2026 a $505 million growth round with NVIDIA valued the company at $5.5 billion.
Scaling innovation to reduce data center water consumption
For digital infrastructure to grow sustainably, data centers need cooling that can handle extreme heat loads without draining local water supplies. The startups above show the range of answers, from cooling inside the chip to entire campuses designed around closed loops.
Capital is following. Funding for these five companies has grown rapidly since late 2025, led by multi-billion-dollar rounds for AI infrastructure builders that treat water efficiency as a design requirement. Sustained investment in data center cooling and further technical breakthroughs will decide how quickly the rest of the industry follows. For the wider picture, see the environmental cost of data centers.