Vancouver-based startup Wafr Technologies has raised $100 million (USD) from a group of private investors to scale its water-efficient cooling technology for artificial intelligence data centres. Founded in 2025, the company says its proprietary thermal battery system can cut AI data centre water consumption by up to 95 percent and reduce cooling power use by up to 80 percent — a significant proposition at a time when the environmental cost of AI infrastructure has become a major point of contention for regulators, utilities, and communities hosting large data centre campuses.
What Happened
Wafr announced the $100 million financing round on July 3, 2026, and confirmed it will use the funds to launch an AI research lab and commercial data centre in Canada, while advancing ongoing commercialisation discussions with international partners. The company is already in negotiations with firms involved in constructing AI data centres in the United States and Europe, with Germany cited as a priority market alongside the US.
The $100 million represents the first tranche of a broader $300 million fundraising campaign. Wafr says it intends to raise the remaining $200 million “as soon as possible” from a combination of government funds and additional private investors. The company has announced letters of intent from international data centre construction partners, though it has not named specific customers or investors publicly.
At the core of Wafr’s technology is a thermal battery that captures cooling capacity when grid electricity is cheap — typically overnight or during periods of low renewable generation surplus — and releases it as needed during peak operational hours. This load-shifting approach reduces both water evaporation (a major consumption driver in conventional evaporative cooling towers) and peak electricity demand, two of the most pressing sustainability pain points for hyperscale AI data centres.
Why It Matters
The timing of Wafr’s funding announcement reflects a rapidly escalating crisis in AI infrastructure sustainability. Google’s AI expansion recently drove a record 37 percent spike in the company’s power consumption, drawing sharp criticism from environmental groups and sparking regulatory inquiries in several European jurisdictions. The water picture is equally stark: a single large-scale AI training cluster can consume millions of gallons of water per month for cooling, straining municipal water systems in arid regions where data centres tend to cluster due to land costs and tax incentives.
A 95 percent reduction in water use — if Wafr can demonstrate it at commercial scale — would be transformational for the industry. It would substantially reduce the permitting friction that has slowed data centre construction in water-stressed markets like the US Southwest, the Middle East, and southern Europe, opening up geographies that are otherwise off-limits for large cooling-intensive deployments.
Wafr is entering the market at a moment when cooling technology is attracting serious venture capital. The company faces competition from established liquid cooling specialists, immersion cooling startups, and major HVAC players who are all trying to capture share in what analysts project will be a multi-billion-dollar cooling infrastructure upgrade cycle driven by AI compute density.
Background and Context
Data centre cooling accounts for roughly 30 to 40 percent of total facility energy consumption in conventional air-cooled facilities, and the problem is intensifying as AI chips run hotter than the general-purpose server hardware they are displacing. Nvidia’s latest GPU accelerators and the custom inference chips that AI labs are increasingly developing in-house all generate substantially more heat per rack than previous generation equipment, requiring facilities to fundamentally rethink their thermal management approaches.
Microsoft’s $2.5 billion Frontier Company initiative has made AI infrastructure deployment at scale a central industry challenge, and the bottleneck is increasingly physical — power grid connections, water availability, and cooling capacity — rather than software or model capability. Wafr’s technology, if it scales as claimed, would remove one of the three key physical constraints limiting where and how fast AI data centres can be built.
What Comes Next
Wafr’s next milestone will be the construction and commissioning of its first commercial data centre in Canada, which the company says will also serve as a proof-of-concept facility for potential enterprise customers. Success there will be critical for closing the remaining $200 million in financing and attracting the named anchor customers it will need to pursue its US and European expansion plans.
The broader question is whether Wafr can move fast enough. AI data centre construction timelines are compressing as hyperscalers race to provision compute for the next generation of AI model training. A cooling technology that requires a purpose-built thermal battery installation will face competition from faster-to-deploy alternatives, including modular liquid cooling systems that can be retrofitted into existing facilities. How Wafr positions its technology for both new builds and retrofits will likely determine how large a share of the market it can capture.
