Wafr says it raised $100 million to cool AI data centers with less water

A Vancouver company called Wafr Technologies says it has raised $100 million to build cooling systems that use far less water than the equipment keeping today's AI data centers from overheating. It announced the round on July 3. It will not say who put in the money.

That last part is where a technology reporter starts asking questions, so let me separate what is solid from what is a pitch.

The problem Wafr is aiming at is real

Cooling a large AI facility is, before anything else, a water problem. Racks of accelerators packed close together throw off a lot of heat, and a common way to move that heat outside is evaporative cooling, which works by boiling off water. The numbers are not small. A 100-megawatt AI data center in the United States goes through roughly 2 million litres of water a day, about what 6,500 households use, according to a 2025 International Energy Agency analysis. Multiply that across the build-out the hyperscalers are financing, and water stops being a footnote. Amazon, Microsoft, Alphabet and Meta are on track to spend more than $700 billion on AI infrastructure this year by Goldman Sachs' June estimate.

So a company selling a way to cut water and power at the cooling stage is aiming at a genuine bottleneck, not an invented one. Cooling has become a place where real money moves: in March, Ecolab bought Calgary's CoolIT Systems, which makes liquid-cooling gear, for $4.75 billion.

What Wafr says it built

The disclosed mechanism is a "thermal battery." Strip the branding and that is thermal energy storage: a material that soaks up and holds heat, or stores cooling capacity, so the timing of when you cool and when you pay for power can be pulled apart. Wafr's version, according to the company, lets a data center do its cooling work when grid electricity is cheap and off-peak, bank that capacity, and coast through the expensive peak hours. A diagram on its site shows exactly that load-shifting loop.

None of that is new physics. Buildings have been cooled with ice storage and chilled-water tanks for decades on the same principle. What Wafr is claiming is a passive package that, in co-founder Darrell Kopke's words, "virtually eliminates the need for water and stops evaporative water for cooling."

Then come the headline figures: up to 95 percent less water, and up to 80 percent less cooling power. Read those carefully. They are the company's own numbers, they carry an "up to," and the trade press covering the raise notes that few technical details of the system have actually been revealed. There is no published methodology and no stated baseline, so it is not clear 95 percent less water than what. Wafr has letters of intent with data-center builders, not a running site you can point a meter at. Until there is one, treat the percentages as a target the company is selling, not a result it has measured.

The parts that are still a pitch

The round itself is reported on Kopke's word to the Financial Post. Wafr declined to name a single investor, saying an announcement is "coming soon." A $100 million raise with no named backer is a claim, and it should be labeled as one until the cap table shows up.

The plan for the money is the part that reads oddly for a cooling vendor. Wafr, founded in March 2025, says it will use the $100 million to stand up its own AI research lab and its own data center, hire dozens of technical staff, and find sites across Canada. It also wants another $200 million "as soon as possible" from government and private money, for roughly $300 million total. That is a one-year-old company trying to be a cooling supplier, an infrastructure operator and an AI lab at the same time. Any one of those is a hard business. Kopke runs the Vancouver arm of the Creative Destruction Lab startup incubator; his co-founder, Bikram Singh, spent close to two decades in the cooling industry across the Middle East and India. The cooling pedigree is there. The rest of the ambition is unproven.

Wafr is going after the United States first, with Germany next. Kopke's framing is blunt: "The U.S. has 5,300 data centres. The country that has the next highest amount of declared data centres is Germany at 400." He also called it "the market that has the least sustainable AI," which is a sales line, but the water figures behind it are not.

The idea is sound and the water problem it targets is worth solving. The number that will matter is not the size of the raise or the percentage on the slide. It is the first operating data center that can show its water and power draw against a named baseline. Until that exists, this is a good pitch with money behind it that no one will name yet.