They Cheered the Water
Data centers don’t have to be win-lose for the community and nature.
It’s a warm evening in Mount Pleasant, Wisconsin, and you’ve brought a glass of wine out to the back deck the way you have for years. The grass, the fence, the slow blue dark coming down. And underneath all of it, something new. A hum. The kind you feel in your chest. A low, constant drone, loud enough to carry over a television and conversations around the fire pit. Some nights you give up on the backyard altogether. You’re considering moving. Three miles away, a 315-acre campus is running 24/7. The hum is the sound of it exhaling.
Your backyard is where this story really starts, but not how it was told this week at Microsoft Build.
Satya Nadella shared the great news during his keynote: the new AI data center design, Fairwater, would use about as much water in a year as one neighborhood restaurant. Fill the cooling loop once, recycle the same water continuously, barely a top-off needed from that point on. The room cheered. And they were right to — the water win is real.
The hum is real too, though. And it’s only going to grow worse as the site grows bigger.
Unfortunately, Microsoft closed one loop and left others wide open. That hum is heat being thrown away. The same heat blowing over your fence in Wisconsin is the kind Microsoft captures elsewhere and pipes straight into people’s homes. In Espoo, Finland, recovered data-center heat warms a quarter of a million people. So picture what leaves the building here, every night: efficiency Microsoft could have banked, warmth and quiet you’re owed, and energy made at real cost to the planet doing one job instead of two. Three kinds of value, drifting off into the cold.
Quiet doesn’t scale. Reuse does.
Low-frequency noise is very good at moving around and through walls, and the only way to really kill it is to make less of it. In a data center, that leaves two levers: quieter fans and capturing the waste heat instead of exhausting it.
There’s a whole field dedicated to quieter fans, much of it borrowed straight from animals that have moved through air and water silently for millions of years — the serrated edge of an owl’s wing, the ribbed flipper of a humpback, the grooved skin of a shark. None of it is exotic. All of it sitting on the shelf today, cheap to design in before the steel goes up and a costly afterthought once it’s standing.
But quieter fans only help so much. Stack enough of them to cool 3 GWs of compute — Fairwater’s projected size — and the noise climbs right back.
Heat capture is the better first lever, because it does two things at once. It cuts the drone — less heat to shed, less air through fans, less noise made — and it creates value instead of destroying it. Closing the water loop makes closing the heat loop easier than ever. Water-to-water heat exchangers have been standard equipment since the early 1900s, ubiquitous in power generation, chemical processing, and modern district energy.
Microsoft already runs the largest data-center heat-recovery scheme in the world. The whole industry knows the move. This isn’t a technology challenge. It’s a business-model challenge.
Build with the place, not around it
The water achievement is something operators can claim alone, from a stage. Reusing the heat asks for more — a partner, a contract, a real conversation with the place you’ve landed in. Which is the same conversation neighbors keep saying never happened. Because it was optional, it waited. And collaboration that stays optional is collaboration that doesn’t happen.
No single user can absorb 3GW of heat, so reuse at this scale becomes a coalition — a paper mill, a year-round food line, a greenhouse sized to the stream. Done well, it reclaims a real share of the heat and takes that much load off the fans. Not silent and solved. But a stream of value sent back into the place instead of thrown over our fences.
Timing is the rest of the challenge. If a fraction of the load is already emptying backyards, the full build — more power, no heat capture easing the fans — carries further. Low-frequency sound is notoriously hard to block, so the ring of homes that can hear it will widen as the site does. The same math runs the other way too though. Every MW added is more heat worth reusing, and one more reason to design the loop now, while the site and the protests are relatively small.
External pressure creates internal permission
I’m going to say it until it gets boring — external pressure creates internal permission.
When Satya listed what a new data center build has to protect — electricity prices, jobs, the tax base — the list was real and worth honoring. But look at what it leaves off. The reality felt in our backyards. The heat loop hasn’t been closed in Wisconsin for exactly one reason: nothing and no one required it. I’ve asked people I know in the industry, and plenty of them want to build it.
Known solution. Real business case. Real people who’d say yes.
The water, the heat, the noise — three faces of the same thing, and not the only ones. There’s the light washing the stars from the sky, the years of heavy construction rolling through town, the diesel generators that outlast every climate pledge they contradict. Each one has options. Each is cheapest to address with the community and nature in the room on day one. That’s the whole of it. Bring every dimension to the table before the design sets, and build with the place instead of around it.
Cheer the step. Then ask for the rest.
The Mount Pleasant campus is mostly still unbuilt. The heat is right there. The fix that quiets the fans is the same one that reuses the heat. And the water already proved Microsoft can close a loop the moment they decide it matters.
So cheer the water. Then ask for the rest — the heat reused, the fans quieted, the neighbors counted in, not sacrificed. Not a different project. Not a trade-off. The same instinct carried all the way through, so the value keeps circulating through the system instead of leaking out of it. The company, the community, and the nature that supports us all will come out stronger.
The first step deserves a cheer. It just isn’t the place to stop.
Connections to The Insider’s Guide to Innovation at Microsoft:
Pattern #4: Innovating More Than Technology — The piece’s spine: “not a technology challenge, a business model challenge.” Every component already exists — heat exchangers since the early 1900s, quieter-fan designs sitting on the shelf. History is clear: the decisive innovation is rarely the device but the business model and value chain around it.
Pattern #3: Innovating With Everyone — “Build with the place, not around it.” Bring community and nature to the table before the design sets. The Regenerative Datacenter team’s own precedent in the book — 200 conversations inside and outside before a single concept was drawn — is the proof that the early, optional conversation is the one that changes the outcome.
Aim for Positive / Regenerative Design — Closing the loop to send a stream of value back into the place instead of throwing waste heat over the fence: surplus by design, not harm reduction. The book quotes Satya asking whether, at the core of the business model, “are you creating a surplus around you?” This piece asks him to finish the question.
Top-down, Bottom-up, Outside-in — All three are in the room and none are wired together: top-down names what to protect (electricity, jobs, tax base); bottom-up is the engineers who already want to build the heat loop; outside-in is the backyard neighbors going unheard. Durable change needs the three connected — which is exactly what “external pressure creates internal permission” names: the outside-in voice unlocking what the bottom-up already wants to build.
This piece was developed in the open, with AI as a thinking and drafting partner. The argument, the judgment, and the final words are mine. Say it Ugly, Build it Better! Onward!