How a UW-Madison Startup Is Rethinking Data Center Cooling

We talk about "the cloud" like it's weightless. Like our data lives somewhere clean and abstract, floating above us in the digital ether.

It doesn't.

The cloud has a body. It lives in massive buildings filled with millions of servers running 24/7. Those servers generate enormous heat. And right now, the primary way the industry keeps them from overheating is by evaporating staggering amounts of clean water through giant cooling towers on the roof. A single large data center can drink five million gallons of water per day. That's roughly the daily usage of a town of 50,000 people.

As AI demand accelerates, these facilities are getting bigger, hotter, and thirstier. And communities are starting to push back. Projects are getting blocked over water rights. Local governments are rejecting permits. The industry is running headfirst into the physical limits of its own infrastructure.

Three UW-Madison students think the answer is right beneath our feet.

‍

Flipping the Script on Cooling

Substrata is a deep-tech startup building what its founders call "the thermal intelligence layer for the AI economy." Instead of fighting heat with more water and more hardware, Substrata uses subsurface physics to transfer data center heat directly into the ground beneath the facility. The earth, which maintains a constant cool temperature year-round, acts as a massive thermal sponge. No water evaporated. No water wasted.

Mara Zwicker is one of the three co-founders. She studies GIS, data science, and statistics at UW-Madison. Her co-founders are Trinity Krohn, a CS and data science student who founded Women in Entrepreneurship at UW, and Joshika Nachiappan, an electrical engineering student with experience at Meta and Milwaukee Tool.

Together, they cover the entire data center stack. Software, hardware, and earth.

"The cloud might be digital," Mara told me, "but the infrastructure that powers it is very much physical."

‍

From Soil Microbes to Thermal Physics

Mara's path to founding an AI infrastructure startup didn't follow a straight line.

Before Substrata, she interned at the Wisconsin Institute for Discovery in the Handelsman Lab, working with Tiny Earth on antibiotic resistance research. She was analyzing soil and microbial data under a microscope. It was her first exposure to how technical research could translate into real-world impact.

"It's funny to think about now," she said, "because I've essentially gone from looking at the earth's soil at the microscopic level to mapping its thermal capacity at the macro, infrastructural level with GIS."

That GIS training turns out to be the key to what makes Substrata different. A pure computer science background teaches you to build software that scales infinitely without worrying about physical constraints. Mara's training does the opposite. It grounds everything in spatial reality. Where things are. How the physical environment shapes what you can actually build.

"When you look at data centers through a GIS lens, you stop seeing a purely technical cooling problem and start seeing a spatial infrastructure opportunity," she said.

Substrata's role isn't just engineering a cooling system. It's providing the thermal intelligence, the spatial and physical modeling, to show developers exactly how much heat the ground at their specific site can safely absorb. It's a fundamentally different way of thinking about the problem.

‍

A Team Built by Design

The three founders didn't meet in a class or a lab. They met through gener8tor's gALPHA CSNest program, a free venture-creation workshop for student entrepreneurs.

Mara signed up on a whim. At the kickoff event, she pitched her idea first in front of the entire group. By the end of the program, Trinity and Joshika approached her directly.

The all-female founding team was intentional. Mara said that deep tech and AI infrastructure are notoriously male-dominated, and the three of them share an understanding of what that means and how to navigate it.

"But what really makes us a 'dream team' is how perfectly our skills interlock," she said. "We aren't just three coders. Trinity is a brilliant full-stack developer who brings incredible business acumen. Joshika is our electrical engineer who deeply understands hardware and power, and actually saw how massive facilities run firsthand during her time at Meta. And I bring the GIS architecture and data science. Between the three of us, we have the software, the hardware, and the earth completely covered."

‍

The Model Is Broken

The data center industry has a water problem, and it's only getting worse.

Projections suggest that water consumption by data centers could increase dramatically in the coming years as AI demand pushes facilities to run hotter and denser. Some communities are already drawing the line. Site permits are getting rejected. Local governments are choosing drinking water over server farms.

Mara frames it bluntly: "The industry is treating physical resources like they are infinite cloud compute. You can't just 'use more water' because municipalities don't have it to give."

The current cooling model relies on what Substrata calls "brute-forcing" the problem. Evaporate clean water through expensive hardware. When that's not enough, use more electricity to run bigger cooling systems. Both approaches hit walls. Water-intensive cooling drains local supplies. Energy-intensive cooling strains power grids and increases carbon footprints.

‍

"In five years, I want building a data center without considering subsurface physics to feel as outdated as building one without a fiber connection," Mara said.

‍

Substrata's approach sidesteps both. By pumping heat into a closed-loop system underground, the facility works with the specific physics of the land it sits on rather than extracting from it. The ground absorbs the heat naturally. The cooled fluid loops back to the servers. Nothing is lost.

The idea isn't purely theoretical. Ground-source thermal exchange has been used in other contexts for decades. What Substrata brings is the intelligence layer: modeling which sites can support it, how much thermal load the subsurface can handle, and how to integrate it into data center design from the beginning.

‍

Wisconsin Made the Founder

Mara is from San Francisco. She's planning to move back to the Bay Area after she graduates this spring. Substrata will likely scale from there.

But she couldn't wait until graduation to start building. And Madison turned out to be the perfect incubator.

"Because I am currently writing my senior thesis on this exact issue, my academic research and my startup venture go completely hand-in-hand," she said. "Building here meant I had direct, immediate access to incredibly supportive professors, world-class research facilities, and student founder organizations that helped me get this off the ground while still in school."

Her time at the Wisconsin Institute for Discovery shaped her mindset as a founder. Working under Jo Handelsman in an all-female research team planted the seed for how she wanted to build Substrata. The research showed her that academic work doesn't have to stay in a lab. It can be spun out to solve massive problems.

And the support infrastructure surprised her. Before starting Substrata, Mara said she had no idea how deep the resources ran for student founders in Madison. Programs like gALPHA and Transcend UW Madison provide advising at every level, from initial concept to legal structuring.

"More than anything, the ecosystem here proves that you don't have to wait," she said. "It is entirely possible to be a young entrepreneur and build something highly meaningful right now."

This is a different version of the brain drain conversation. Madison didn't keep Mara. But it made her a founder. The DNA of Substrata, its early prototypes, its technical foundation, and the team that built it will always trace back to UW-Madison. That's not a loss for Wisconsin. That's a proof point.

"San Francisco will be where Substrata scales," Mara said, "but I am incredibly proud that its DNA will always be deeply rooted in my education here at UW-Madison."

‍

Bigger Than One Startup

Substrata is early. The founders are transparent about that. They're still in the gALPHA accelerator, refining their models and conducting the customer discovery calls that will shape their first pilot. They haven't built a prototype yet.

But the timing of what they're working on is hard to ignore.

UW-Madison just approved a standalone College of Computing and Artificial Intelligence, the first new college at the university in 42 years. Data center water consumption has gone from an industry footnote to a front-page crisis. And the Midwest is increasingly in the conversation for new data center development, which means the water and energy tradeoffs are about to become very real for communities across the region.

"In places like the Midwest, where we have abundant land but sensitive water ecosystems, you can't just drop a traditional data center into a community and expect them to be okay with it draining millions of gallons of water a day," Mara said. "Substrata's approach makes these facilities good neighbors by default."

There's also a generational dimension to this story that's worth paying attention to. Mara didn't mince words about why she felt compelled to start building now.

"With the current administration bringing government support for climate science essentially to a halt, it became very clear that the public sector isn't going to save us," she said. "Young, educated people have to take the future into our own hands. If we have a deep understanding of technology and infrastructure, it feels like a moral imperative to use that knowledge to actually do something good for the world."

Whether Substrata becomes the company that changes how data centers are cooled or not, the fact that three women at a Midwest university are even taking a serious swing at a trillion-dollar infrastructure problem says something about what's possible here.

‍

What's Next

The next six to twelve months are about the transition from research to reality. As Mara wraps up her senior thesis and graduates, the focus shifts to taking the foundation built in Madison to the Bay Area to scale. The team is looking for early-adopter partners ready to move away from water-cooled systems.

"In five years, I want building a data center without considering subsurface physics to feel as outdated as building one without a fiber connection," Mara said.

It's a bold claim from a company that's still in its first months of existence. But if you've spent any time watching the data center industry collide with water scarcity, community resistance, and the sheer physics of cooling AI infrastructure, it's not hard to see why someone would want to solve this problem.

And it's awesome to see that, that someone just might come from a GIS program in Madison, Wisconsin.

‍

Jacob Miller is Marketing & Brand Manager at Headway and Marketing Director for Startup Wisconsin, as well as host of their podcast. As a Yooper based in Green Bay, Wisconsin, he enjoys all the Midwest has to offer.