Before Chase Lochmiller ever built an AI-infrastructure company, he'd been obsessed with climbing mountains. His middle school teacher was married to Eric Weihenmayer, the first blind climber to reach Everest. And that story, combined with the Rockies outside his window in Colorado, set the future founder of Crusoe on a quest to conquer distant peaks.

His first attempt to climb Everest in 2014 was derailed by a catastrophic avalanche near the Khumbu Icefall. He was safely away in base camp but heard the tremendous crash of snow that would claim the lives of 16 sherpas.

Four years later he returned to Everest and successfully scaled the peak. But just as he descended back into communication range he got a call from his family telling him his mom had died.

Lochmiller had known going into it that she was ill—she'd had a heart transplant years earlier—but his father gave him his blessing to climb the mountain and then return home quickly. As it turned out, it wasn't quick enough.

"I came back from the summit and it was this high, high. And then you get back, and then you talk to your family, and they're like, hey, Mom passed away," Lochmiller said.

And so Lochmiller returned home to Colorado where he reconnected with a childhood friend, Cully Cavness. There, while hiking through the state's 14ers, they began putting together the pieces that would become Crusoe.

Years later, Crusoe would place itself in the middle of the AI revolution by building data centers where no one else thought to look. What started as an unconventional energy play would evolve into something far bigger: a bet that AI's insatiable appetite for computing power would require an entirely new approach to infrastructure. Lochmiller had been preparing for a mountain many others couldn't yet see.

On their hikes, Lochmiller's conversations with Cavness kept returning to the topic of flaring—the process of burning off natural gas that oil drilling produces. Wells yield oil, natural gas, and water, but while the water can be trucked back to a treatment facility, drilling companies in remote locations without pipeline infrastructure have no economical option for the natural gas other than setting it on fire.

The two came at the topic from different mindsets. Cavness had grown up in an oil and gas family but attended the environmentally progressive Middlebury College, and was embarrassed by the waste.

Lochmiller meanwhile had spent years in high-frequency trading as a quant, picking up a graduate degree in machine learning from Stanford and doing a stint at a crypto hedge fund started by Olaf Carlson-Wee, Coinbase's first employee.

Combining Cavness's energy problem with Lochmiller's computing and crypto background, the two envisioned a company that would use flared gas to generate power for cryptocurrency mining rigs. They initially called these remote facilities "Island Clouds"—envisioning an archipelago of small, isolated data centers. Later they rebranded it Crusoe Cloud, a nod to the fictional island adventurer Robinson Crusoe.At first it was just a name. Eventually it would become the business model.

Capturing the energy from the flare gas to mine bitcoin

Beyond Bitcoin

But as much as Lochmiller, then an ardent libertarian, believed in the importance of bitcoin, it was never his real target with Crusoe—AI was. During his master's in computer science, deep learning and neural networks were all the rage, and he recognized how much computing power it would take to process the massive datasets these models required.

In 2018 when the company was raising its first rounds of financing, investors weren't much interested in AI. Through a relationship with Stefan Cohen, now the partner leading Bain Capital Crypto, Bain Capital Ventures co-led the company's seed round with Founders Fund as Crusoe set out to scale its bitcoin mining business. The pitch was straightforward. BCV's investor memo from July 2018 described the company as "a mobile data center operator that has semi proprietary access to free electricity at remote shale drilling sites."

It continued: "initially they intend to focus their data center compute power on cryptocurrency mining. Over time we think there will be other use cases, including AI/ML model training via GPUs."

For BCV and likely Founders Fund, AI was a "call option"—upside if it hit, but not the reason they were writing the check.

“This idea of access to unused power at great terms was an important part of the thesis,” said Cohen. “And that was part of what he said on day one: I'm going to find stranded power. I'm going to use it in ways that people aren't using it, and I'm going to keep doing that.”

Generative AI as a mainstream technology was still in its nascent stages. Google had published its seminal transformers paper "Attention is all you Need" in 2017 but the subject was mostly a topic for academic conferences like NeurIPS.

When Crusoe made the rounds to raise its Series A, investors were puzzled by Lochmiller's continuing fixation on AI customers.

"You have this great bitcoin thing that's going really well with all these oil companies," Lochmiller recalled his early investors telling him. "Why don't you focus on that one thing?"

Even his early backers had doubts.

“When he started moving away from crypto, we were skeptical that it was the right time,” said Enrique Salem, partner at Bain Capital Ventures who led the initial investments into Crusoe. “But his intuition was ultimately spot on, and we trusted his ability to execute on complex projects like this one.”

There were some early signs of industry interest in AI data centers. In 2019, around the time Crusoe raised its Series A, Microsoft made its initial $1 billion investment in OpenAI. An executive on Microsoft's infrastructure team, who had previously worked at Exxon with Crusoe, reached out about a potential partnership to power Azure data centers with flared gas.

Lochmiller saw a bigger opportunity and suggested to the Microsoft executive that maybe it could provide compute to support OpenAI. Nothing happened at the time, but that conversation would resurface a few years later.

By 2021, the bitcoin mining business was humming along, but Crusoe's AI ambitions were gaining ground. In subsequent funding rounds, powering large clusters of GPUs moved from a footnote in the pitch deck to the headline.

As a capital-intensive infrastructure company with hard assets, Crusoe didn't fit the typical venture capital mold. But the company assembled an eclectic coalition: Founders Fund, which has taken on far-off moonshot bets like SpaceX; Ribbit Capital, focused largely on fintech investments; and Bain Capital, more accustomed to asset-heavy businesses. Even then, it was a stretch.

"Only making consensus bets doesn't make you a lot of money. You have to sometimes say, this is a very special founder who has a thesis and has ambition," said Salem. "We took a swing on Chase and Cully early with this vision around energy and power dislocations in the country. Over time, we got much more comfortable with the capex intensity. But I think it's very hard for other venture investors to digest that model."

Cully Cavness, Enrique Salem and Chase Lochmiller

Renewable Opportunity

Going into the company's Series C in 2022, Crusoe expanded beyond flared gas to pursue a second massive opportunity: renewables. In markets like Texas, federal production tax credits were incentivizing developers to build wind and solar farms in remote locations where the sun and wind were strongest — if not quite not where people actually needed power. The result was massive oversupply and frequently negative electricity prices, as generators paid customers to take excess power while still collecting their subsidies.

Crusoe again spotted an arbitrage opportunity of massive amounts of cheap, wasted energy with nowhere to go. With utilities not building billion-dollar transmission lines to move that excess energy to cities, Crusoe planted data centers right next to the turbines and solar panels. It was a perfect match for AI computing, which gulps power by the megawatt but can ramp up and down to chase whatever electrons are available.

Its initial customers were research labs at universities like MIT doing tasks like computer simulations of the universe. Then the ChatGPT moment hit in November 2022 and the industry's explosion around AI kicked off a frenzy of building data centers.

From 2022 through 2024, US tech giants spent roughly $485 billion on AI infrastructure, according to Goldman Sachs. The growth catapulted Nvidia into the most valuable company in the world and turned data center capacity into the tech industry's most precious resource.

Companies like CoreWeave, also a former crypto mining operation, built custom clusters of GPUs numbering in the thousands. Investors bet that these “neoclouds” had unique skills well-suited for the AI era, where surging demand created a bottleneck for chips, components—and most of all, power.Crusoe’s bet was that speed would belong to the companies that owned more of the pipeline. Instead of splitting the work across vendors, utilizing one company to build, another to operate the cloud, and yet another to supply the electrical gear, Crusoe pushed toward a vertically integrated model: data centers at the edge of cheap power, a cloud product to monetize the GPUs, and in-house manufacturing for key electrical components that could otherwise slow deployments.

Stargate's flagship data center campus in Abilene, TX

Abilene

In January 2024, Lochmiller caught wind of an RFP for a new 100-megawatt data center. Most bidders were quoting two-and-a-half years, which was already a breakneck pace for a project that large.

Lochmiller turned to the company's new chief data center officer, Chris Dolan. Could they do it faster? Like in a year?

"Chris and I were brainstorming how we might be able to do this, and I realized yeah, I think we could," Lochmiller said.

While other data center builders were looking to typical hubs like Northern Virginia or Columbus, Ohio, none had the capacity Crusoe needed. Lochmiller then remembered a site he'd come across in Abilene, Texas, a railroad town in the western part of the state that had become a hub for renewable energy development. He happened to be friends with the CEO of Lancium, a local energy infrastructure company that was developing a site there for bitcoin mining.

The timing was perfect: Lancium had just completed a 100-megawatt substation with approval to scale up to one gigawatt. Lochmiller asked if they'd be interested in repurposing the facility for AI; the Lancium CEO loved it. It was symbolic of Crusoe's own shift: bitcoin mining had gotten them started, but AI was where they were heading.

"To me, this was a gold mine," Lochmiller said. "We put Abilene on the map by putting a shovel in the ground there. It's really becoming one of the most important data center places by virtue of us building there."

With the site secured, Crusoe needed customers. Lochmiller turned to Oracle, the database giant trying to break into cloud computing during the AI boom. Oracle was similarly interested in building at breakneck speed. The company was on board.

The first potential client interested in the Abilene site was reportedly Elon Musk's xAI. But after Musk decided to pull out and build a megacluster on his own, they turned to OpenAI—the account Lochmiller had courted a few years earlier.

The deal OpenAI had with Microsoft required Azure to be its exclusive cloud provider. But when Microsoft balked at the size of the data center OpenAI was looking for, it gave its partner a waiver to go elsewhere. Oracle won the deal and had the perfect site for its newest AI factory.

In September 2025, when word got out that Oracle’s contract with OpenAI was worth $300 billion, its stock surged 40 percent in a single day.

Now Oracle and Crusoe are racing to deliver. Lochmiller said 5,000 workers are on site daily in Abilene, toiling in the red dirt to assemble the facility. Part of the work involves building gas turbine generators that will provide primary power.

The Abilene project is just the beginning. In October 2025 the company announced a $1.375 billion Series E funding round that valued that company above $10 billion. Crusoe is now scouting locations for new data centers, including a 5-gigawatt facility—the same amount of power that exists across all of the data centers in Northern Virginia. For Lochmiller, the scale keeps expanding, but the approach remains the same.

Like any climber, he's always preparing for the next peak, whether or not it's visible yet. The AI infrastructure demands that seem impossible today will be baseline tomorrow.

"It's hard to imagine anything that can't benefit by having more intelligence embedded in it."