Chase Lochmiller on Building Crusoe and Scaling an AI Infrastructure Company

Chase Lochmiller describes Crusoe as a vertically integrated business spanning power generation, data centers and a managed AI cloud. Its teams range from power specialists and electrical-equipment manufacturers to researchers working on model serving. He says the breadth makes it essential for leaders to understand how their part contributes to the whole.

The host introduces Lochmiller and says Crusoe has recently closed financing at a $31 billion valuation. Lochmiller’s path has included academic research, quantitative trading, mountaineering and energy entrepreneurship. Asked what connects those pursuits, he says he has consistently been drawn to difficult, inspiring problems.

The host asks what people misunderstand about becoming a CEO. Lochmiller says a snapshot of a mature leader can obscure the gradual process: Crusoe began with him and his co-founder, then added its first employees one by one. Even ten people once felt hectic, but the company’s demands grew in increments.

Lochmiller says a founder must keep investing in personal growth if the company is to scale. At Crusoe, the idea is an “infinite growth loop”: neither the company nor its people are finished products. Small improvements compounded over months and years can make a striking difference.

The host describes a group of Sequoia founder CEOs whose businesses range from roughly 20 to 2,000 employees. One said that the company was becoming dizzyingly complex beneath them. Lochmiller replies that he too would be overwhelmed if dropped into Crusoe without the years of context he accumulated along the way.

That history alone is not enough. Lochmiller says he has surrounded himself with capable leaders and can entrust them with resources and decisions. He can go deep with each specialist, but expects everyone also to understand the broader objective and how their individual work affects the rest of the company.

The host asks whether trust comes naturally. Lochmiller says his team would not describe him as entirely hands-off. He wants to understand each area, agree on its direction and its connections to other parts of the business, and then hold the responsible leader accountable for executing the plan.

Cross-functional work remains difficult even with a program-management function, Lochmiller says. His weekly staff meeting brings together senior executives and direct reports, allowing friction between groups to surface. In a vertically integrated company, cooperation at the boundaries between layers is necessary for the model to work.

Lochmiller learned that communicating once is insufficient in a large organization. In a small team a single conversation could align everyone; now he often has to repeat what the company is doing, why it matters and who owns each part. By the time he is tired of saying it, the message may only just be reaching everyone.

The host asks how a larger company maintains speed. Lochmiller says Crusoe still moves urgently and warns against the sense that a company has already made it and can coast. He admires Elon Musk’s ability to combine a compelling vision of changing the world with intense pressure to act quickly.

Lochmiller says he is less extreme than Musk, but considers the tension between opportunity and urgency an important management problem. Crusoe has surpassed 2,000 employees, making it the largest company he has worked for. He has studied business books, sought mentors and learned through mistakes to develop as a leader.

One early mistake was letting news about competitors consume too much attention. Lochmiller recalls advice that startups are more likely to fail through their own execution than be killed by rivals. In his view, meaningful markets often support multiple winners, so founders should concentrate on culture, talent and the work in front of them.

He does not see AI infrastructure as a winner-take-all market. Organizations that appear to compete with Crusoe may also be valuable partners. He argues that building the infrastructure for intelligence is large enough to create many opportunities to collaborate, though each business still has to execute.

Lochmiller compares the AI computing stack to oil and gas. Upstream companies produce oil and gas, midstream firms transport it, and downstream operations refine it into products. He wants Crusoe to resemble an integrated energy major, with different activities offering opportunities both to compete and to serve other companies.

In his analogy, margins at different parts of the value chain can move in opposite directions. He cites Exxon’s upstream exposure and downstream refining as an example of natural balancing when oil prices change. The comparison is his account of a business model, not a guarantee that AI infrastructure will behave the same way.

The host asks how Crusoe’s energy approach has evolved, mentioning wasted energy and natural gas. Lochmiller says the company seeks to activate power for computing through different sources: gas, wind, solar, batteries and possible nuclear technologies. He describes a partnership aimed at an SMR-powered AI factory at Idaho National Laboratory as an initiative, not a completed facility.

In Iceland, Lochmiller points to geothermal and hydropower resources, relatively low population and efficient cooling as reasons to bring computing demand to the energy source. He also discusses hydropower in Norway. His “energy-first” idea is to locate compute where suitable power is available rather than assume one energy solution works everywhere.

He is interested in several storage approaches, including repurposed electric-vehicle batteries and longer-duration systems. He believes rising demand for AI computing will require a mix of energy technologies. These are his expectations about future demand and technology, not established outcomes.

The host asks about moving from a thriving Bitcoin mining operation into AI infrastructure. Lochmiller says that, although outsiders saw a pivot, the company’s longer-term plan was always to bring computing to underused energy. His prior work in quantitative finance and graduate study at Stanford had already exposed him to AI research.

Lochmiller recalls the influence of writing about superintelligence and his Stanford professor Andrew Ng’s analogy that AI is like electricity: useful across many activities rather than just one. His thesis was that if compute becomes a bottleneck for AI, providing sufficient power becomes a central part of the problem.

Crusoe initially focused on natural gas produced alongside oil where there was no pipeline to take it away. Instead of burning the gas as waste, the company placed modular generators and small data centers near it. Bitcoin mining was an early way to turn that local electricity into revenue.

Lochmiller says people questioned why Crusoe should invest in an AI cloud while Bitcoin mining was profitable. The early AI efforts involved failures and several iterations. He recalls asking university researchers whether they could use GPUs powered by gas in North Dakota’s Bakken oil field.

He identifies the launch of ChatGPT in November 2022 as a turning point: a wave of startups began demanding the kind of compute, storage and networking infrastructure Crusoe had been trying to build. This is his retrospective account of how the market developed around the company.

Lochmiller says Bitcoin mining and AI computing can have complementary power profiles. Mining can be adjusted to a relatively steady draw and moved up or down with available power; AI workloads and cooling demand vary. He describes these differences as one reason he saw the two activities fitting together.

Pressed on why Crusoe sold a good business, Lochmiller says NYDIG offered what he regarded as a suitable home for it, with experience in both Bitcoin and natural gas. The central reason for selling, however, was focus: the company had to decide where to direct capital, resources and management attention.

Some prospective investors liked the AI business but were prohibited by their firms from investing in crypto-linked activities. Lochmiller says divesting made the company’s AI infrastructure story easier to understand and allowed it to concentrate. He adds that he remains personally supportive of Bitcoin; the business decision was to go all in on AI infrastructure.

The host notes Lochmiller’s polished public speaking and asks whether it came naturally. Lochmiller says it did not. His quantitative-finance background involved working with technical colleagues while keeping information private, so speaking publicly was a skill he had to prepare for and practise.

With repetition, he says, he became more comfortable explaining his thinking and less dependent on preparation for every appearance. Some situations still require extensive preparation, but he sometimes speaks better off the cuff because he has developed the ability to make his ideas clear.

Lochmiller credits his coach Shane Metcalf with helping him grow as a leader. He also describes an investor, Felicis, setting aside resources for founders’ personal development, such as executive coaching and offsites. To him, that recognizes that a venture investment depends partly on a founder’s capacity to improve as the company grows.

He says a coaching relationship builds trust over time, even when an individual session does not seem immediately productive. When a major problem appears, the coach knows the leader’s blind spots and can help them adjust. The host compares some of his own work with founders to listening and offering a form of support.

For Lochmiller, many of the hardest company-building problems involve people: aligning smart, ambitious colleagues, helping people commit despite disagreement, and deciding when to keep investing in someone or change who occupies a role. Getting those relationships right can also be a company’s greatest strength.

The host asks whether he has waited too long to replace people. Lochmiller believes CEOs often do, though he says no single signal tells the whole story. Feedback from the team is one data point among many; the CEO must combine information from different parts of the organization and exercise judgment.

He describes a difficult mismatch: companies can grow much faster than individuals develop, so a person who was right for a role at one stage may not be right for the same role when the company is much larger. That is painful when colleagues have worked hard together and care about the mission.

The host contrasts leaders who keep the same executive team for years with those who turn over executives more frequently. Lochmiller says there is no single correct pattern. He places himself between the extremes and compares a struggling team member to a sports teammate: the personal bond matters, but so does the team’s ability to succeed.

The host asks how a company with welders and AI researchers can build a common culture. Lochmiller points to what Crusoe calls collective genius: no individual has the expertise to deliver the whole system. He says innovations from electricians and other people working on physical infrastructure can ultimately benefit researchers and customers.

The conversation turns to public opposition to data centers, including concerns in communities where facilities are built. Lochmiller says he believes infrastructure can be developed in ways that benefit communities, citing questions of energy, water, tax revenue and schools. These are his claims and aspirations, not an independent assessment of any particular project.

He argues that the industry has not explained its work effectively and describes meeting local officials and residents to discuss projects. He says he tries to explain what the company is doing and listen without assuming the audience is uninformed. He also says Crusoe has increased its security in response to concerns about threats.

Asked about facing angry criticism, Lochmiller says leaders need to understand what is behind objections and be willing to change their minds when new facts emerge. He distinguishes, from his perspective, claims he regards as unfounded from legitimate feedback that could help the company be a better community steward.

He says a CEO must receive bad news and still make decisions clearly. As the company expands, he worries about the growing distance between himself and people on the front lines. He wants employees to feel they can speak honestly and know their concerns will be heard.

His method is to ask persistently what is going wrong and demonstrate that criticism can change his decisions. When people see a leader accept feedback, he says, they may be more willing to speak up. He wants to seek out problems rather than hear only reassuring news.

Asked about Everest, Lochmiller connects mountaineering with a Crusoe value: resilience, planning, preparation and endurance. He recounts an attempted climb in 2014 interrupted by a deadly avalanche in the Khumbu Icefall. Afterward he still wanted to return, although his wife asked that he do so before they had children.

He describes two especially dangerous parts of the southern route: the high-altitude “death zone” above roughly 8,000 meters and the Khumbu Icefall between base camp and Camp I. He felt the latter involved unpredictable hazards, such as falling ice and avalanches, that could not simply be trained away.

A conventional expedition’s repeated acclimatization trips can require eight crossings of the icefall, according to Lochmiller. He and his climbing partner instead spent time on nearby mountains to acclimatize, reducing their Everest icefall crossings to four. He presents this as reducing exposure, not eliminating the mountain’s risks.

Lochmiller says the plan required extensive preparation, long-distance trail running and high-altitude training. He reached the summit in 2018. The lesson he emphasizes is conscious risk-taking: identify the hazards, decide which risks to accept and prepare carefully for them.

Near a point called the Balcony, at about 8,600 meters, he nearly turned back. He recalls darkness, dehydration, little food and cramping legs. His climbing partner encouraged him; after sunrise, the surroundings lifted his spirits and helped him make the final push.

At the summit, Lochmiller felt awe and a sense of being small in the world. He says the weather was sunny and the team spent about an hour there taking pictures and appreciating the achievement. The host thanks him for the conversation.

The host closes with his own takeaways. He values Lochmiller’s appetite for hard challenges and his willingness to learn through reading, coaching and practising public speaking. He also notes that founders can collaborate with competitors and should build relationships before they know exactly when they will need them.

Finally, the host says leaders must train their organizations to deliver bad news, including by welcoming disagreement in front of others. He praises having strong opinions while remaining willing to change them when facts change. The conversation ends with his encouragement to continue learning as a company grows.