Crusoe CEO Chase Lochmiller overrates the AI supermajor hedge
By Klaus Berger · Reporting from Frankfurt ·
Chase Lochmiller claims Crusoe can operate as an integrated AI supermajor like ExxonMobil, but layering massive capital expenditure over speculative cloud demand concentrates operational leverage rather than de-risking it.
On the Sequoia Capital podcast, Crusoe chief executive Chase Lochmiller articulated an ambitious vision for his infrastructure business: building an integrated “AI supermajor” modeled directly on oil titans ExxonMobil and Chevron. Fresh off a funding round valuing his firm at $31 billion, Lochmiller detailed Crusoe’s transformation from a flare-gas harvester into a vertically integrated powerhouse with over 2,000 employees. The company now builds power plants, manufactures custom electrical equipment, operates gigawatt-scale data centers, and runs a managed artificial intelligence cloud platform. To finance this single-minded focus on computing capacity, Lochmiller sold off the lucrative Bitcoin mining business that initially paid the bills, divesting the division to natural gas operator NYDIG. On paper, it is a masterclass in corporate scaling; on the balance sheet, it is a dangerous concentration of duration risk.
The Myth of the Natural Hedge
Lochmiller’s argument for vertical integration rests on a flawed financial analogy. In oil and gas, integrated majors like ExxonMobil possess a genuine structural hedge. When crude prices fall, upstream wellhead margins collapse, but downstream refining margins expand because raw input costs decline. Lochmiller claimed on the podcast that Crusoe enjoys a similar balance-sheet hedge by controlling everything from power generation to cloud software.
This misapprehends basic industrial economics. Power generation and AI compute do not share counter-cyclical margin dynamics. If demand for AI model training or inference cools, lower electricity costs do not magically expand the profit margins of an idle server cluster carrying billions in debt. When end-user demand drops, a vertically integrated infrastructure provider does not capture an offsetting margin; it simply owns both sides of a non-performing asset—the underutilized turbine and the uncontracted silicon. Far from de-risking the enterprise, integrating heavy power generation with speculative cloud hosting doubles down on the same directional exposure to artificial intelligence capital expenditure.
Capital Intensity and Moral Hazard
The decision to sell the Bitcoin mining business illustrates the moral hazard embedded in venture-backed infrastructure. Bitcoin mining offered Crusoe an immediate, liquid, cash-generating customer for stranded power. Replacing that liquid monetization engine with long-gestation AI infrastructure requires enormous capital expenditure before a single token is billed. Lochmiller highlighted partnerships with nuclear firms like Oklo at the Idaho National Lab and Blue Energy, alongside projects in Iceland and Norway.
Yet building power generation alongside high-density data centers is exceptionally capital-intensive. In fixed-income markets, utility-scale power projects demand long-term power purchase agreements, low cost of capital, and regulated rates of return. Venturing into gigawatt-scale power development under a private equity framework creates a classic moral hazard: management assumes that private capital markets will indefinitely refinance capital-heavy assets before long-term demand is established. When equity investors value a capital-intensive utility builder at $31 billion based on tech-cloud multiples, they misprice both credit risk and operational complexity.
Crusoe’s evolution from monetizing wasted flare gas to underwriting power generation represents a dramatic shift in financial leverage. Modeling a technology startup on an oil supermajor sounds reassuring in pitch decks, but physical assets obey the unyielding rules of capital allocation, not software economics. If the current wave of artificial intelligence investment slows, heavy physical balance sheets cannot be recompiled, and unhedged liabilities will fall squarely where they always do—on the equity holders who mistook leverage for structural advantage.