Crusoe abandons $1.25B plan to use Boom turbines at AI data centers

Crusoe has terminated its $1.25 billion agreement to buy 29 of Boom Supersonic's 42MW natural-gas Superpower turbines, eliminating the stationary-power venture's launch customer ahead of planned 2027 deliveries. The loss is a material setback for Boom, which raised $300 million last year and intended to use turbine-business profits to help fund development of its Overture supersonic jet. Boom said it expects to deliver roughly 250MW of turbines to other sites next year and targets 1GW in 2028, while Crusoe will retain flexibility across turbines, renewables, batteries and grid power for its expanding AI data-center portfolio.

Analysis

The relevant signal is not a public-equity earnings event but a procurement warning: AI-campus developers appear unwilling to anchor primary-power architecture to an unproven turbine platform when grid interconnection, modular generation, storage, and established OEM equipment offer lower execution risk. That favors bankable suppliers with installed-service networks—GE Vernova (GEV), Siemens Energy (ENR.DE), Caterpillar (CAT), Vertiv (VRT), and Eaton (ETN)—because financing and construction schedules increasingly value delivery certainty over nominal equipment innovation.

For the next 1-3 months, this is modestly supportive of the listed AI-power ecosystem rather than a standalone catalyst. GEV and CAT have the clearest upside if hyperscale-related onsite generation demand shifts toward proven gas turbines and reciprocating engines; VRT and ETN benefit regardless of whether the marginal watt comes from grid, generation, or batteries because power conditioning, distribution, and thermal infrastructure remain required. The second-order risk is that a more distributed power mix raises campus-level capex and integration complexity, potentially slowing AI capacity commissioning and shifting spend from compute hardware toward electrical infrastructure.

Over 6-18 months, the key contrarian point is that this does not prove gas generation demand is weakening. It more likely shows that startup OEM financing, certification, warranty support, fuel economics, and delivery credibility are gating deployment. Boom's stated replacement pipeline should be treated as unverified until binding customer commitments, project financing, and manufacturing capacity are disclosed; failure there would reinforce the valuation premium assigned to incumbent power-equipment vendors.

The thesis is falsified if major AI-campus operators disclose sustained preference for grid-connected renewable-plus-storage builds with no incremental firm-generation procurement, or if GEV/CAT report AI-related order growth failing to convert into backlog and service revenue. A sharp decline in US power prices or natural-gas generation margins would also reduce the urgency of onsite-power solutions.

Key Decisions for Investors

• Maintain or initiate a 3-6 month long GEV / short XLI pair: GEV is the cleaner listed beneficiary of firm-power scarcity and service-backed turbine demand, while XLI limits broad industrial-beta exposure. Target 10-15% relative upside; exit if GEV's next reported gas-power equipment backlog or 2026 delivery outlook weakens.
• Buy VRT on 5-8% pullbacks rather than chase event strength; use a 6-12 month horizon. The trade captures power-distribution and cooling spend that persists across generation choices, with risk defined by hyperscaler capex guidance cuts or evidence that AI deployments are materially delayed.
• Watch CAT and CMI for disclosed data-center power orders before establishing size. CAT has a more credible near-term route to incremental reciprocating-generation demand, but the missing data are order size, utilization duration, and whether deployments displace rather than supplement grid power; treat this as an earnings-call alert, not a fresh directional recommendation.
• Avoid treating the private turbine developer's claimed replacement demand as validation of a new stationary-power market. For public-market positioning, wait for independently observable equipment orders, project financing, or named end customers; absent those, the better risk-adjusted exposure remains incumbent OEMs and electrical-equipment suppliers.