Compute & Market Power
Boom's New Delivery Target Is 21% of Crusoe's Order
Crusoe has dropped its Boom turbine partnership. Boom's reported 250MW next-year target equals 20.7% of the old order, not replacement backlog.
Crusoe has abandoned its launch partnership for Boom’s Superpower turbines, so infrastructure buyers should remove that relationship from their evidence of committed supply. Boom’s reported 250 MW delivery target for other sites next year is 20.7% of the 1.21 GW order in its original announcement—a scale comparison across different time horizons, not a measure of replacement backlog.
A launch customer is not a commissioning certificate
The denominator comes from December’s primary release. Boom announced an order for 29 Superpower units from Crusoe, describing it as 1.21 gigawatts and a backlog exceeding $1.25 billion. It also announced $300 million of funding. The September cancellation report says the partnership is no longer proceeding. Those are different events: the original order made an industrial plan more credible; its loss requires buyers to reassess which part of that plan still has a customer behind it.
The new target comes from CEO Blake Scholl’s statement reproduced in TechCrunch’s reporting: about 250 megawatts of deliveries to other sites next year, with a one-gigawatt target for 2028. Convert the original 1.21 GW to 1,210 MW, then divide 250 by 1,210: 20.7%, rounded. The calculation deliberately uses Boom’s announced rounded order capacity, rather than inventing precision from unit ratings. It shows how modest the near-term delivery target is relative to the lost anchor order.
It does not show that Boom has replaced only that share of its total sales pipeline. An annual delivery goal and a multi-year order are not equivalent quantities. The reporting does not disclose the contracts, customer identities, deposits, or acceptance terms behind every prospective delivery. Calling the difference an unfilled backlog would go beyond the evidence. The useful procurement conclusion is that the named launch reference has disappeared and the replacement evidence is less specific.
There is also a disputed explanation to preserve. TechCrunch reports that Scholl removed language suggesting turbines were no longer part of Crusoe’s near-term primary power mix. Crusoe’s spokesperson said the energy plans had not changed and that the company still intended to use turbines, just not Boom’s. The cancellation is supported; a general retreat from turbine power is not. Buyers should not let a supplier-specific breakup become an unsupported claim about an entire energy technology.
The financing context reinforces why the distinction matters. TechCrunch’s original funding coverage described the stationary-turbine business as a way to support Boom’s aircraft ambitions. Losing the first customer therefore touches an important commercial reference. It does not establish insolvency, an engineering failure, or the end of the product. None of those stronger conclusions follows from the retrieved record.
Our earlier Crusoe analysis covered its financing and delivery-backed contracting problem. This development is a separate event with a sharper implication: a specific power-supply arrangement can disappear even while the infrastructure buyer continues expanding. Company-level financing and supplier-level continuity belong in different rows of the risk register.
Replace the reference with evidence, not a new promise
For an infrastructure procurement team, the immediate action is to check whether a proposal or internal approval relied on Crusoe as the Superpower launch customer. If so, update that assumption and ask what now supports the delivery schedule. Require named manufacturing milestones, site readiness, acceptance tests, and remedies for delay. A supplier can still be worth evaluating after losing an anchor order, but it must qualify on the remaining evidence.
Price must stay at the right boundary. The cancelled arrangement’s reported $1.25 billion value does not reveal the cost of a new customer’s fully installed, fueled, permitted, maintained power supply. Nor does dividing it by megawatts establish a useful electricity tariff. Public reporting supplies neither a current replacement quote nor the full operating cost stack. Ask for a site-specific offer and identify who bears each dependency before treating the technology as cheaper capacity.
Crusoe’s recent financing coverage describes an expanding data-center and modular-infrastructure business. That broader growth is compatible with choosing a different turbine supplier. The operational question is not which company’s narrative wins. It is whether the buyer’s own capacity plan still has a credible path from equipment order to accepted power at the required site.
The strongest counterpoint is commercial flexibility. A first partnership can end because requirements or priorities diverge, while other customers remain a better fit. Boom’s reported next-year delivery target is a reason to monitor execution rather than write off the program. But a target remains weaker evidence than commissioned equipment and a satisfied operating customer. The burden of proof should rise with the irreversibility of the buyer’s commitment.
This is the physical counterpart to today’s Docker execution-boundary analysis: an available component is not the same thing as a verified service. The software team needs a result it can trust; the infrastructure team needs accepted power under documented conditions. In both cases, the appealing headline becomes useful only after its dependencies are made explicit.
Evidence that would change the recommendation includes completed factory testing, named replacement customers with binding schedules, and actual operation at the promised site. Evidence that would weaken it includes slipping milestones or a proposal that depends on the cancelled relationship as though it still existed. These are forward-looking evaluation criteria, not allegations of undisclosed failures.
The verdict is to requalify the supplier, not abandon turbines. Buyers exposed to this particular reference should reopen diligence now and keep alternatives alive until acceptance. Buyers using other power arrangements should not infer a problem that the reporting does not establish. The lost order changes the evidence behind one commercial plan; it does not erase every way to power an AI campus.