Crusoe is one of the most aggressive expanders in the AI data center space. On September 25, 2026 (US time), Boom Supersonic CEO Blake Scholl announced on X that Crusoe would no longer move forward with its $1.25 billion gas turbine partnership with Boom. Crusoe later confirmed the news to TechCrunch. That kills the December order for 29 turbines totaling about 1.21 gigawatts, with first deliveries originally slated for 2027.
How a $1.25 billion order fell apart
Boom Supersonic's main business is developing the Overture supersonic airliner. Last year it opened a side business: converting jet-engine technology from its in-development Symphony engine into stationary gas turbines called "Superpower," sold to data centers as power plants. Superpower shares roughly 80% of its parts with Symphony, with each unit rated at 42 megawatts. When Crusoe signed on as launch customer, it effectively vouched for Boom's new venture — Boom even raised $300 million to plow turbine profits back into aircraft development.
Scholl's post was diplomatic: first congratulating the Crusoe founders on their freshly closed $3.9 billion raise, then announcing the turbine launch partnership was off. He later deleted one TL/DR line from the post: "turbines are no longer part of Crusoe's near-term primary power mix at Abilene/etc., so a launch partnership just didn't make sense." Deleted or not, the message was clear: this wasn't a price dispute — Crusoe's power roadmap had changed.
Crusoe spokesperson Andrew Schmitt offered TechCrunch a different framing: "Our energy plans haven't changed — we still plan to use turbines, just not Boom's." He described a site-by-site, flexible approach — wind, solar, batteries, conventional gas turbines and the grid — "choosing the energy solutions that are right for each site." In plain terms: they still need the power, but they won't bet on a single vendor's unproven hardware.
What losing the launch customer costs Boom
For Boom, this is a genuine setback. The whole commercial logic of Superpower was "use data center orders to fund the airliner," and losing the launch customer dents both the delivery timeline and the fundraising story. Scholl's post included a consolation line: about 250 MW to other sites next year, targeting 1 GW in 2028. TechCrunch also noted Scholl could not be reached for comment before publication.
Crusoe's own power puzzle is actually well filled in: its 1.2-gigawatt Abilene, Texas campus built for Oracle and OpenAI runs on grid power, with a gas-turbine plant kept as backup only; the 900-megawatt campus under construction for Microsoft will run directly on on-site gas turbines. In other words, Crusoe isn't abandoning gas turbines — it already has suppliers with more certain delivery. For Crusoe, Boom's Superpower went from "must-have" to "optional."
Power, not GPUs, is the real AI infrastructure bottleneck
The weight of this story isn't the $1.25 billion itself — it's the industry reality it exposes: the binding constraint on AI data centers is no longer GPUs, it's electricity. Substation approvals can take three to five years, and grid expansion can't keep up with compute growth, which is why gas turbines, battery storage and behind-the-meter generation have become fiercely contested ground. Crusoe's playbook has always been "build AI factories from the power up" — Schmitt's own words — and the company that started in 2018 mining bitcoin on stranded oilfield gas now treats power as the first principle of AI infrastructure.
Boom's crossover attempt shows exactly how attractive this lane has become: even a supersonic-jet company wants to sell electricity. This time, though, the buyer voted with its feet — Crusoe chose a more mature power solution with lower delivery risk. The AI infrastructure power race will go on, but the game is shifting from "betting on novel technology" back to "competing on delivery certainty." For anyone planning a new campus, that's a signal worth noting: where your power comes from can decide whether you break ground before which model you run does.
