Converted Jet Engines Won a $1.465 Billion Order — and the AI Power Queue Still Runs to 2031

This is no longer a pilot project. The $1.465 billion purchase order flows through a joint venture rather than through FTAI Aviation directly, which tells you how new the plumbing is: Jereh Group packages and distributes the sets while FTAI Aviation supplies the cores, the conversion and the aftermarket behind them. The customer is described only as a leading international cloud service provider, and deliveries run in batches through November 2027.

The feedstock sets a ceiling no order book can lift. ProEnergy expects roughly 1,000 aircraft engines to retire over the next decade, and each CF6-80C2-derived PE6000 turns into 48 megawatts, implying about 4.8 GW a year of repurposed capacity. Against that, GE Vernova alone shipped 5 GW of gas equipment in the third quarter of 2026. The converted-engine channel is rounding error in megawatts and decisive in months.

Hyperscalers are not buying converted engines because they are cheap per kilowatt. They are buying them because new-build turbines are gone until 2029, with orders running to 2031, while a converted unit can be delivered in 2027. ProEnergy's two sold-out projects expect to run bridging power for five to seven years, until grid interconnection arrives. The buyer is paying a premium to start earning on GPUs it has already bought, not to lower its lifetime power bill.

That reframes the pricing. At 25 megawatts a unit, the $1.465 billion purchase order implies roughly $585 to $1,170 per kilowatt of peak capacity, depending on whether it covers 50 or 100 units. That price only works because the buyer values a 2027 start higher than it values lifetime cost, which puts the commercial risk on the operator rather than the engine seller. If the data center is not paid for and connected, the premium is never collected.

The counter-intuitive part is that the bridge channel makes the new-build order book more valuable, not less. GE Vernova entered the third quarter of 2026 with 116 GW of gas power backlog and slot reservations and left with 122 GW, converting 13 GW of reservations into firm orders while shipping 5 GW. Every data center that takes a converted engine instead of a 2029 slot stops competing for the scarce slot, and the slot holder can wait without losing the project.

None of this requires the bridge channel to fail. It requires only that the wait stays long, and GE Vernova guides to 2026 revenue of $46.5 billion to $47.5 billion at a 13.5% to 14.5% adjusted EBITDA margin, with free cash flow of $16 billion to $17 billion. Third-quarter revenue of $12.6 billion and net income of $1.29 billion show what that backlog is currently worth.

The second-order effect sits in aerospace, not power. GE Aerospace grew commercial engines and services revenue 21% to $10.34 billion in the third quarter of 2026, with services up 17% and spare parts up more than 20%. A converted engine burns through the same parts catalogue as a flying one, and FTAI Aviation signed a multi-year agreement with CFM International on January 22, 2026 covering OEM replacement parts, thrust performance upgrades and component repair for CFM56 engines.

So the toll is not a single toll. Engine owners collect it on the conversion, parts suppliers collect it on the annuity, and the manufacturer that refuses to sell new turbines collects it on the queue. The feedstock constraint that caps the first two is what makes the third one valuable.

The next wave of this trade is being built in Korea and Japan, and it is aimed at the same American customers. On October 6, 2026, Doosan Enerbility shipped its first 380-megawatt gas turbine to the United States for a project led by SpaceXAI, a subsidiary of SpaceX, marking the Korean maker's entry into the American power and AI infrastructure market. The turbine is the first of a package Doosan says it contracted for twelve US customers.

For SpaceX, self-supplying power is now a capital allocation question rather than a procurement question. Third-quarter 2026 revenue reached $14.6 billion with $4.4 billion of net income, and the first 380 MW unit for its SpaceXAI project has already shipped. A 6.4 GW site with no turbine supplier chosen is either the largest power decision the company has made or the largest it has deferred.

The synthesis is narrow, and it is mispriced by a market reading this as a power story. It is a timing story with a hard supply ceiling: roughly 1,000 engines, two certified converters, one 25-megawatt design that has already passed critical testing, and a hyperscaler willing to write a $1.465 billion check for 2027 delivery. The engine owner captures the conversion spread, the parts supplier captures the annuity, the turbine maker captures the queue, and the data center operator captures five to seven years of revenue it could not otherwise have. The last of those is the largest prize, and it is the one no public company in this chain owns.