Retired Narrowbody Engines Finding Second Life in AI Data Centers, and What It Means for the Used Parts Market

A new class of buyer has entered the market for end-of-life commercial jet engines - and it isn’t an airline. Retired CFM56 turbofans, the engines that powered the majority of the world’s narrowbody fleet for the past three decades, are being evaluated as working power generators for artificial intelligence data centers. The implications for asset values, used parts availability, and MRO cost structures are only beginning to surface.

The Engine at the Center of This Story

CFM International - a joint venture between GE Aerospace and Safran Aircraft Engines formed in the 1970s - built one of the most prolific commercial engine programs in aviation history. The CFM56’s first flight on a modified Boeing 707 testbed came in 1974. Full commercial service on the Boeing 737 followed in the mid-1980s.

Total production now exceeds 33,000 engines. Those engines powered the 737 Classic, 737 Next Generation, and 737 derivative variants, as well as the entire Airbus A320 family in its original CEO (current engine option) configuration. If you’ve flown commercially on a narrowbody at virtually any point in the last 30 years, there’s a strong chance a CFM56 made that flight possible.

That fleet is now aging out. 737 Classics are largely out of first-tier service in developed markets. Earlier 737 NG aircraft are approaching mid- and end-of-life decisions. A320 CEO aircraft face competitive pressure from the NEO family and the 737 MAX. As those aircraft retire, the engines come with them - in substantial numbers.

How Jet Engine End-of-Life Economics Work

An operator decides an engine has reached end of commercial service when the cost of the next major maintenance event no longer makes economic sense. The driving factor is life-limited parts - high-stress components in the hot section and rotating assemblies with fixed cycle limits established under FAA regulations. A cycle is one complete takeoff-and-landing sequence. When a life-limited part hits its limit, it comes out. No extensions.

When multiple life-limited parts converge on their limits simultaneously, the shop visit cost to replace them all can exceed both the engine’s resale value and any operational return a continued run would generate. At that point, the engine typically goes to teardown.

Teardown is methodical. A specialist evaluates every component. Parts with remaining service life get certified as used serviceable material (USM) and sold into the market. Parts at or near limits get scrapped. That USM flows into MRO shops as an affordable alternative to new OEM parts - which is a core reason commercial shop visit costs remain manageable at all.

Why AI Data Centers Are Buying Retired Jet Engines

The alternative path, the one reshaping this market, involves a buyer who doesn’t need the parts individually. A buyer who doesn’t need the engine to be airworthy. A buyer who needs it to generate electricity, continuously, at scale.

Industrial gas turbines derived from aero engine cores have powered pipelines, offshore platforms, and utility generation for decades. GE Vernova’s aeroderivative turbines, Rolls-Royce industrial programs, and Pratt & Whitney’s equivalent products are purpose-designed for fixed-installation service. What the data center market is now pursuing is different: engines designed for flight, retired from commercial service, being evaluated for a second working life as generation equipment rather than teardown.

The engineering logic is sound. Stationary ground-level power generation is a substantially less demanding environment than airline service. The thermal cycling, pressure swings, and vibration loads of flight impose fatigue on hardware in ways that steady-state generation on a concrete pad does not. An adapted turbofan in this application can produce roughly 20 to 40 megawatts of electrical output depending on engine type and generator configuration - enough to power between 15,000 and 30,000 average homes.

The Grid Problem Driving the Demand

Computing workloads for large-scale AI training and inference are unlike anything the data center industry has historically handled. Training a frontier AI model requires sustained operation of thousands of specialized processors running in parallel, often for weeks or months at a time. A large AI training cluster can consume tens to hundreds of megawatts without interruption.

Grid operators and utilities are describing the resulting load growth as unprecedented. Northern Virginia, the Phoenix area, parts of Texas, and the Pacific Northwest corridor have seen data center construction accelerate at a pace the existing grid cannot match. Utility interconnection queues in many of these markets carry backlogs measured in years. A developer who needs significant power within the next 12 months may be looking at a connection date toward the end of this decade.

That gap has driven serious interest in behind-the-meter generation - generating power on-site, independent of the interconnection queue. Diesel generators are not economical for continuous baseload. Natural gas distributed generation is attractive but permitting and equipment lead times run long. Used jet engines, available through established commercial channels, fit a specific niche.

What This Means for Asset Values and the Parts Market

The business model creates a new bidding dynamic. An engine that might yield $2 to $4 million through the traditional teardown and parts sale process may command a higher intact price from a power generation buyer with the right application and timeline. That shifts the floor on end-of-life asset values - a development that matters to operators, lessors, and anyone projecting residual values on aging narrowbody fleets.

For operators and lessors, the development is potentially favorable. If a new buyer class competes with the teardown market for intact engines, residual values on end-of-life assets improve. That matters during asset retirement planning and lease return negotiations.

For the used serviceable material market, the picture is more complicated. Parts that don’t get torn down don’t return to the pool. If the pipeline of CFM56 USM tightens, prices for key components could rise. MRO shops that depend on used parts to hold down shop visit costs would feel that in their cost structure. Operators budgeting maintenance reserves based on current part cost assumptions may need to reassess.

The scale of the effect is the critical variable. The global retirement wave for first-generation CFM56 aircraft is substantial. How many engines ultimately find data center buyers versus teardown buyers will determine whether the impact on parts availability is marginal or significant. Industry analysts and parts brokers will be tracking transaction patterns closely over the next several years.

The Environmental Dimension

An adapted turbofan running on natural gas for data center power is a combustion source. It produces CO2. Whether that represents a better or worse environmental outcome than the alternative power source depends on local grid mix - in many markets, behind-the-meter gas generation displaces grid power that itself includes coal or legacy gas generation.

What’s clearer from a materials standpoint is that extending the working life of existing hardware delays the smelter. The nickel superalloys in the hot section, titanium compressor blades, and precision-machined structural components in a CFM56 represent enormous manufacturing energy. Using that hardware in a second application is, from a narrow materials perspective, a more efficient use of what it took to build it. Whether the full lifecycle accounting comes out favorable is a question for energy systems analysts.

The historical parallel is the post-World War II surplus. Military aircraft and engines were sold off and repurposed in ways nobody planned for. Pratt & Whitney Wasp engines ended up in boats, experimental applications, and places far outside the original design intent. Scale and context differ, but the pattern is familiar. Aviation technology has a track record of finding its way into applications it wasn’t originally designed for.

Key Takeaways

Retired CFM56 engines - with over 33,000 produced, the dominant narrowbody powerplant of the past three decades - are being purchased intact by AI data center operators for stationary power generation rather than going to traditional teardown.
AI data center construction has outpaced grid infrastructure in key U.S. markets, creating demand for behind-the-meter generation with utility interconnection backlogs measured in years.
An intact engine sale to a data center buyer may yield more than the $2–4 million traditional teardown parts value, creating a new bidding dynamic for end-of-life assets that improves residual values for operators and lessors.
If engines bypass teardown in significant numbers, used serviceable material supply for CFM56 components could tighten, potentially raising MRO shop visit costs for the broader operator community.
The scale of this shift - marginal disruption versus genuine market impact - depends on how many engines ultimately find power generation buyers over the next several years, and that number is not yet known.