HomeArticle

Retired aircraft engines have been "rehired" by Elon Musk.

字母AI2026-11-19 16:26
Decommissioned aero engines have become hot commodities in the AI industry.

Among the power supply equipment for AI data centers, there are aeroengines dismantled from decommissioned passenger planes.

After refurbishment and modification, these engines are reassembled into gas turbine power generation equipment and put back into operation. They once powered aircraft at an altitude of 10,000 meters, and now they have transformed into the power guarantee for ground data centers.

A large number of decommissioned military aircraft are parked in the desert at the aircraft storage site of Davis-Monthan Air Force Base in Arizona, USA. This site was also used by the U.S. Energy Information Administration to measure the potential scale of power generation transformed from decommissioned aircraft engines.

Musk's SpaceXSI (formerly xAI) has installed 8 such generating units.

These devices are provided by the American energy equipment company ProEnergy, using the core of aeroengines produced by General Electric for large passenger aircraft such as Boeing 747. The U.S. data center developer Crusoe also placed an order subsequently, ordering 13 units at a time with a total capacity of 650,000 kilowatts. China's Yantai Jereh Group also participated in this business, cooperating with the U.S. FTAI to transform decommissioned engines used by Boeing 737 and Airbus A320 into power generation equipment.

Decommissioned engines have suddenly become hot commodities, behind which is an increasingly fierce power competition.

Re-employment of Decommissioned Engines

The American energy equipment company ProEnergy can be regarded as an active recruiter to "find jobs" for decommissioned aircraft engines.

ProEnergy has set its sights on a group of "veteran employees", which are the core components of the General Electric CF6-80C2 engine. This engine was once widely used in Boeing 747.

The GE CF6-80C2 aeroengine was once widely used in Boeing 747 passenger aircraft. Based on the core components of this engine, ProEnergy produces PE6000 generator sets after refurbishment and transformation.

However, to make these engines, which were originally responsible for propelling aircraft, drive generators instead, a series of modifications are still required. ProEnergy disassembles them for inspection and overhaul, then equips them with new power turbines, generators and control systems, allowing these veteran employees to change positions and continue doing their old jobs.

The working principle of an aeroengine is to convert fuel into mechanical power. After air enters the engine, it is first compressed, then mixed with fuel for combustion, and the generated high-temperature gas pushes the turbine to rotate.

On an aircraft, the turbine drives the fan to generate thrust. After being transformed into power generation equipment, the high-temperature gas will push a dedicated power turbine, which then drives the generator to convert mechanical power into electricity.

In fact, there have long been mature products that apply aeroengine technology to ground power generation.

LM6000 under the U.S. energy giant GE Vernova is one of them. It is technically based on the CF6-80C2 aeroengine, has sold more than 1,500 units, and has a cumulative operating time of about 60 million hours.

However, GE Vernova produces specially developed industrial gas turbines, while ProEnergy renovates and reuses the cores of decommissioned aeroengines. One is to use aeroengine technology to develop new equipment, and the other is to rehire old engines back to work.

ProEnergy has turned this transformation plan into the PE6000 generator set, with a gross output power of about 50 megawatts per unit and a start-up time of only 5 minutes. Musk's xAI has installed 8 units.

U.S. data center developer Crusoe also placed an order with ProEnergy, ordering 13 units at a time in April 2026 with a total capacity of 650,000 kilowatts, scheduled to be delivered before the summer of 2027. For ProEnergy, the recruitment of these veteran employees is obviously not over yet.

Jereh Group from Yantai, China, has also cooperated with the U.S. aeroengine leasing company FTAI to prepare for another batch of aeroengines to return to work. The selected model this time is CFM56, which was once widely used in Boeing 737 and Airbus A320. FTAI is responsible for the engine transformation technology, and Jereh is responsible for the design, manufacture and integration of the power generation system. The two sides have made the whole set of equipment into a transportable and rapidly deployable mobile generator set.

The Mod-1 mobile aero-derivative gas turbine power generation equipment launched by FTAI Power adopts the CFM56 engine transformation scheme, with a designed output power of 25 megawatts. The picture is the product renderings in the official investor materials of the enterprise, not the actual delivery photos.

In July 2026, J&F Power Systems, a joint venture established by the two parties, signed a five-year supply framework agreement with an international cloud service provider, and received the first purchase order of 1.465 billion US dollars. The equipment is planned to be delivered in batches before November 2027.

An Energy Rush Triggered by AI

The fundamental reason why so many decommissioned engines suddenly have new jobs is that the AI industry is extremely short of electricity.

Servers in data centers need to operate 24 hours a day, especially when training large AI models, thousands of GPUs work at the same time, and the power consumption is quite staggering.

A report released by the Lawrence Berkeley National Laboratory in June 2026 estimated that U.S. data centers consumed about 192 terawatt-hours of electricity in 2024, accounting for 4.7% of the total U.S. electricity consumption. By 2030, this figure is expected to reach 649 terawatt-hours, accounting for 11.8% of the total.

If calculated according to the high-growth scenario, it may even reach 843 terawatt-hours, accounting for 15.3% of the total U.S. electricity consumption. Among them, AI servers will become one of the main driving forces for growth.

The problem is that AI data centers can be built in a very short time, but supplying power to them is not so easy.

Building a new power plant requires approval and construction. To deliver electricity to data centers, it is also necessary to lay transmission lines, build substations, and go through a long review process to connect to the grid.

According to statistics from the Lawrence Berkeley National Laboratory in the United States, in regions with complete data, for power projects that were put into commercial operation in 2025, the median time from applying for grid connection to official operation was 61 months, more than 5 years.

AI companies are not willing to let servers worth billions of dollars wait in the computer room for so long.

The data center built by Musk's AI company xAI in the Memphis area of the United States. As the AI training cluster continues to expand, data centers need to continuously obtain large-scale power supply, and on-site power generation has also become one of the ways for enterprises to meet their power demand.

As a result, data center developers began to find electricity on their own. Building new gas power plants takes time, and orders for traditional gas turbines are also queuing up. In this way, enterprises turned their attention to aeroengines. After all, these engines have already been manufactured, and the core components can be reused after inspection and overhaul, and the transformed equipment can also be directly deployed near the data center.

For companies that are in a hurry to obtain electricity, this batch of "retired employees" comes just in time.

Even the real "aircraft graveyard" has been taken to measure its power generation potential.

In January 2026, the U.S. Energy Information Administration studied the military aircraft storage site at Davis-Monthan Air Force Base in Arizona, where a large number of decommissioned military aircraft are parked.

Researchers estimated that if all the engines on these aircraft are transformed into power generation equipment, the theoretical maximum installed capacity can reach 40,000 megawatts, which is equivalent to about 110% of Arizona's existing power generation capacity.

How Long Can Old Engines Help Ease the Urgency

After seeing the 40,000 megawatt estimate, it is easy to feel that there are so many decommissioned planes parked in the desert, and AI companies can dismantle as many engines as they need.

When you actually arrive at the aircraft graveyard, you may not even see the shadow of the engine.

Aviation maintenance companies have long been collecting these old engines. About an hour's drive from the aforementioned air force base, there is the Pinal Airpark, which is dedicated to recycling decommissioned passenger planes. The account manager there once said that as soon as the plane arrives, the first thing to be dismantled is the engine. In many cases, the engine on an aircraft is more valuable than the rest of the fuselage.

Even if the whole machine can no longer fly, the good parts removed can still be used to repair other planes. Now that companies making power generation equipment for data centers have also taken a fancy to the same batch of goods, the fastest one gets it.

The CEO of FTAI calculated an account in February this year. According to his estimate, about 400 CFM56 engines are dismantled for parts every year in the world. As long as 100 units are purchased before dismantling, it is enough for the company's power generation business for one year. According to his algorithm, to meet the demand of FTAI alone, a quarter of them have to be bought.

After finally grabbing the goods, you have to buy a few more units.

Take the aforementioned unit transformed from the Boeing 747 engine as an example, one unit can generate 50 megawatts. A data center that requires 1 gigawatt of power, with 20 units running together, is theoretically just enough, but this obviously has no fault tolerance at all.

The International Energy Agency estimates that for such a large data center to rely on gas turbines for continuous power supply, the installed capacity may need to reach 1.3 to 1.7 gigawatts. Roughly calculated at 50 megawatts per unit, that is 26 to 34 units. The extra units are all real money. Even if only 6 more units are installed, according to the construction cost estimate for such units in the report, the additional investment will be about 750 million US dollars.

Another challenge is where the energy required by such engines comes from. Data center developers have already started to select sites along natural gas pipelines. CloudBurst in the United States has discussed with the pipeline company Energy Transfer that when selecting sites for computer rooms in the future, the two parties will work together to find locations along or near gas transmission pipelines.

In addition, the emission pollution of gas turbines is also a long-standing problem.

Musk's xAI has installed 8 units transformed from aeroengines in Mississippi. The local permit documents show that the upper limit of the permitted carbon monoxide emission rate of these 8 units is twice that of other units at the same site.

The Lincoln Gas Turbine Power Station of Duke Energy in the United States. Gas turbines generate carbon dioxide and nitrogen oxides during operation. When data centers adopt on-site gas power generation, they also need to face environmental permits and pollutant emission issues.

For the batch of servers waiting to be powered on at the moment, the engines dismantled from the Boeing 747 can indeed provide great help. However, the old engines suitable for modification in the aircraft graveyard are getting fewer and fewer as each one is removed. When all these ready-made engines have found buyers, when the next computer room lacks electricity, it is impossible to urge airlines to replace their planes earlier.

This article is from the WeChat official account "Alpha AI", the author is Jin Xiaoyá, and 36Kr is authorized to release it.