Retired Aircraft Engines Are Powering AI Data Centres

In a development that looks more like a stopgap than a long-term strategy, operators of AI data centres have begun stripping down decommissioned aircraft engines, converting them to run on natural gas, and deploying them as standalone power generators. The reason is simple and direct: demand for electricity has outpaced the grid's ability to respond, and waiting for an official connection is now measured in years rather than months.

Why jet engines specifically?

The decisive advantage here is speed. According to estimates from energy sector analysts, a data centre relying on generation units built from recycled aircraft engines can be up and running in roughly 24 months. By contrast, connecting 100 megawatts of capacity to the grid in Northern Virginia — one of the world's densest data centre clusters — takes nearly seven years, according to International Energy Agency estimates.

The bottleneck isn't limited to the grid; it extends to the supply chains for industrial turbines themselves. Statements from executives at major power generation equipment manufacturers have indicated that most of the new capacity required won't be available before 2031, leaving a wide timing gap that companies are scrambling to fill by any means available.

The cost of delay: contracts evaporating

Delays in securing electricity are no longer merely an engineering problem — they have translated into tangible commercial losses. One of the most prominent examples is Microsoft's withdrawal from a key customer site at a data centre project in the UK, after the local utility advised that it would be unable to supply the required power before the early-to-mid part of the next decade. Decisions like this make clear that power has become the determining factor in a project's viability — not the availability of land, chips, or capital.

Who is actually using these solutions?

• xAI: Reports based on official permits and satellite imagery indicate the installation of eight generation units built from repurposed aircraft engines in Mississippi, supporting operations at the "Colossus 2" data centre in Memphis. The units come from ProEnergy, which rebuilds a General Electric engine widely used on Boeing aircraft.
• OpenAI: Has likewise turned to gas-fired turbines at the first site of its "Stargate" project in Abilene, Texas.
• Crusoe: The startup specialising in building AI data centres, which is constructing the Abilene campus, has ordered 13 converted units from ProEnergy. In September it also walked away from a $1.25 billion deal to purchase 29 gas turbines from Boom Supersonic, while confirming it will continue relying on turbines more broadly.

Notably, Boom Supersonic — a company developing a supersonic passenger aircraft — has begun marketing a version of its aircraft engine as a power generation unit for data centres, a sign that the market for fast-deployment power has become attractive to manufacturers from outside the traditional sector.

A clear trade-off: speed versus efficiency

This approach comes at a price. Converted jet engines are less efficient than turbines designed from the outset for conventional gas plants, they produce higher emissions, and they require a greater number of units to generate the same amount of electricity. That means higher fuel, operating, and maintenance costs over the medium term, along with a heavier environmental burden on surrounding sites.

For that reason, a number of experts view this option as closer to a transitional fix: a way to bridge the current waiting period, with efficiency and environmental considerations rising in importance over the next year or two and pushing companies toward more sustainable alternatives — from modern turbines to renewable power contracts and battery storage, and perhaps small-scale nuclear solutions further down the line.

What this wave means for the sector

The most important lesson is that energy infrastructure has become the real bottleneck for AI expansion, after years in which the conversation centred on chip supply. With policy support encouraging data centres to generate their own electricity, the industry is moving toward a "behind-the-meter power" model that offers greater independence but carries heavier operational and environmental responsibility.

For companies planning compute-intensive projects in the Arab region, the message is practical: planning for the power source must begin in parallel with designing the facility itself, not after. Whoever secures electricity first wins the race — while those waiting their turn in the queue may find the customer has already moved on.

✦ بقلم فريق دروب أيديا