Some AI data centres are generating their own electricity with jet engines pulled from retired aircraft and rebuilt to burn natural gas, because they cannot wait years for a connection to the grid. The New York Times reported the trend on Tuesday.
"It's all about getting power as fast as possible," said Rystad Energy analyst Audun Martinsen. With converted jet engines, "they can have a data centre running in 24 months."
Watch: news reports on the trend in the US…
Insights on data centres using jet engines
The waits are long. Connecting 100 megawatts to the grid in Northern Virginia, the world's biggest data centre market, takes about seven years, according to the International Energy Agency, and new power plant turbines are also scarce, with one major manufacturer saying most new capacity will only arrive by 2031. The delays are already costing deals.
Last month Microsoft pulled out as the anchor customer for Nscale's data centre in Loughton, England, after the local utility said it could not supply enough power until the early-to-mid 2030s. Elon Musk's xAI has installed eight converted jet-engine turbines in Mississippi to help power its Colossus 2 data centre in Memphis, the Times reported, using units from ProEnergy, which rebuilds a GE engine commonly found on Boeing jets.
Crusoe, which is building OpenAI's first Stargate site in Abilene, Texas, ordered 13 of ProEnergy's units in April. In September, it walked away from a $1.25-billion deal, about R20.6-billion, to buy turbines from Boom Supersonic, a startup building a supersonic passenger jet that has started selling a version of its engine as a power plant.
The trade-offs are real. Converted jet engines are less efficient and produce more pollution than turbines designed for gas power plants, and more of them are needed to make the same amount of electricity. Some experts expect efficiency and environmental concerns to matter more within a year or two, which may make them a short-term fix.
What others are saying about data centres using jet engines
TechCrunch's Aditya Mehta reported the trend from the Times' findings and described it as another sign of how far power supply has fallen behind the AI buildout. The New York Times identified xAI's turbines using state permits and satellite images. The International Energy Agency's report on energy and AI sets out the grid connection delays.
Power is the real bottleneck
The AI race is increasingly a race for electricity. Companies will burn gas in old aircraft engines, put chips in space and rename data centres to win communities over, because without power, the most advanced chips are useless.
That flips the usual story about South Africa. We are used to thinking of power as our weakness, but Eskom is selling less electricity than it did 20 years ago, while big markets abroad are queuing for years. Where spare capacity and grid connections exist, that is a selling point for data centres, and for the kind of factory-built data centres African founders could make.
The catch is the grid itself. As we found with South Africa's onshore wind, transmission, not generation, is often what holds projects back here. Fix the connections, and power could become one of South Africa's advantages in the AI economy rather than a liability.
You might also like our piece on OpenAI's AI compute campuses, Google's space data centre, and Eskom ruling out price cuts.
Get more SA tech and business news and subscribe to The Open Letter.


