The Secret Power Source: Data Centers Repurposing Old Jet Engines

Stuck in years-long lines for grid connections and new turbines, some AI data centers are making their own electricity with converted jet engines, the New York Times reported Tuesday.

Key Takeaways

• **Urgent Power Demand:** The explosive growth of AI is severely outstripping traditional energy grid capabilities, leading to multi-year delays for data centers seeking conventional power connections.
• **Jet-Powered Solution:** To circumvent these bottlenecks, major AI players like xAI and OpenAI are deploying repurposed aircraft jet engines, converted to burn natural gas, for rapid on-site electricity generation.
• **Short-Term Fix, Long-Term Questions:** While offering unprecedented speed to market, these off-grid solutions come with significant trade-offs in efficiency and environmental impact, posing sustainability challenges for the future of AI infrastructure.

The relentless march of artificial intelligence demands an equally relentless supply of power. But as the silicon brains of tomorrow scale at exponential rates, the very foundation of their existence – electricity – is proving to be a formidable bottleneck. Imagine building a hyper-speed train only to find the tracks won’t be laid for another decade. That’s the conundrum facing AI data centers today, pushing them towards an unconventional, high-octane solution: converting retired jet engines into on-site power plants.

This startling development, first highlighted by the New York Times, underscores a critical disconnect between the dizzying pace of AI innovation and the glacial speed of traditional energy infrastructure development. Companies eager to bring their computational behemoths online are staring down multi-year waits for grid connections and new, purpose-built turbines. The result? A scramble for any power source that can be deployed quickly, even if it means harnessing the roar of a Boeing engine.

The Need for Speed: Why Jet Engines?

For companies stuck in those interminable lines, the allure of converted jet engines is their unparalleled speed of deployment. “It’s all about getting power as fast as possible,” Rystad Energy analyst Audun Martinsen told the Times. He elaborates on the stark contrast: with jet-powered units, “they can have a data center running in 24 months.” This two-year timeline is a game-changer in a market where every month of delay means lost revenue, competitive disadvantage, and deferred innovation.

Compare this to conventional grid connections. Hooking up a substantial 100 megawatts to the grid in key data center hubs like Northern Virginia, for example, is an arduous process that can take about seven years, according to the International Energy Agency. Furthermore, an executive at Mitsubishi Power, a leading manufacturer of power generation equipment, recently told Bloomberg that most new power capacity isn’t expected to come online until 2031. This chasm between AI’s immediate power needs and the grid’s long-term planning horizon creates a desperate void that these rapidly deployable jet engines are designed to fill.

The Cost of Waiting: Business Lost

These grid delays aren’t just an inconvenience; they’re already costing data center operators significant business. A stark example emerged last month when Microsoft, a colossus in the cloud and AI space, withdrew as the anchor customer for Nscale’s data center in Loughton, England. The reason? The local utility couldn’t guarantee sufficient power delivery until the early to mid-2030s. Such setbacks highlight the immense financial and strategic pressure on data center providers and their AI clients to find alternative power solutions, no matter how unorthodox.

Who’s Flying Solo? Early Adopters and Their Strategies

The race to self-power has already seen some high-profile players take the plunge. Elon Musk’s xAI is at the forefront, having reportedly installed eight converted jet-engine turbines in Mississippi. These units are crucial for powering its ambitious Colossus 2 data center in Memphis, as revealed by state permits and satellite imagery cited by the Times. The engines themselves come from ProEnergy, a specialist firm that rebuilds General Electric engines commonly found on Boeing aircraft, repurposing them for stationary power generation.

Following in Musk’s footsteps, OpenAI, another titan in the AI realm, is planning to utilize gas turbines for its inaugural Stargate site in Abilene, Texas, slated for 2025. This massive undertaking is being constructed by Crusoe, a Denver-based startup specializing in building AI data centers. Crusoe has already placed an order for 13 of ProEnergy’s converted units, signaling a significant investment in this distributed power strategy.

Crusoe’s journey into self-generated power hasn’t been without its twists. Last September, the company notably walked away from a hefty $1.25 billion deal to procure 29 natural gas turbines from Boom Supersonic. Boom, a startup developing supersonic passenger jets, had also branched into selling a version of its jet engine as a power plant for data centers. While that particular deal fell through, a Crusoe spokesperson confirmed to TechCrunch at the time that the company remained committed to using turbines for its operations, underscoring the strategic importance of on-site power for their business model.

The Broader Context and Future Outlook

The pressure on data centers to generate their own electricity isn’t solely market-driven. In the United States, there’s also a political dimension, with the Trump administration actively encouraging data centers to pursue self-generation capabilities. This push could be seen as part of a broader energy independence agenda, further accelerating the trend towards distributed power solutions for energy-intensive industries.

However, this expedient solution comes with significant caveats. Converted jet engines, while fast to deploy, carry inherent trade-offs. They are generally less efficient than turbines specifically designed for conventional gas power plants, meaning they consume more fuel to generate the same amount of electricity. This reduced efficiency translates directly into higher operational costs and, crucially, a larger environmental footprint. They also pollute more than their purpose-built counterparts, raising concerns about emissions and air quality, particularly as AI companies increasingly emphasize their commitment to sustainability.

The fact that it takes more of these repurposed units to achieve the same power output as a single, modern conventional turbine further compounds the issues of land use, noise, and maintenance. These factors lead many experts to view converted jet engines as a short-term play, a necessary evil to bridge the current power gap. While the urgency of getting AI infrastructure online currently outweighs these concerns, it’s widely anticipated that efficiency and environmental impact will grow significantly in importance within the next year or two, forcing a reevaluation of these stop-gap measures and accelerating the search for more sustainable long-term solutions.

Bottom Line

The reliance on repurposed jet engines for AI data centers is a vivid illustration of the energy crisis brewing beneath the surface of the tech boom. It highlights a critical inflection point where the sheer scale and speed of technological advancement have outpaced the foundational infrastructure meant to support it. While these roaring turbines offer a quick fix to an immediate, pressing problem, they simultaneously underscore the urgent need for innovative, scalable, and sustainable energy solutions to truly power the future of artificial intelligence. The current strategy is less about efficiency and sustainability and more about pure, unadulterated speed, a temporary measure that sets the stage for a much larger and more complex energy transformation in the years to come.