Key Takeaways

  • Boom Supersonic and its aviation peers utilize jet-engine technology that is now being adapted to power data centers.
  • Developer Crusoe is deploying gas turbines for data centers supporting OpenAI as the sector seeks rapid power solutions.
  • Developers are accelerating "behind the meter" natural gas generation as utilities face long interconnection delays.
  • Experts warn that rapid gas expansions tied to AI growth could lock in high emissions for years.

Boom Supersonic is best known for its ambition to build the first commercial supersonic airliner. Yet the engineering principles behind high-speed aviation are now intersecting with an entirely different market: utilizing jet-engine turbines to power data centers. The shift says a lot about how strained the energy landscape has become as artificial intelligence workloads surge. It also shows how far developers are willing to go to secure their own electricity when utilities cannot keep up.

In a significant industry development, developer Crusoe is securing jet-engine style gas turbines for data centers being built for OpenAI. At first glance, relying on aviation-grade technology might feel like an odd pairing. Jet engines were engineered for cruising at Mach speeds, not keeping server racks running. But the pressure for rapid, flexible, self-managed power has created an opening for these experimental solutions. The traditional manufacturers of large combined cycle turbines are now so backlogged that developers are turning to aeroderivative engines—essentially jet engines mounted on the ground—which are available on shorter timelines.

Other tech firms are moving even faster. Reports indicate Meta has explored plans to rely on mobile mini turbines at facilities in the El Paso region. Caterpillar has supplied gas engines for at least one West Virginia installation. Crusoe itself has utilized aeroderivative turbines at its massive facility in Abilene, where demand is projected to exceed 1.2 gigawatts. If that number feels eye-popping, it is worth remembering that a single data center cluster can now rival a mid-sized city in power use.

The broader trend is global. According to Global Energy Monitor, developers added more than 1,000 gigawatts of proposed natural gas capacity worldwide over the past year. The United States leads the expansion, and more than one-third of new US capacity is tied to the data center pipeline. Two-thirds of American developers have not yet identified who will build their turbines, which suggests continued supply chain improvisation.

Climate groups have been warning about exactly this outcome. Early on, many boosters claimed AI compute would run on renewables. Some of it has. Yet the current build-out is locking in long-life fossil assets at a pace no one expected. Cornell researchers estimate the added emissions could reach 44 million metric tons of carbon dioxide by 2030. That is roughly the same as 10 million passenger cars. The Sierra Club’s analysts have been tracking the numbers, noting that existing coal units and newly proposed gas plants are squeezing out cleaner alternatives.

The situation is compounded by interconnection delays. Utilities are overloaded with requests. Securing large new service commitments can take years. That has pushed developers toward behind-the-meter generation, meaning on-site power plants that do not connect to the broader grid. Cleanview counts at least 46 data centers using this approach with a combined 56 gigawatts of capacity. For comparison, the Hoover Dam produces about 2 gigawatts. The scale is difficult to wrap your head around.

Fuel cell vendors say demand for on-site systems has doubled. Their business used to cluster in states with high retail electricity prices. Now the fastest growth is in places with strong natural gas infrastructure and permissive policy frameworks. It is an interesting reversal because on-site power was once seen as a niche solution. Today it feels like a standard operating tactic.

Texas illustrates the shift more clearly than any other state. Almost 58 gigawatts of natural gas projects are in development there. That is more than the next four states combined. Nearly half of new Texas gas construction will serve data centers exclusively. Developers plan hundreds of modular generators at sites from Abilene to El Paso.

Some teams prefer smaller reciprocating engines rather than turbines. They are easier to ramp and can handle the sudden load spikes that AI training clusters produce. They also start up in about a minute. Turbines need closer to an hour. That comparison sounds useful, although researchers note that engines emit more carbon per unit of electricity because they are less efficient.

Rural areas are also seeing a wave of proposals. The BorderPlex region is reviewing a massive project called Jupiter in southern New Mexico, supported by two simple cycle microgrids. Simple cycle means they burn gas to turn a turbine and stop there. No waste heat recovery. The fuel efficiency gap is sizable. The facility would generate 2,880 megawatts, surpassing the main utility serving central New Mexico.

Community groups question whether Jupiter should even qualify as a microgrid. Because it is behind the meter, the project avoids regulatory review that would typically apply under the state’s climate laws. Critics argue the emissions could wipe out years of statewide climate progress. And then there is the market risk. If AI demand softens or capital spending slows, the region could be left with unused gas turbines that cannot legally operate. Stranded assets in the most literal sense.

There is a pattern emerging here. As artificial intelligence accelerates, power requirements scale even faster. Developers are racing to secure capacity in whatever form they can find. Natural gas has become the default solution because it is dispatchable, available, and familiar. The result is a rapid build-out with long-term consequences that are difficult to unwind later. It is not entirely clear how long this phase will last. But the decisions being made today could shape grid emissions for decades.