The Power Crunch for AI Data Centers
As the artificial intelligence boom accelerates, power supply is struggling to keep pace with the massive energy demands of new data centers. According to reports from the New York Times cited by TechCrunch, companies are facing years-long waiting lists for grid connections and new traditional turbines. Hooking up 100 megawatts to the grid in Northern Virginia, for example, takes approximately seven years, as stated by the International Energy Agency. Mitsubishi Power executives note that most power capacity is not expected to come online until 2031.
These severe delays are already impacting business operations. Last month, Microsoft withdrew as the anchor customer for a data center in Loughton, England, after the local utility warned it could not deliver sufficient power until the early-to-mid 2030s.
Deploying Converted Aircraft Engines
To bypass grid delays, several tech companies are turning to temporary, self-generated electricity using aircraft engines pulled from retired planes and rebuilt to burn natural gas. According to Rystad Energy analyst Audun Martinsen, the primary advantage is speed, allowing a data center to become operational within 24 months.
Elon Musk's xAI has utilized this approach, installing eight converted jet-engine turbines in Mississippi to support its Colossus 2 data center in Memphis, based on state permits and satellite imagery cited in reports. These units were sourced from ProEnergy, a company that rebuilds General Electric engines commonly found on Boeing aircraft.
OpenAI has followed a similar path for its first Stargate site in Abilene, Texas. Crusoe, a Denver-based AI infrastructure startup constructing the Abilene campus, announced a contract with ProEnergy to acquire 13 PE6000 aeroderivative gas turbine generator sets, with each unit capable of generating 50 MW to deliver a 650-MW power generation solution by summer 2027. Additionally, reports noted plans involving GE Vernova LM2500XPRESS gas turbines capable of producing 34 MW each for the site.
Trade-Offs and Short-Term Solutions
While aeroderivative gas turbines offer quick setup times and the flexibility to adjust output rapidly to maintain continuous power, they come with significant drawbacks. Converted jet engines are less efficient and generate more pollution than turbines specifically designed for conventional gas plants, requiring more units to produce an equivalent amount of power. Experts suggest these environmental and efficiency concerns may cause companies to treat them strictly as a short-term fix.
Other alternatives, such as small modular nuclear reactor (SMR) technology pursued by major tech companies including Microsoft, Meta, Google, and Oracle, are still under development and generally not expected until the 2030s. Meanwhile, government pressure, such as encouragement from the Trump administration, continues to push data center operators toward generating their own electricity.