Grid Queues Too Long: Data Centers Turn to Retired Jet Engines for Power

AI compute demand is surging, but data centers are stuck at an unexpected bottleneck: waiting for power. According to RSS sources, the timelines for grid interconnection and procuring new gas turbines have stretched to years, and builders are starting to look at an unconventional solution—retired jet engines.

A Power Solution Born of Desperation

Data center power demand is growing far faster than the grid can expand. New transmission lines, substations, and large gas turbines all require lengthy approval and construction cycles, and grid interconnection queues often run in years. For compute projects racing to go live, that wait is tantamount to失控 opportunity cost.

Retired jet engines have thus come into view. These machines originally belong to the aviation sector; after retirement, they can be converted for use as generator sets. Compared with waiting for new turbine deliveries, old engines can be acquired faster and fill the power gap sooner. The report describes it as an "unusual, less efficient, and less environmentally friendly" solution—these three qualifiers themselves speak to its emergency nature.

Why Old Engines

Jet engine power generation is not entirely new. Aero-engine-derived ground gas generators have already been used in industry, with advantages including high power density, fast startup, and relatively flexible deployment. For loads like data centers that need continuous, stable power, paralleling multiple units can compensate to some extent for single-unit reliability issues.

The real driver is still time. New gas turbine manufacturing and delivery face clear bottlenecks, and grid interconnection queues are an even more uncontrollable variable. Old engines essentially bypass these two hurdles, trading existing equipment for time to go live. The cost shows up in fuel efficiency and emissions: older models have higher energy consumption per unit of power generated, and their nitrogen oxide emissions performance is worse than that of new units.

Converting aero-engines into ground gas generators already has a mature engineering path, usually called "aeroderivative gas turbines." The core approach is to retain the core engine section of a retired turbofan or turbojet, remove the fan and bypass structure, and add a power turbine, reduction gearbox, control system, and power generation module, so that it shifts from producing thrust to outputting shaft work to drive a generator. The typical feature of such units is extremely high power density—at the same output power, their size and weight are far smaller than industrial heavy-duty gas turbines, so they can be containerized and vehicle-mounted. Startup time is usually on the order of minutes, making them suitable as peaking or emergency power sources. However, aero-engines are designed for high altitude, short duration, and high thrust; when converted for continuous ground power generation, both thermal efficiency and service life suffer, overhaul intervals are often shorter than those of dedicated generator sets, and maintenance costs are higher. In addition, aeroderivative units are more sensitive to fuel quality and usually run on aviation kerosene or light diesel; switching to natural gas requires an additional fuel treatment system.

A Double Compromise on Efficiency and Environment

The controversy around this route is direct. In recent years, the data center industry has invested considerably in renewable energy procurement and energy efficiency metrics, and turning to old aviation engines for power forms a sharp contrast. The report characterizes the choice as "less efficient and less environmentally friendly" without giving specific emissions data, but the direction is clear.

From a system perspective, this looks more like a transitional arrangement. When grid expansion cannot keep up with compute expansion, temporary power becomes a necessary evil. The problem is that once temporary solutions are put into use, they often have their lifespans extended due to sunk costs and continued demand pressure, turning a short-term efficiency compromise into a medium-term emissions burden.

Key metrics for measuring the environmental performance of generator sets include CO2 emissions per unit of power generated and nitrogen oxide concentration. Modern heavy-duty gas turbine combined cycles can achieve efficiency above 60%, while older aeroderivative units in simple cycle typically have efficiency of only 30% to 40%, meaning the same amount of electricity requires burning more fuel and emitting more CO2. On nitrogen oxides, new units generally come with dry low-emission combustors that keep emissions at low levels; converted old engine units often retain original combustor designs, with noticeably higher NOx concentrations, and some models require selective catalytic reduction systems to meet local environmental permits. This is also why such solutions may face resistance in regions with stricter environmental approvals, while being easier to implement under relatively lax regulation or emergency exemption mechanisms.

Power Is Becoming the Real Bottleneck for Compute

The significance of this goes beyond a single technology choice. It shows that the focus of competition in AI infrastructure is shifting from chips to power: GPUs can be bought at a premium, but electricity cannot be conjured out of thin air. Every year spent in the interconnection queue is a direct delay in project go-live time.

For data center operators, power access capability is becoming a siting and expansion variable as important as capital and land. Old jet engines are just one of many emergency measures; the fact that they are being discussed at all reflects the urgency of the supply-demand gap. When the wait for new turbines becomes unacceptably long, the industry will look for imperfect but usable substitutes.

The interconnection queue is the key mechanism for understanding this bottleneck. In most power markets, new or expanded electricity-consuming facilities must submit interconnection applications to grid operators and go through multiple processes—feasibility studies, system impact assessments, and determination of upgrade plans—before obtaining interconnection approval. In recent years, with a flood of renewable energy and storage projects plus surging data center loads, queues in many regions have become severely backlogged, and in some markets (such as PJM and ERCOT in the U.S.) queue timelines have stretched to several years. Projects in the queue must also bear allocated grid upgrade costs, and there is a phenomenon of withdrawal after queuing, further slowing overall progress. For data centers, this means that even if the building and servers are ready, they may be unable to energize and operate because they cannot obtain interconnection approval—power has thus shifted from a back-end supporting factor to a front-end constraint that determines project pace.

One-line Summary

Grid interconnection and turbine delivery timelines run in years, so data centers are turning to retired jet engines for emergency power.

Paragraph Summary

The article notes that surging AI compute demand has put data centers under a severe power access bottleneck: approval and construction cycles for new transmission lines, substations, and gas turbines are lengthy, and grid interconnection queues often run in years. To go live as soon as possible, builders are converting retired jet engines into generator sets, using their short acquisition timelines, high power density, and flexible deployment to fill the power gap. But this solution is less efficient and has worse emissions, making it an emergency transitional measure. The article argues that this reflects a shift in the focus of AI infrastructure competition from chips to power, and that power access capability has become an expansion variable as important as capital and land.