DEEP RESEARCH · DATA-CENTER POWER

Data-Center Power Bottlenecks: Demand for Aircraft-Engine Generation

Why aeroderivative gas turbines are rising as AI data centers prioritize speed over efficiency

Date: 2025-12-27 · AI infrastructure power-supply analysis · Naver Blog source and TNBfolio

0. Bottom line first

Official fact: The source links to TNBfolio Telegram and fact-checks grid-connection delays, use of aircraft-engine-based generators, and the Sam Altman / Boom Supersonic discussion.

1. The core problem: time, not cost

From a traditional power-engineering perspective, mounting aircraft engines on the ground as generators looks inefficient. Large combined-cycle gas plants can deliver thermal efficiency above 60% and low generation cost, while aircraft-engine-based generation has lower efficiency and higher maintenance cost.

Interpretation: Even so, time efficiency now dominates cost efficiency in the data-center market. Data-center buildings can be ready for servers in 18 to 24 months, while large gas plants require at least four to five years for site selection, environmental review, construction, and commissioning.

2. Technology comparison: heavy-duty vs aeroderivative

Gas turbines fall into two broad groups: heavy-duty turbines used in conventional power plants and aeroderivative turbines adapted from jet engines for ground power generation.

Official fact: The source describes GE’s CF6-based LM6000 and LM2500 models and says LM2500XPRESS is shipped 95% factory-assembled and can be installed on site in two weeks.

3. Power architecture: hybrid microgrid

Modern data centers are moving toward behind-the-meter generation rather than relying only on external grids. In normal operation, they draw from the grid or fuel cells. During peaks or emergencies, aeroderivative gas turbines produce power within minutes. BESS and UPS bridge the two-to-five-minute gap before turbines are online.

Official fact: The source says fuel cells can be highly efficient, above 60%, but have difficulty reacting instantly to load changes. Aeroderivative turbines are described as starting within five minutes to cover AI workload spikes or grid instability.

4. Global supply chain: who makes the engines

The source frames aircraft-engine technology as concentrated around Big 4 players because of precision and durability requirements: GE, Rolls-Royce, Pratt & Whitney, and Safran. In derived power-generation markets, GE Vernova, Siemens Energy, Mitsubishi Power Aero, and Rolls-Royce Power Systems are emphasized.

5. Korean companies: Hanwha Aerospace and Doosan Enerbility

Korean companies have limited independent aircraft-engine design capability, but they participate through partnerships and manufacturing capacity with global Big 4 players.

• Hanwha Aerospace: Korea’s only aircraft-engine manufacturer and an RSP partner to GE, P&W, and Rolls-Royce. It supplies core parts such as blades and cases and performs licensed production.
• Hanwha Power Systems: Works on gas turbine retrofit and hydrogen co-firing, and with GE on LM2500 marine gas-turbine packaging and modularization. The source sees potential expansion into data-center generation modules.
• Doosan Enerbility: Not aircraft-engine based, but independently developed the DGT6-300H heavy-duty gas turbine as the fifth company in the world to do so. A recent US data-center gas turbine supply contract targets large campus data-center demand.

6. Outlook: the bridge-power era

In the long term, large gas plants can be the better answer. But data centers that need power right now cannot accept five-to-seven-year construction and transmission waits. Aeroderivative generators can be moved by trailer and produce electricity in two to three months where grid connection is missing.

Interpretation: The next five to ten years may be a bridge-power era for data-center electricity. Until grids expand or SMRs commercialize, aeroderivative gas turbines and fuel cells can fill the gap. For investors, the key is which need each company’s portfolio satisfies: speed, scale, or lower-emission power.