Bottom Line
Korean ship engines have entered the US data center power supply chain for the first time — HD Hyundai's HiMSEN has emerged as the core bridge-power axis amid gas turbine and transformer bottlenecks.
Reading depth
- 1. Why Now — Three Simultaneous ShortagesAs grid interconnection delay becomes a bigger bottleneck than GPU supply, immediately deployable ship engines and fuel cells gain a structural advantage.Jump to section
- 2. Demand Scale — What a 1GW Data Center Really MeansUS data center power demand rises far beyond current supply-chain capacity under every scenario.Jump to section
- 3. HiMSEN — Korea's Only Proprietary 4-Stroke Medium-Speed EngineHiMSEN combines fast start, load-following, and clean-fuel extensibility, giving it a structural edge over gas turbines in always-on data centers.Jump to section
- 4. AEG 684MW Deal — Dissecting 2026.4.22HD Hyundai's first data center order is 1.66x the scale of Wärtsilä's single deal that same month, reflecting HiMSEN's large-unit positioning.Jump to section
- 5. Landscape of Korea's 4 Engine Makers — Role DifferentiationKorea's four engine makers differentiate by role, and Hanwha Engine's benefit is a mid-term story tied to its 4-stroke pivot and group energy transition.Jump to section
- 6. Scenario Analysis — Trigger-BasedThe scenarios split by triggers without odds. HD Hyundai's cumulative 1 GW break in 2026 and follow-on DC wins are the key markers across Bull, Base, and Bear.Jump to section
- 7. Risk & Deal RadarThe biggest risk is an AI capex overshoot correction. HiMSEN capacity is a near-term constraint, and price competition like Caterpillar's 4 GW deal pressures margins.Jump to section
- 8. Conclusion & Action PointsThe theme's essence is not ship engines replacing gas turbines but being the core axis of 2027-2029 bridge power. Cashing in wins in that window sets the valuation.Jump to section
1. Why Now — Three Simultaneous Shortages
AI data center power is not simply about "not enough electricity." The three axes of supply-chain infrastructure — generators (gas turbines), transmission (transformers), and timing () — are simultaneously in bottleneck, creating a structure where AI buildout speed is tied to infrastructure speed.
1-1. Gas Turbine Shortage — GE Vernova Backlog Sold Out Through 2029
Large gas turbines are the core equipment for combined-cycle power, historically the main source for DCs. All three global leaders (GE Vernova, Siemens Energy, Mitsubishi ) are in extended backlog due to DC-driven orders.
Per GE Vernova's Q4 2025 release, Gas Power segment backlog surged from 29GW (end-2024) to 83GW (end-2025). CEO Scott Strazik expects slots through 2030 to be fully booked by end-2026. That's 3–4 years of pre-bookings against annual capacity of 24GW (2028 target).
1-2. Transformer Shortage — Transmission Is Clogged Too
Even if you generate electricity, you need transformers to carry it. Large power transformer (LPT) lead times extended from 24–30 months (pre-2020) to 36–60 months (3–5 years) today. Roughly half of planned US DC capacity is being delayed or canceled due to transformer/equipment shortages. HD Hyundai Electric and Hyosung Heavy's North American HV transformer exports benefit directly.
1-3. Time-to-Power — Time Is Money
Oracle CEO K.R. Sridhar has called "Bring your own power" the new DC mandate. For hyperscalers, grid interconnection delay is a bigger bottleneck than GPU supply. Ship engines and fuel cells now hold a decisive advantage with "delivery within 1–3 years.