Analysis-AI power rush sends data centres hunting for used turbines and jet engines

Global gas-turbine orders exceed 100 GW against estimated annual production capacity of 60–70 GW; capacity is expected to reach about 87 GW in 2028, still below projected demand of 129 GW. US delivery times average six years, with market discussions pointing to 2030–31, driving data-centre developers to buy used or refurbished turbines and alternatives. Used units cost about $1.3 million–$1.7 million per MW versus $2 million–$2.5 million for new turbines, but the secondary market is limited and cannot meaningfully resolve the broader supply bottleneck.

Analysis

The investable signal is scarcity rent, not simply AI power demand. Long OEM backlogs can support pricing and service demand, but they do not guarantee near-term earnings: delivery delays defer revenue, while rushed deployments using mixed-age equipment raise commissioning, reliability and warranty risks. Used-turbine transactions mainly reallocate existing capacity; they are unlikely to close the supply gap, and can bid up replacement equipment for utilities and cooperatives, potentially delaying non-data-center projects.

Wärtsilä (WRT1V) is the clearest substitution angle: gas engines can win projects when turbine timing, rather than technology preference, determines procurement. Its reported data-center orders are a useful demand signal, not proof of attractive margins or durable share. Sulzer (SUN) and CAT’s Solar Turbines business may see aftermarket and equipment demand, but the article provides no revenue materiality; GE (GE) exposure through jet-engine-based units is similarly indirect. For GEV and Siemens Energy (ENR), backlog scarcity is supportive, but watch execution, order quality and cash conversion rather than treating backlog as immediate earnings.

Over 1–3 months, look for evidence of engine awards, OEM delivery-slot pricing and guidance revisions. Over 6–18 months, permits, grid access, gas supply and data-center financing determine whether orders become operating capacity. The contrarian point: turbine scarcity can be bullish for suppliers but bearish for project returns—higher equipment costs and slower energization may defer data-center utilization and monetization. Thesis weakens if hyperscaler capex is cut, turbine lead times fall materially, or engine alternatives fail reliability/permitting tests.

Key Decisions for Investors

• Prefer a modest, staged long in WRT1V over a broad turbine-OEM chase; add only if subsequent disclosures confirm data-center engine orders convert into revenue without margin deterioration. Reassess on order cancellations, project delays or weaker energy-segment guidance.
• Keep GEV and ENR on a backlog-quality watchlist rather than buying solely on scarcity headlines. Track delivery schedules, pricing, cash conversion and any capacity-expansion commitments; prolonged delays or rising execution costs would undermine the earnings benefit.
• Treat CAT and SUN as indirect aftermarket/equipment beneficiaries, not clean AI-power proxies. Seek segment-level evidence of incremental orders or service revenue before positioning; absent disclosure, no standalone trade.
• Monitor used-equipment prices, gas-engine awards, permitting and data-center capex over the next 1–3 months. A material decline in lead times or cancellations by large developers would falsify the scarcity thesis; persistent delays alongside rising engine orders would strengthen the substitution case.