Gas Turbine Supercycle: The Hidden Beneficiary Is Industrial HVAC & Cooling, Not Power Producers

The thesis fails if gas turbine installations shift toward modular, factory-integrated skids where the OEM bundles all ancillary components in-house, eliminating the tier-2 supplier opportunity. The concrete observable signal: if GE Vernova or Caterpillar announces a vertical integration strategy for heat recovery or fluid management systems (e.g., an acquisition of a mid-cap ancillary supplier), it would signal the OEM is capturing the margin rather than passing it through the supply chain. Watch GEV and CAT M&A announcements in the industrials space.

The strongest bear case is that the $50B figure represents announced intent, not contracted spend — and that US data center power demand growth slows faster than expected as AI model efficiency improves (inference compute per query dropping sharply with each model generation). If Anthropic, OpenAI, and hyperscalers find that next-gen models require 40-50% less compute per task, the power demand projections underpinning the gas turbine supercycle get revised down sharply. The specific assumption most likely to be wrong: that data center power demand is relatively inelastic to AI efficiency gains. The data point that would flip conviction entirely: a public hyperscaler disclosure (e.g., Google or Microsoft earnings commentary) indicating data center power consumption is growing slower than new capacity additions, signaling demand saturation rather than a structural shortfall.

The 2000s natural gas combined-cycle (NGCC) buildout provides the closest analog. When US utilities accelerated NGCC construction from 1998-2002 (~150 GW of new capacity), the headline trade was on turbine OEMs (GE, Siemens). But the outsized returns came from tier-2 suppliers: Watts Industries (fluid control) returned 3x the S&P 500 from 1999-2003, and Roper Technologies (then Roper Industries, precision flow measurement) compounded at 28% annually through the same period. The OEMs faced margin compression from rapid scaling; the component suppliers held pricing power because switching costs were high once a component was engineered into a specific turbine skid. A second analog: the 2010-2015 shale midstream buildout. The market bought pipeline MLPs; the real alpha was in mid-cap compression equipment suppliers like Chart Industries and CIRCOR — both re-rated 2-4x as OEM backlogs stretched to 18-24 months and tier-2 suppliers gained pricing leverage. The pattern is consistent: infrastructure supercycles create bottlenecks one level down the supply chain from where the market is looking.