Dispatchable Power Supply Constraints And Thermal Capex Inflation
Key takeaways
- Global gas turbine production capacity is about ~60 GW per year and current orders imply demand exceeding that by roughly ~20–40 GW annually in the near term.
- ERCOT and Texas transmission service providers have sharply divergent 2030 electricity-demand forecasts: about 1,000 TWh versus about 1,600 TWh.
- Globally, electricity spending has been remarkably stable at roughly 3–4% of GDP for decades.
- The cited 2025 tech capex share excludes external power, transmission, and water investments needed to support data centers.
- China’s electricity share of final energy rose from roughly 3% in 1970 to about 30% today, while North America and Europe rose more gradually to just over 20%.
Sections
Dispatchable Power Supply Constraints And Thermal Capex Inflation
The corpus quantifies a turbine manufacturing capacity ceiling (~60 GW/year) against an order environment that exceeds it, including heavy forward backlog into 2028. It also provides a mechanism for why OEMs may prefer sustained tightness given historical boom-bust experience. Consistent with scarcity, planned gas plant capital costs are cited as rising sharply, and substitution patterns (aeroderivatives for data centers; recips at high $/kW with short pipeline) are described as responses to constraints rather than first-best choices.
- Global gas turbine production capacity is about ~60 GW per year and current orders imply demand exceeding that by roughly ~20–40 GW annually in the near term.
- Booked gas turbine order volume for 2028 is already over ~100 GW versus ~60 GW of annual production capacity.
- Planned U.S. combined-cycle gas plant capital costs roughly doubled from about ~$1,200/kW for 2026-online projects to just under ~$2,500/kW for 2030–2031-online projects.
- The gas turbine market saw a boom-bust in the early 2000s with roughly ~80–90 GW of orders in 2001 collapsing to well under ~40 GW the following year.
- Gas turbine manufacturers have incentives to avoid overexpanding capacity because undersupply improves contract selectivity and pricing and firms remember oversupply after the early-2000s collapse.
- Planned simple-cycle gas plant capital costs are up roughly ~50%, from about ~$1,000/kW to about ~$1,500/kW.
Ercot Load-Queue Inflation And Forecast Disagreement
ERCOT’s large-load queue growth is quantified as extreme and is contextualized versus Texas peak load. The corpus describes compositional detail (co-location) and introduces mechanisms for queue inflation via speculative behavior and site-control monetization. Separately, a major institutional dispute is described: ERCOT versus TSP 2030 demand forecasts differ by ~500 TWh, with an incentive-structure mechanism offered as partial explanation; the magnitude is framed as nationally material.
- ERCOT and Texas transmission service providers have sharply divergent 2030 electricity-demand forecasts: about 1,000 TWh versus about 1,600 TWh.
- ERCOT’s large-load interconnection request pipeline grew from roughly 40–42 GW in January 2024 to 226 GW by November 2025.
- If realized, ERCOT’s 226 GW large-load queue would be about 2.5× Texas’s current peak load of roughly 85 GW.
- The ERCOT-versus-TSP 2030 forecast delta is about 500 TWh, more than 10% of total U.S. annual electricity demand.
- Much of the Texas load-queue inflation is driven by site developers and speculators who believe interconnectable land can be monetized via sale or lease to hyperscalers and colocation providers.
- The ERCOT-versus-TSP 2030 forecast gap is partly incentive-driven because ERCOT optimizes for reliable least-cost operations while TSPs are paid to build assets and rely more directly on customer requests with uncertain discounting.
Macro Energy Expenditure Stability And Potential Crossover Risk
Oil spending is presented as volatile and historically large, while electricity spending is presented as stable at ~3–4% of GDP for decades. A forward-looking watch item raises whether electricity could break that stability and even exceed oil’s GDP share, with conditions involving low oil prices and higher electricity spending or weaker GDP. Another watch item ties the outcome to whether AI-driven load growth is accompanied by AI-driven GDP/productivity gains.
- Globally, electricity spending has been remarkably stable at roughly 3–4% of GDP for decades.
- A key open question is whether electricity spending could break its historical stability and potentially exceed oil’s share of GDP given low oil prices and upward pressure on electricity prices.
- Whether electricity’s GDP share rises may hinge on whether AI-driven electricity demand growth is matched by AI-driven GDP growth versus coinciding with weaker productivity outcomes.
- Oil spending as a share of per-capita GDP was just under ~9% globally in 1980, fell below ~5% by the late 1990s, and remains volatile around ~4–7% depending on oil prices.
- For electricity spending to exceed oil spending as a share of GDP would likely require both much lower oil prices and either materially higher electricity spending or weaker GDP growth.
Data-Center/Compute Capex As An Economy-Scale Investment Wave
Tech/data-center capex is benchmarked at just under 2% of U.S. GDP, exceeding several historic investment waves, and explicitly excluding key enabling infrastructure (external power, transmission, water). The corpus suggests capex cycles can persist after market sentiment turns and includes an expectation that capex may not yet be at peak. An additional expectation-range quantifies potentially very large incremental utility/grid capex required to energize these loads.
- The cited 2025 tech capex share excludes external power, transmission, and water investments needed to support data centers.
- In 2025, tech capex (largely data-center/compute related) is just under 2% of U.S. GDP and exceeds historic capex booms such as broadband (~1.2% at the 2000 peak) and the interstate highway system (~0.6%).
- Large capex cycles can lag financial-market turning points, as illustrated by broadband capex continuing after the 2000 NASDAQ crash.
- Tech capex is expected to be not yet at its peak because company-guided estimates imply higher spending next year and significant projects are already committed.
- Utility and grid-related capex tied to energizing data centers is plausibly in the tens to hundreds of billions of dollars in addition to tech capex.
Electrification Divergence And Energy Sovereignty Framing
The corpus asserts that China’s electrification has advanced much faster than North America/Europe and has been structurally stagnant in North America since ~1990. A separate mechanism frames electrification as a sovereignty play because electricity supply is more domestically controllable than globally traded fuels. The combined delta is a reframing: electrification progress and strategic motivation may not align with common West-ahead assumptions.
- China’s electricity share of final energy rose from roughly 3% in 1970 to about 30% today, while North America and Europe rose more gradually to just over 20%.
- North America’s electricity share of final energy has been roughly ~20% since 1990 (about 22–23% today), while China increased from ~7% in 1990 to ~30% today.
- China’s economy is substantially more electrified than the United States when measured by electricity’s share of final energy consumption.
- Electrification can increase energy sovereignty because electricity is largely generated within national boundaries even when produced from thermal fuels, reducing exposure to geopolitics and global commodity markets.
Watchlist
- A key open question is whether electricity spending could break its historical stability and potentially exceed oil’s share of GDP given low oil prices and upward pressure on electricity prices.
- Whether electricity’s GDP share rises may hinge on whether AI-driven electricity demand growth is matched by AI-driven GDP growth versus coinciding with weaker productivity outcomes.
- Data centers’ global ranking as a driver of electricity-demand growth may rise meaningfully in future IEA publications.
Unknowns
- What fraction of ERCOT’s 226 GW large-load interconnection requests will reach commercial operation, and on what timeline?
- What specific assumptions and methodologies drive ERCOT’s ~1,000 TWh versus TSPs’ ~1,600 TWh 2030 demand forecasts, and how are customer requests discounted for uncertainty?
- Will global electricity spending remain near ~3–4% of GDP or break upward, and what would be the primary driver (higher tariffs, faster load growth, slower GDP growth, or some combination)?
- Will AI-driven electricity demand growth be matched by AI-driven GDP and productivity growth, or will load rise without commensurate macro output gains?
- Will gas turbine OEMs expand manufacturing capacity beyond ~60 GW/year, and if so, by how much and how quickly relative to booked orders (including >100 GW for 2028)?