Episode Summary
Executive Summary: Nat Bullard and Sho Khan unpack Bullard’s annual energy/decarbonization slide deck, focusing on electrification, electricity’s long-stable share of GDP, surging data-center-driven power demand, and severe shortages across gas turbines and grid equipment. The conversation frames today’s energy transition as a supply-chain and buildout bottleneck rather than a technology question alone.
Main Topics: China’s much higher electrification than the U.S. (Priority: 5/5): Bullard explains that China’s economy is substantially more electrified than the U.S., with a much faster rise in electricity’s share of final energy since 1990. He ties this to China’s industrial scale and desire for energy sovereignty through domestic electricity generation. Electricity spend has stayed range-bound while oil spend has been volatile (Priority: 5/5): The hosts discuss a striking historical pattern: electricity has consumed roughly 3%-4% of GDP for decades, while oil spending has been far more cyclical and historically much larger. This raises the question of whether rising electricity demand can finally break the long-standing pattern. Gas turbine supply is badly constrained (Priority: 5/5): Bullard shows that global gas turbine order books are now far above production capacity, reflecting a severe supply shortage. The discussion highlights how utility and data-center demand is colliding with limited manufacturing capacity and past industry scars from overbuilding. Power plant costs are rising sharply (Priority: 4/5): Using regulatory filings, Bullard tracks planned gas plant capital costs and shows that combined-cycle projects expected online later in the decade are materially more expensive than near-term projects, with additional inflation risk from EPC labor and supply constraints. Tech/data-center capex is enormous by historical standards (Priority: 4/5): A comparison of current tech capital spending with major historical buildouts shows that 2025 tech capex is nearly 2% of U.S. GDP, exceeding earlier booms like the Interstate Highway System, Apollo, and broadband buildout. Much of this spending is tied to AI infrastructure. Texas is a focal point for speculative load growth (Priority: 5/5): ERCOT’s large-load interconnection queue has ballooned, but the real level of future demand is highly uncertain. The gap between ERCOT’s and transmission providers’ forecasts underscores how speculative and incentive-driven today’s demand projections are. Global electricity demand growth is broader than data centers (Priority: 4/5): A global IEA outlook shows data centers are important but not the largest source of new electricity demand; industrial electrification, appliances, EVs, and cooling currently rank higher. Bullard argues this mix could shift quickly as data-center growth accelerates.
Key Arguments: China’s electricity share of final energy has risen from about 7% in 1990 to roughly 30% today, far outpacing the U.S. and Europe, which have changed slowly over decades. China’s electrification is not just about efficiency; it is also about energy sovereignty because electricity is produced within national borders, reducing exposure to imported oil and gas. Electricity spending has stayed remarkably flat as a share of GDP for about 50 years, suggesting a powerful structural collar that AI-driven demand may test. Oil spending as a share of GDP has been far more volatile than electricity, reflecting oil’s status as the historically shock-prone fuel. Gas turbine markets are now structurally undersupplied: order books exceed annual production capacity by a wide margin, and manufacturers are reluctant to expand too aggressively because of memories of prior boom-bust cycles. Rising gas plant costs are being driven not only by turbine scarcity but also by EPC bottlenecks, labor shortages, tariffs, and construction inflation. Texas illustrates the next phase of the power crunch: interconnection queues are exploding, but much of the queue is speculative and not all projects will be built. Forecasts for future Texas electricity demand differ massively between ERCOT and transmission providers, showing that even experienced market participants lack a common view of near-term load growth. Globally, data centers are a major demand driver but not yet the largest one; broader electrification trends still dominate total demand growth. The current buildout cycle is likely to outlast current market expectations because capital spending commitments already underway tend to lag financial-market sentiment.
Data Points: China electricity share of final energy: ~30% today - Bullard cites Ember data showing China is now significantly more electrified than the U.S. China electricity share of final energy in 1990: ~7% - Used as a benchmark for China’s long-run electrification trajectory. North America electricity share of final energy: ~20%-23% - Bullard says the U.S./North America has been around this level since 1990. Electricity share of GDP: ~3%-4% - Global electricity expenditure has remained range-bound for decades. Oil share of GDP in 1980: just under 9% - After the second oil shock, global GDP expenditures on oil were unusually high. Oil share of GDP today: ~5% - Oil spending has fallen from its 1980 peak but remains volatile. Oil share of GDP at 2020 low: below 4% - Oil spending dipped sharply during the pandemic period. Gas turbine production capacity: ~60 GW/year - Approximate annual output limit for the small number of turbine manufacturers. 2028 gas turbine order book: over 100 GW - Current orders exceed production capacity by a large margin. 2026 combined-cycle plant cost: ~$1,200/kW - Planned near-term delivery cost for combined-cycle gas plants. 2030-2031 combined-cycle plant cost: just under $2,500/kW - Planned future delivery cost, nearly double the near-term figure. Simple-cycle turbine cost: ~$1,000/kW to ~$1,500/kW - Costs have risen roughly 50% in the plan set discussed. RICE plant cost: ~$2,500/kW to ~$3,000/kW - Reciprocating internal combustion engine projects appear expensive and short-dated in the pipeline. Tech capex as share of U.S. GDP: just under 2% - 2025 tech capital spending, mostly compute/data-center related, compared with historical buildouts. Broadband capex peak: ~1.2% of U.S. GDP - Year 2000 broadband buildout peak, used as a historical comparison. Interstate Highway Project peak: ~0.6% of U.S. GDP - Historical public-works comparison for major buildout spending. ERCOT large-load queue Jan 2024: ~42 GW - Starting point for Texas large-load interconnection requests. ERCOT large-load queue Nov 2025: 226 GW - Queue growth over roughly two years in Texas. Texas peak load: ~85 GW - Shows the queued large loads would vastly exceed current peak demand if all materialized. ERCOT demand forecast for 2030: ~1,000 TWh - Grid operator’s expected load trajectory from 2024 levels. Transmission service providers’ 2030 forecast: ~1,600 TWh - More aggressive forecast based on requested builds and load assumptions. U.S. total electricity demand: ~4,500 TWh annually - Used to frame the scale of Texas forecast divergence. Global electricity demand growth from electrification of industry: ~30% - IEA projection of the biggest source of new electricity demand through 2030. Global electricity demand growth from data centers: ~8% - IEA projection shows data centers are important but not dominant globally. Global electricity demand growth from space cooling: ~10% - Air conditioning/buildings remain a larger driver than data centers globally.
Pivotal Quotes: "we basically spend between 3 and 4% of GDP on electricity and that is that essentially" — Nat Bullard: On the long-run stability of electricity spending as a share of GDP. "everything is turbine" — Sho Khan: On the rush to convert every available turbine-related asset into data-center power supply. "no one knows anything" — Nat Bullard: On the huge gap between ERCOT and transmission-provider load forecasts in Texas.
Implications: The energy transition is being shaped less by technology alone than by physical bottlenecks, capital intensity, and uncertain load growth. Utilities, developers, and manufacturers face a tighter, more expensive, and more speculative buildout environment.