2026 trends: Gas turbines, Texas’ load queue, and China electrifies

Start with supply, where the behavior is deliberate. Manufacturers can build roughly 60 gigawatts of gas turbines a year; orders ran about 20 gigawatts above that last year, about 30 above this year, and the 2028 book is over 100 gigawatts. The obvious reading is that they cannot keep up. Bullard’s is that they largely do not want to. His series starts in 2001, when the world ordered close to 90 gigawatts, manufacturers built out to meet it, and orders collapsed to well under 40 the next year and stayed below capacity for most of two decades. Anyone senior enough to approve an expansion remembers that, and undersupply is the better commercial position anyway, helping you book durable contracts with the customers you want and helping on price. The cost shows up in utility filings: combined-cycle plants arriving this year at roughly $1,200 per kilowatt against 2030 and 2031 projects at just under $2,500.

On the demand side the numbers stop meaning what they appear to mean. The large load interconnection queue at the Electric Reliability Council of Texas, the state’s grid operator, went from about 42 gigawatts in January 2024 to 226 in November 2025, against a state peak load near 85. Neither speaker thinks it will be built. What Bullard argues is new is not the speculation but where it sits: energy people expect speculative supply-side queues, where a developer permits ten solar projects to build one or two, but a speculative queue on the demand side is unfamiliar, and nobody sites a hospital across seven states at once. Kann pushes back on the mechanism, arguing these are mostly not developers choosing among sites for one facility but speculators developing whatever site they can monetize by sale or lease. Either way nothing nets out the duplicates, because interconnection is regulated state by state. Hence the “no one knows anything” slide: the grid operator forecasts Texas going from under 500 terawatt-hours in 2024 to about 1,000 by 2030, while the transmission service providers, serving the same market, forecast about 1,600. The utilities see every request they are asked to build and are paid to build assets; the operator forecasts what it thinks will happen. Bullard puts the truth between them, closer to the operator.

Two slides then supply proportion, pointing opposite ways. Against the great American build-outs, 2025 technology capital expenditure is larger than any of them: the Manhattan Project, postwar electrification and Apollo each peaked at or barely above half a percent of gross domestic product, broadband around 1.2% in 2000, and technology capital expenditure just under 2%, probably not yet at its peak. But that figure covers only what sits inside the companies’ own fence, and excludes the utility spending that follows. Pointing the other way, on International Energy Agency projections of global demand growth to 2030, data centers rank fifth at around 8%, behind industrial electrification at roughly 30%, electrified transport, and space cooling and buildings at around 10%.

Underneath all of it sits a fifty-year regularity the episode opens with and never resolves. Spending on electricity has stayed between roughly 3% and 4% of global gross domestic product for five decades, while oil’s share swung from just under 9% in 1980 to below 4% in 2020 and sits near 5% now. The flat line means electricity spending has grown with the economy and never faster. Whether AI breaks that depends less on the spending, which everyone expects to rise, than on whether output rises with it, which is precisely the contested question about AI. Bullard declines to guess. The China slide asks it from the other end: China went from about 3% of final energy delivered as electricity in 1970 to roughly 30% today, while North America moved from about 20% in 1990 to 22 or 23% now. Bullard’s explanation is partly sovereignty. Electricity generated inside your borders is yours whether it comes from coal or from solar, and China imports oil and gas while sitting on centuries of domestic coal, so electrification buys energy independence even where it does not buy decarbonization.