#AI's Power Grab Is Splitting the US Grid in Two
Rhodium Group's Taking Stock 2026 report, released July 29, projects that US electricity demand growth is about to hit its fastest pace since before 2000 [1]. Under the firm's central assumptions, annual demand growth over the next 15 years could more than triple the average pace seen from 2000 through 2023.
The single biggest driver is data centers. Rhodium's model assumes their power demand more than triples by 2040 versus 2025, driving 59% to 77% of total demand growth depending on the year, and making up 17% of all US electricity use by 2040. But the firm is explicit that this is one assumed path, and that the real uncertainty here is large. It notes the gap between the highest and lowest analyst estimates for 2030 data center demand alone is bigger than the entire power use of California and Florida combined.
For investors, that gap matters as much as the growth story itself. It shapes the case for transformers, turbines, transmission lines, and every company building or servicing the grid data centers plug into, but it also means nobody really knows how big that opportunity actually is.
#A Grid Torn Between Gas And Renewables
Once that demand estimate is fixed, Rhodium runs three scenarios for how it gets met, based on different assumptions about fuel prices, technology costs, and economic growth. The outcomes by 2040 could hardly be more different.
A shortage of gas turbines has roughly doubled the cost of building new gas plants through 2030, which limits how fast utilities can add gas capacity even as demand keeps rising. That, plus tax credits set to expire, is why solar, wind, and battery storage are being built at a near record pace. Developers are rushing to build now, through 2027, specifically to lock in tax credits before they disappear. That's a pull-forward effect, some of this construction is happening earlier than it otherwise would, precisely because the credit window is closing.
After that, the paths split sharply. In the scenario with high fuel prices and cheap clean technology, renewable construction speeds back up through the 2030s and the grid ends up 72% clean. In the scenario with cheap gas and expensive clean technology, renewable construction slows to a crawl, gas ends up supplying half of US power, and the clean share falls to 47%.
A sharper way to see the markets in just 5 minutes.
Same news, different lens. We cut through the noise and hand you the overlooked ideas and the deeper read the crowd misses. Join 38,000+ investors seeing the markets differently.
#Prices And The Political Risk
None of this comes free. Rhodium projects real electricity prices will keep rising faster than inflation no matter which scenario plays out, with the average residential bill up 39% to 48% in nominal terms by 2040, a 9% to 13% increase after inflation. Some states and utilities are already considering rules that make data centers cover more of their own costs instead of spreading them across everyone's bill, or that slow how fast new ones can connect to the grid. Those are political decisions no energy model can predict.
#Who Actually Captures The Upside
Higher prices do not automatically mean fatter profits for utilities. Regulated utilities earn a set return on the capital they invest, not a share of every extra dollar on a bill, so their earnings track approved capital spending more than the price trend itself. If regulators push back on data center related costs, as some already are, that caps the upside even while demand keeps climbing.
Equipment makers sit on the other side of that trade. When new gas turbine capital costs roughly double, as Rhodium's model assumes through 2030, that pricing power flows straight to manufacturers rather than through a regulatory approval process. The same dynamic applies to transformers and other grid hardware the report flags as running short.
Put simply, rising prices are a better signal for equipment and infrastructure builders than for utility profit margins. What actually decides utility upside is whether regulators approve their spending and whether large customers like data centers get charged fairly for the capacity they use, not the price trend itself.
That is the real tension for investors weighing the grid buildout theme. It hinges on assumptions that could move either way, whether tax credits get extended, whether turbine supply catches up, and whether data center growth tracks anywhere near what Rhodium assumes. Watching how much new capacity actually gets built each year is a reasonable way to track which path is winning.