How the Lights Stay On · The Grid That Has to Grow

The Grid That Has to Grow, part 2

Here is what makes this hard for the grid. Data centers are not spread out evenly. They cluster in specific places — Northern Virginia, central Ohio, eastern Texas, Phoenix, Arizona, and a handful of others. The cluster effect compounds the engineering challenge. A single new gigawatt-scale data center being built near an existing cluster does not just add a gigawatt to national demand. It adds a gigawatt to a local grid that may not have a gigawatt of spare capacity within reasonable transmission distance.

In Loudoun County alone, S&P Global’s 451 Research projects that data centers will draw 12 gigawatts by 2030 — roughly tripling from 2024 levels. In Texas, demand is growing nearly as fast. The local utilities have to either generate this electricity nearby or build the transmission lines to bring it in from elsewhere. Often, neither option is fast enough.

The second force is everything else getting plugged in.

For the last decade, federal and state policy has been pushing American households and businesses to switch from fossil fuels to electricity for as many uses as possible. Heat pumps to replace gas furnaces. Electric induction stoves to replace gas burners. Heat pump water heaters to replace gas tanks. Electric vehicles to replace internal-combustion cars.

Each of these changes saves carbon emissions, because most modern electricity is significantly cleaner than burning gas in your basement, and because the technology itself is more efficient. A heat pump can move three or four units of heat for every unit of electricity it uses. An electric car converts electricity to motion at over 90 percent efficiency, compared to a gasoline car’s roughly 30 percent.

But the same shift also adds electric load. American homes that fully electrify use significantly more electricity than they did before — they have just stopped burning gas separately.

In the early 2020s, the rate of electrification accelerated. EV sales in the United States grew 10 percent year-over-year in early 2025. Heat pump sales grew 14 percent in 2024 and 9.5 percent in early 2025. Each of these adoptions, multiplied across millions of households, adds a steady increase to grid demand.

For many decades, grid planners in northern Virginia and Texas could ignore residential electrification because it was small and slow. By 2024, they could no longer ignore it.

The third force is the most recently visible: American manufacturing is rebuilding.

For decades, the United States offshored most of its industrial production. The trend reversed in the late 2010s and accelerated in the 2020s. The reasons are several — pandemic supply-chain shocks, US-China political tensions, government subsidies under the CHIPS Act and the Inflation Reduction Act, and the long-term recognition that critical industries should not all be located in any single foreign country.

The result has been a manufacturing construction boom, particularly for two categories: semiconductor fabrication plants (the “fabs” that make computer chips) and battery factories.

Semiconductor fabs are some of the most electricity-intensive buildings ever built. A single new TSMC fab in Arizona, or Intel fab in Ohio, can draw a gigawatt of continuous power — comparable to a small city. Battery factories are not far behind. Both are also highly concentrated geographically, like data centers, because they need huge supplies of trained labor and water and zoning permits, all of which are easier to assemble in some places than others.

By the first quarter of 2025, US construction spending on manufacturing facilities was more than double what it had been three years earlier. Most of those facilities, when they come online, will be heavy electricity consumers for the next several decades.

So that is the demand picture. Twenty-five percent growth by 2030, conservatively. Concentrated in a few places. Driven by data centers, electrification, and reshored manufacturing.

Where will the new electricity come from?

Every source we have talked about in this book is being deployed, fast.

Natural gas is, in the short term, the biggest beneficiary. Gas plants can be built faster than almost anything else — eighteen to thirty months from groundbreaking to operating — and they provide firm, dispatchable power. Several states are seeing new gas plants approved for the first time in years. This is controversial. Gas is a fossil fuel. Building new gas plants in 2026 means committing to those emissions for the next thirty or forty years. Some clean-energy advocates argue this is a betrayal of climate goals. Some utilities argue there is no other option that can be built fast enough. Both sides have a point.

Solar and wind continue to be built at record pace. The United States added about 50 gigawatts of new solar and wind capacity in 2024. These remain the cheapest sources of new electricity per kilowatt-hour, even after accounting for intermittency. But, as Chapter 6 showed, you cannot run a grid on solar and wind alone without storage and transmission — both of which are also being built, and both of which are still bottlenecks.

Nuclear is back, as Chapter 10 showed. The Three Mile Island restart, the Vogtle reactors, the SMR deals. But most of this capacity will not come online until the late 2020s or the early 2030s. In the immediate term, nuclear is a hope, not a solution.

Geothermal is the dark horse. New techniques — drilling deeper, using methods borrowed from the fracking industry to engineer underground reservoirs — are making geothermal viable in many more places than the few traditional hot-spring regions. It is firm, like nuclear, but cleaner-image and faster to build. Several large geothermal projects are now under construction in Nevada, Utah, and Texas.

All 18 chapters of How the Lights Stay On · Library · All courses