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

The Grid That Has to Grow, part 3

Behind-the-meter generation — power plants built right next to the data center or factory they serve, bypassing the grid — is growing fast. Most of these are gas turbines, but some are nuclear (the Amazon-Talen deal from Chapter 10) and some, increasingly, are solar plus storage.

Across all of these together, the United States is building generation faster than it has built it in decades. But not faster than the demand is growing in some places. The gap is the political problem.

Even where new electricity is being generated, getting it to where it needs to go is increasingly the harder problem.

The International Energy Agency reports that more than 2,500 gigawatts of generation projects are stuck globally in grid connection queues — that is, projects that are built or nearly built but cannot connect to the grid because the transmission system cannot accept them. In the United States, the queue is so backlogged that some renewable energy projects are waiting four to seven years just to be studied for interconnection.

This is the long road problem from Chapter 3, made urgent. We have the engineering. We have the money. We have the demand. We do not have the wires. And we do not have a fast way to build the wires.

Some of the most ambitious responses to the demand surge are trying to avoid the transmission problem entirely. Hyperscalers are co-locating their data centers next to power plants. Manufacturers are building their own gas turbines on site. New startups are putting AI compute nodes on individual homes, where the local distribution grid can already serve them. We will talk about that in the next chapter.

There is one question this chapter has been circling that we have to address directly.

When a utility builds a new power plant or a new transmission line, the cost is recovered from electricity customers — in the form of higher rates on their bills. Traditionally, the cost has been spread across all customers in the utility’s service territory. The logic was that everyone benefits from a reliable grid, so everyone should share the cost.

But when most of the new demand is coming from a handful of giant new data centers owned by some of the wealthiest companies in the world, that traditional logic starts to feel unfair. Why should a retiree on a fixed income in rural Virginia subsidize the transmission upgrade required to serve a Microsoft data center?

This question has become politically explosive in the last two years. Virginia, Texas, Ohio, and several other states have seen residents organizing against new data centers and against rate increases. Polls show electricity affordability rising rapidly as a political issue. Energy prices nationally rose 6.3 percent in the year ending early 2026, according to the federal Consumer Price Index — faster than overall inflation.

In February 2026, President Trump used his State of the Union address to announce something he called the Ratepayer Protection Pledge. A few weeks later, in March 2026, seven of the largest American technology companies — Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI — gathered at the White House to sign it.

Under the pledge, these companies committed to “build, bring, or buy” all the new electricity their data centers will require. They committed to pay for any new infrastructure — transmission, substations, even local distribution upgrades — needed to serve them. They committed to negotiate separate rate structures with utilities so that the cost would be assigned to the data centers, not spread across all customers. They even committed to pay for the electricity whether they used it or not.

The pledge is voluntary. There is no federal enforcement mechanism. American electricity is mostly regulated at the state level, and states have varying rules about how costs can be allocated. Some energy-policy experts have called the pledge largely symbolic — a public commitment by tech companies to do things they were probably going to do anyway, dressed up as a political win.

But the pledge does reflect a real shift. The biggest tech companies do, in fact, want to be seen as good neighbors to the communities they operate in. They have learned that public backlash against data centers can slow their permits, raise their costs, and politicize their existence. The math, for the companies, is to pay for their own infrastructure even when the law does not strictly require it.

Whether the pledge will actually keep residential rates from rising is a question that will be answered over the next several years. Electricity rate decisions are slow. Effects on bills, even if the companies fully follow through, will take at least two years to show up. The fight will continue in state regulatory hearings, in court cases, and in elections, for the rest of the decade.

The honest framing of all this is not “data centers are eating residential electricity.” The grid is not a fixed pie. The data centers are paying for new generation that, in many cases, adds to total supply. Some of that new generation, particularly the renewable power and the nuclear restarts from Chapter 10, would not have happened without their demand.

The honest framing is this: the grid is being asked to grow faster than it has in fifty years. Who pays for the growth, and how the costs are allocated, is a political question that will be litigated state by state, year by year, for as long as the demand keeps rising.

Somewhere in northern Virginia, the planner is still at her desk, looking at the same forecast.

The forecast says her utility’s service area will need to add somewhere between five and ten gigawatts of new electricity capacity over the next five years. That is, in five years, what utilities used to build in fifty.

She will not do it alone. The same forecast is being looked at by planners in Ohio and Texas and Arizona. The same calculation is being run by every electric utility in the country. The same supply chains for transformers, for transmission cable, for skilled electricians, are being stretched to their limits.

For two decades, her job was about efficiency. About squeezing more out of what was there. About making the grid leaner.

For the rest of her career, her job will be about growth. About building things the country has not built in a generation. About arguing with regulators, with neighbors, with environmental reviewers, with state legislators, about where the lines go and who pays for them.

This is the central engineering and political project of the next decade. It is not a problem that data centers caused, or that electrification caused, or that any single industry caused. It is a problem that the entire modern economy has, by deciding it wanted more electrification of everything, and more AI, and more reshored manufacturing, and a cleaner grid, all at once.

We chose all of these. We are going to have to build the grid that can serve them.

The planner closes her laptop and goes home. Tomorrow, the forecasts will be a little higher than they were today.

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