For most of the last two decades, the U.S. power sector operated under a comfortable fiction: electricity demand would remain mostly flat, generation transitions could be managed gradually, and reliability would somehow take care of itself.

In 2025, that fiction collapsed.

Electricity demand surged for the first time in a generation, driven primarily by data centers, AI workloads, and industrial growth. Prices rose. Capacity margins tightened. Regulators leaned on emergency authorities. Plants once slated for retirement were kept online out of necessity. And blackouts stopped sounding like a distant risk and started entering everyday conversation.

At the same time, the system responded. The U.S. added record amounts of new capacity. Natural gas returned to the center of reliability planning. Nuclear began inching from aspiration toward execution. Permitting reform, long stalled, started to move under pressure from grid physics rather than politics.

These are the stories that mattered in 2025 because they forced real decisions, moved markets, and reshaped how the grid is planned.

In 2025, data centers crossed a threshold from “emerging concern” to dominant planning variable. Federal and regional forecasts now show data centers consuming roughly four percent of U.S. electricity today, with credible projections placing them in the high single digits or even double digits before the decade’s end. In certain regions, especially Northern Virginia, Texas, Ohio, and parts of the Midwest, large-load requests alone rivaled the size of existing generation fleets.

This wasn’t speculative growth. Utilities began revising load forecasts upward, acknowledging that AI-driven demand is structural, not cyclical. The era of planning around flat or declining demand officially ended.

Grid Take
Large new loads aren’t a grid failure. They’re the justification for building. Treating data centers as a problem to restrain rather than demand to serve is how you guarantee scarcity instead of growth.

The U.S. added more utility-scale generation capacity in 2025 than in any year on record, dominated by solar and battery storage. Texas and California led deployment, and clean energy made up the overwhelming majority of new additions.

Yet despite the headline numbers, system operators grew increasingly concerned about reliability. Much of the new capacity was intermittent or duration-limited, arriving faster than the firm generation needed to backstop peak demand and extreme weather events.

Grid Take
We’re getting very good at adding megawatts and much slower at adding reliability. Energy abundance isn’t just about volume, it’s about availability when conditions are worst.

As electricity demand accelerated in 2025, utilities and large customers turned decisively to natural gas. By year’s end, project announcements and utility filings pointed to roughly 45–80 gigawatts of new gas-fired capacity planned or under development for later this decade, much of it tied directly to data centers and industrial loads. Individual deals frequently ranged from 1 to 3 GW, reflecting the scale of demand coming online.

These decisions were not driven by ideology or nostalgia. Gas plants offered something no other technology could deliver on comparable timelines: firm, dispatchable power that could be permitted, financed, and constructed in three to five years, rather than a decade or more. In practical terms, gas became the enabling technology for near-term AI and industrial expansion.

Grid Take
Gas didn’t “win” a policy debate. It filled a reliability vacuum. Until the system can deliver firm, non-emitting power at scale and speed, gas will remain the default.

In 2025, the planned retirement of aging coal and gas plants repeatedly collided with reliability needs. Across multiple regions, most notably PJM, utilities delayed or canceled scheduled shutdowns as reserve margins tightened and replacement capacity lagged rising demand. Units that had already been deemed uneconomic or obsolete on paper remained critical in practice.

At the same time, the federal government invoked emergency authorities to keep specific plants available during periods of peak stress, extreme weather, and transmission constraints. These orders covered several gigawatts of capacity across PJM, MISO, and the Pacific Northwest, overwhelmingly involving the same coal and gas units that had been slated for retirement or reduced operation. What made 2025 notable was not any single order, but how routine the mechanism became. Emergency directives were used repeatedly to bridge predictable gaps during winter cold snaps, summer peaks, and localized grid congestion.

Grid Take
This is what underbuilding looks like in real time. When markets and policy push plants out before replacements arrive, reliability gets propped up by emergency powers and last-minute reversals. That may keep the lights on, but it’s a sign the system has lost slack it hasn’t replaced with firm capacity or transmission.

In December, federal regulators were forced to address a rapidly growing workaround: data centers co-locating directly with generation to bypass interconnection delays. By ordering PJM to develop clearer rules and new transmission service options, regulators acknowledged that the existing framework no longer matches how large loads are actually being deployed.

This marked a shift from pretending the issue didn’t exist to trying to manage it explicitly.

Grid Take
Co-location didn’t break the grid. It exposed where the grid stopped keeping up with economic reality. Clear rules are better than informal workarounds, but they’re no substitute for faster interconnection.

PJM’s 2027/2028 capacity auction cleared at the maximum allowable price, sending a stark signal about the availability of firm capacity. Even at record prices, the auction failed to fully meet the region’s reliability requirement, revealing a meaningful shortfall.

For customers, the result was higher bills. For planners, it was confirmation that capacity scarcity is no longer theoretical.

Grid Take
This wasn’t a market failure. It was the market doing its job too late. Price signals only help if the system can respond before shortages become routine.

ERCOT’s large-load interconnection queue grew several times over in a single year, dominated by data centers and crypto facilities. Even in the country’s fastest-moving grid, transmission expansion and interconnection studies struggled to keep pace.

The story wasn’t unique to Texas, but the scale made it impossible to ignore.

Grid Take
Speed matters, but physics still wins. Even the most flexible markets need wires and generation to match growth.

Residential electricity prices rose sharply in 2025, approaching eighteen cents per kilowatt-hour on average, up from roughly the mid-sixteen-cent range in late 2024. Utilities issued record amounts of debt to finance grid upgrades, locking in higher costs for years to come.

Electricity affordability reentered the political spotlight.

Grid Take
Rising prices are what turn grid planning into public policy. If affordability matters, the solution is more supply and faster delivery, not demand suppression.

Federal and regional reliability assessments issued increasingly blunt warnings in 2025, projecting sharply higher blackout risk by the end of the decade if firm capacity additions fail to keep pace with load growth. These weren’t activist reports. They were operational warnings from grid planners.

The tone changed noticeably.

Grid Take
When reliability warnings start sounding repetitive, it’s because the underlying math hasn’t improved. Planning optimism doesn’t keep the lights on.

In 2025, virtual power plants crossed a meaningful threshold. Aggregated demand response, distributed batteries, and smart devices were deployed at scale in several regions, with leading programs now providing hundreds to low thousands of megawatts of dispatchable capacity during peak events. In California alone, VPP portfolios were credited with over 3 GW of peak load reduction during extreme heat events, while utilities in the Midwest and Northeast increasingly incorporated VPPs into seasonal reliability planning.

These resources proved valuable in shaving peaks, easing local congestion, and buying time during stress events. But their limitations also became clearer. VPP output is conditional, duration-limited, and dependent on customer participation, making it fundamentally different from always-available generation or new transmission.

Grid Take
VPPs are a real reliability asset, not a gimmick. But they are a multiplier, not a substitute. They stretch existing infrastructure and improve flexibility at the margins, yet they cannot replace the need for firm generation and major transmission buildout as demand continues to grow.

In 2025, nuclear energy began to move out of the conceptual phase and into early execution. Several large reactors received life extensions that will keep tens of gigawatts of existing nuclear capacity online into the 2040s and beyond. Restart efforts advanced at previously retired sites, while multiple utilities and state governments committed funding and regulatory support for new builds and small modular reactors, shifting discussions from “if” to “where and when.”

At the federal level, DOE backing for advanced reactor demonstrations and fuel supply chains translated into clearer development timelines. While no wave of new reactors broke ground this year, utilities began naming sites, estimating costs, and integrating nuclear into long-term resource plans in a way that hadn’t happened for years.

Grid Take
Nuclear remains the most compelling source of clean, firm power at scale. The challenge is no longer technological feasibility. It’s institutional execution. If licensing, supply chains, and construction timelines can be made predictable, nuclear becomes a cornerstone of energy abundance rather than a perpetual promise.

2025 wasn’t an energy transition year. It was a stress test.

Demand surged. Prices rose. Emergency tools filled gaps that investment hadn’t yet closed. At the same time, the system responded: capital flowed, capacity was added, and long-ignored constraints became impossible to deny.

Whether 2026 brings true abundance or entrenched scarcity will depend less on ambition than on execution: faster interconnection, real permitting reform, and a willingness to build what reliability actually requires.

Thanks for reading GridBrief in 2025. More to come in 2026.

We rely on word of mouth to grow. If you're enjoying this, don't forget to forward Grid Brief to your friends and ask them to subscribe!