The summer of 2026 has become a stress test the American electric system was not designed to pass gracefully. Across the eastern seaboard and deep into Texas, successive heat domes have pushed regional grids to demand levels planners once treated as distant, once-in-a-generation outliers. In early July, the PJM Interconnection—the largest grid operator in North America, serving roughly 67 million people from the Mid-Atlantic to the edge of the Midwest—came within reach of its all-time consumption record.

 Weeks later, in the third week of July, the Electric Reliability Council of Texas shattered its own peak, drawing more than ninety gigawatts of load in a single afternoon. These were not isolated weather events. They were the visible surface of a deeper structural shift: record electricity demand is being driven not only by hotter summers but also by a new and voracious class of consumer—the hyperscale data center. Understanding what happened this summer, and why federal regulators reached for emergency authorities they rarely invoke, is essential to understanding where grid reliability is headed for the rest of the decade.

Three forces converged this summer: extreme heat pushed demand to record levels, hyperscale data centers accelerated load growth beyond traditional planning assumptions, and emergency authorities began treating large flexible loads as reliability resources. The result was not simply a hot-weather story but a preview of the planning model U.S. grid operators will need for the rest of the decade.

A Record Summer, Coast to Coast

The numbers tell a story of a system operating at the edge of its margins. During the Independence Day heat wave, PJM forecast that load would surpass 165,563 megawatts, the record it had held since the summer of 2006. The grid ultimately peaked at roughly 163 gigawatts that Thursday, with a preliminary Wednesday peak of 161,910 megawatts—close enough to the two-decade-old high-water mark to leave operators with little room for error. By Thursday afternoon, operating reserves had fallen to 5,091 megawatts, down sharply from nearly 11,000 megawatts the day before. In a system of PJM’s scale, that reserve cushion was the operational equivalent of driving with the fuel light on.

Texas told a parallel story with its own numbers. On July 22, ERCOT recorded 90,027 megawatts of demand at three o’clock in the afternoon, eclipsing the previous day’s 87,403 megawatts and comfortably beating the prior all-time record of 85,508 megawatts set in August 2023. Heat indices across the state reached extraordinary levels—119 degrees in the Dallas–Fort Worth metroplex, 116 near Houston, and 115 in the Austin area—as an extreme-heat warning blanketed the region. What makes the Texas figures notable is not merely their magnitude but the fact that the grid absorbed them without an emergency declaration. As of the afternoon peak, ERCOT reported roughly 113,543 megawatts of available capacity and issued no conservation appeals, with a spokesperson describing the grid as operating under normal conditions. The contrast between a Texas grid taking a record in stride and an eastern grid invoking federal emergency powers is one of the defining lessons of the season.

The Emergency Toolbox Comes Out

What distinguished PJM’s summer was the recourse to a rarely used federal instrument. Under Section 202(c) of the Federal Power Act, the U.S. Department of Energy can order generators and grid operators to take extraordinary steps to preserve reliability during an emergency. The first such PJM emergency order of the year arrived unusually early, on May 18, when an out-of-season heat wave collided with more than forty gigawatts of generation sitting idle for scheduled spring maintenance. With projected reserves below 5,800 megawatts and peak loads forecast to approach 136,000 megawatts, the DOE authorized PJM to curtail power to data centers and other large loads equipped with on-site backup generation as a last resort before rolling blackouts.

By the July heat wave, that toolbox was open again and used more aggressively. The DOE authorized PJM to maximize output from fossil-fueled generation and, critically, to curtail large loads with backup generation before resorting to voltage reductions or broader load shedding. PJM had already recalled all maintenance outages to service by June 25 to squeeze every available megawatt onto the system. Energy Secretary Chris Wright framed the stakes bluntly, calling affordable and reliable power in the PJM territory “non-negotiable” and asserting that additional dispatch of specified resources was necessary to meet the emergency. The second PJM emergency order in two months signaled something important: mechanisms once reserved for genuine crises are becoming a recurring feature of ordinary summer operations. Buried in the details was an intriguing structural insight—regulators pointed to tens of gigawatts of largely untapped backup generation sitting behind the meter at data centers and large industrial sites, a shadow fleet that the emergency framework is beginning to treat as a reliability asset rather than an afterthought.

When Data Centers Set the Price

If reliability was the headline, price was the subplot that will linger longest in ratepayers’ memories. During the July peak, day-ahead wholesale power prices topped $2,000 per megawatt-hour in parts of the PJM footprint, and the closely watched Western Hub benchmark settled at $1,222.75 per megawatt-hour—nearly triple the comparable peak from the summer of 2025. Wholesale power prices at that altitude are not merely a trading-floor curiosity; they ripple through retail supply contracts, utility fuel-cost recovery, and ultimately the monthly bills of households and businesses across a dozen states.

The uncomfortable question is how much of that run-up traces to data center load growth. PJM’s independent market monitor, Monitoring Analytics, has attributed a striking share of recent cost increases to new demand. In its analysis, roughly 63 percent of a major price run-up was tied to data center consumption, translating into approximately $9.3 billion in added costs borne by ratepayers. The forward trajectory sharpens the concern. PJM projects that demand across its territory will grow by 32 gigawatts between 2024 and 2030, with approximately 30 of those gigawatts—the overwhelming majority—attributable to data center expansion. In other words, nearly all of the incremental load the largest U.S. grid expects to serve over the balance of the decade comes from a single, concentrated, and fast-moving source. When a customer category grows that quickly against a supply base that expands slowly, the arithmetic of scarcity does the rest, and wholesale power prices climb during the hours when the system is tightest.

Texas Runs a Different Playbook

The divergence between ERCOT and PJM this summer is instructive because both faced brutal heat and surging demand, yet only one leaned on federal emergency authority. Part of the explanation lies in what Texas had already added to its grid. ERCOT has brought nearly 11,000 megawatts of new capacity online in recent months, much of it from solar and battery storage. On July 21, the state set a fresh solar production record of 34,665 megawatts—a volume of midday generation that arrives precisely when air conditioners are working hardest and the sun is at its most punishing. Batteries then shift a portion of that abundance into the early-evening hours when solar output fades but demand remains high, smoothing the daily ramp that has historically been ERCOT’s moment of greatest vulnerability.

The result is a grid that met record electricity demand with a comfortable supply cushion and no call for conservation. ERCOT’s own seasonal risk assessment reflected the change, pegging the probability of a grid emergency at well under one percent for July, though the operator flagged elevated risk of roughly six percent for August, when solar’s seasonal contribution begins to wane and demand stays stubbornly high. None of this makes the Texas grid invulnerable; the August and September numbers are a reminder that the reliability question is seasonal and dynamic. But the contrast underscores a point that grid reliability planners are increasingly making aloud: the resource mix a region builds in the years before a record-demand summer largely determines whether that summer is a routine operational challenge or a public emergency.

Reliability as a Design Problem, not a Tail Risk

The through-line connecting these events is a shift in how the industry must think about extreme conditions. For decades, planners treated the hottest afternoons and tightest reserve margins as tail risks—rare enough to manage with a modest cushion and the occasional appeal for conservation. The summer of 2026 makes the case that those conditions have moved from the tail of the distribution toward its center, becoming a design baseline rather than an exception. Federal regulators have begun to respond in kind. In June, the Federal Energy Regulatory Commission issued Section 206 show-cause orders directing all six of the nation’s regional grid operators to justify or reform the tariff provisions that govern how data centers and other large loads connect to the system.

FERC’s directives sketch the outline of a new interconnection regime built for the age of data center load growth. The Commission floated a standardized definition of a large load as any facility with a peak capacity of 50 megawatts or more, pressed for study procedures that conclude within 60 to 90 days, and demanded greater transparency in how the cost of network upgrades is assigned and recovered. It also opened the door to flexible transmission services for customers willing to curtail their usage during system peaks—an idea that dovetails neatly with the behind-the-meter backup fleet that emergency orders have started to tap. Several commissioners signaled that if the grid operators do not move quickly, FERC is prepared to take a more prescriptive approach. The message to industry is that the regulatory posture toward large-load interconnection is hardening from invitation into obligation, and that reliability rules once tailored to conventional customers are being rewritten around the realities of hyperscale computing.

What the Summer Is Teaching the Industry

For utilities, developers, and large energy buyers, the practical lessons of this summer are already coming into focus. The first is that speed of supply now matters as much as scale. Regions that fared best had already added dispatchable and renewable capacity, paired with storage, before the heat arrived; regions that struggled were those whose load outran their supply and whose maintenance schedules left them exposed. The second lesson is that data center load growth is no longer a forecasting abstraction. It is showing up in real time in reserve margins, emergency orders, and wholesale power prices, and it is doing so faster than transmission and generation can be permitted and built. The gap between how quickly a hyperscale campus can be energized and how slowly a transmission line can be approved has become the central tension of grid planning.

The third lesson concerns the social contract around reliability and cost. When an independent monitor attributes billions of dollars in added ratepayer costs to a specific and concentrated source of demand, the question of who pays for the grid that serves that demand becomes unavoidable. FERC’s push for transparent cost allocation is an early attempt to answer it, but the politics of the answer—balancing economic development, affordability, and fairness to existing customers—will occupy state commissions and legislatures for years. The behind-the-meter generation that data centers already own may prove to be part of the solution, turning some of the largest new loads into flexible resources that support the grid at its moments of greatest need rather than straining it.

Conclusion

The summer of 2026 will be remembered less for any single record than for the operating pattern it revealed. Extreme heat exposed how thin the margins have become, but heat alone did not create the strain; it amplified a load-growth story already underway and accelerating. Data center expansion is pushing electricity demand into a new baseline faster than generation and transmission can expand, and the mismatch is showing up in near-record peaks, repeated PJM emergency orders, and wholesale prices that briefly tripled. The regions that navigated the season most smoothly did so because they had already treated reliability as a design problem, investing in generation, storage, and flexibility before the crisis rather than during it.

As FERC rewrites the rules for large-load interconnection and the DOE grows more comfortable using emergency authorities, the industry’s task is to internalize the summer’s core message: grid reliability in the era of hyperscale computing will be earned in the planning years, not improvised in the heat. Utilities and grid operators that build for the new baseline—rather than hoping old tail risks stay in the tail—will be best positioned to keep the lights on and prices in check through the record summers still to come.