America's heatwave exposes AI's real energy bill

Introduction: when the heat reveals what AI really costs

A heat dome that tests the whole system

A massive heatwave swept across the United States in early July, pushing heat-index temperatures above 38.9 degrees Celsius in Washington and 37.7 degrees in New York, according to Al Jazeera. This heat dome didn't just make the air unbreathable in several major cities: within just a few days, it exposed the fragility of an electrical grid already under strain from the explosion of AI data centers.

Grid operator PJM Interconnection, which serves 13 states and the U.S. capital, asked the Department of Energy to order data centers to switch to backup generators within 15 minutes of an emergency signal, in order to free up electricity for homes and businesses.

Let's check the numbers driving the public debate

The central claim we're checking here: do AI data centers really consume enough electricity and water to justify the current level of public concern? According to the U.S. Department of Energy, data centers today account for 4% of national electricity demand, a share expected to climb to 9% by 2030. That figure is confirmed and consistently cited by Al Jazeera.

Together, American data centers consume roughly 176 terawatt-hours of electricity every year, enough to power nearly 16 million households, or roughly 40 million people — a population comparable to Canada or California.

What a hyperscale data center really represents

Energy needs on par with entire cities

A typical hyperscale data center requires between 100 and 300 megawatts of electricity, enough to power up to 300,000 American homes, or a city of roughly 750,000 people, according to data reported by Al Jazeera. That's comparable to the population of cities like Nashville, Charlotte, or Edmonton.

This isn't an activist exaggeration: it comes directly from data compiled by researchers and relayed by a recognized international news agency. The fact deserves repeating because it changes the nature of the debate: we're no longer talking about a server in a basement, but about infrastructure whose electricity appetite rivals that of entire metropolitan areas.

A geographic footprint concentrated in a handful of states

According to data compiled by the Pew Research Center and cited by Al Jazeera, 38% of Americans live within eight kilometers of one of the 3,000 data centers currently operating in the United States, most located in the South and Midwest. Virginia leads with 398 operational centers and 287 more planned, while Texas has 296 active and 170 in the pipeline.

In Henrico County, Virginia, already home to 37 data centers, authorities asked schools to cut back on electricity use amid rising grid demand, according to emails obtained by tech publication 404 Media. This isn't an isolated anecdote — it's a documented warning sign.

The water question: an underestimated problem

Consumption that spikes during heatwaves

Data centers' water demand rises during periods of extreme heat, since cooling systems must keep servers at a safe temperature. Many of these systems don't recycle the water they use, and nearly 80% of the drinking water used for cooling simply evaporates, according to Al Jazeera.

Currently, data center water consumption reaches roughly 627 million gallons, or 2.85 billion liters, per day across the United States. That's far less than cattle farming, which consumes 137 billion gallons, but it's a fast-growing figure that's geographically concentrated, unlike the more diffuse consumption from agriculture.

Regions already under water stress targeted for new projects

According to data cited in the report, two-thirds of all new data centers built or in development since 2022 are located in regions already facing water stress. A single large AI data center can consume up to 5 million gallons of water per day, as much as a city of 50,000 people, according to the Environmental and Energy Institute.

In Corpus Christi, Texas, where three data centers are already operating, some reservoirs have dropped below 10% of capacity after five consecutive years of drought. Local officials say a data center project planned an hour north of the city has delayed emergency water supply plans for the coastal municipality.

Public opposition is growing against this expansion

A massive rejection in the polls

According to a recent Gallup poll cited by Al Jazeera, seven in ten Americans oppose building data centers in their local community. Half of these opponents explicitly cite excessive resource use, especially electricity and water, as their main concern.

This level of opposition isn't marginal: it cuts across the usual partisan lines and affects conservative rural areas as much as more progressive suburbs. When seven out of ten citizens oppose the same issue, it becomes difficult to dismiss the reaction as an isolated activist movement.

Elected officials from opposite ends of the spectrum converge

Texas Republican Governor Greg Abbott called for banning data center construction in rural areas during a campaign stop, insisting these facilities generate their own electricity and recycle their water. Meanwhile, Vermont Senator Bernie Sanders and New York Representative Alexandria Ocasio-Cortez have called for a complete moratorium on new data center construction.

When officials this far apart ideologically land on the same position, it should put the tech industry on notice: this is no longer a partisan issue — it has become a public resource management issue that no one can conveniently ignore anymore.

What industry experts are actually saying

A cooling model overtaken by the climate

Arif Gasilov, managing partner of the natural resources and built environment practice at Gasilov Group, told Al Jazeera that "heatwaves are showing that the current siting and cooling model was designed for average conditions, and average conditions are disappearing more every day." He added that when a facility uses evaporative cooling during a heatwave, it draws from the same already-strained water supply as residents, who are often already under restrictions.

Alex MacColl, a project manager at British firm Datamove, added that data centers' reliance on water for cooling arrives at exactly the moment when local supply is scarcest, calling it "a circular problem, not a seasonal inconvenience."

Consequences already visible for ordinary consumers

Earlier this year, roughly 50,000 California customers near Lake Tahoe were told they'd need to find a new electricity provider because of rising demand tied to data centers. These concrete cases show the impact isn't theoretical: it's already hitting ordinary families who never asked to host artificial intelligence infrastructure near their homes.

Utility companies are now being forced to delay retiring aging power plants while accelerating investment in new generation and transmission capacity, a race against time that will inevitably cost taxpayers and residential ratepayers dearly.

How the tech industry is responding to these criticisms

Efficiency promises that remain vague

Facing public pressure, several major tech companies say they're working on more efficient cooling systems and renewable energy purchase agreements to offset their footprint. Some data center operators even claim to use a "zero water" system, as is reportedly the case for one facility near Corpus Christi, according to data reported by Al Jazeera.

But these promises remain hard to verify independently, and Corpus Christi's own local authorities say a nearby project has delayed the city's emergency water supply plans, illustrating the persistent gap between corporate messaging and the reality experienced on the ground by affected communities.

An investment race that shows no sign of slowing

Despite the controversy, no major tech company has announced a significant slowdown in its data center construction plans. Investment keeps pouring in, driven by the conviction that falling behind in the artificial intelligence race means losing a strategic advantage that would be hard to recover, especially against rivals like China.

This relentless-race logic partly explains why environmental and energy concerns are struggling to meaningfully slow the pace of announcements, even as mayors, governors, and senators from opposite camps voice their concern publicly.

The fact-check verdict: a grounded concern, not unjustified panic

The figures cited hold up under scrutiny

After verification, the main figures cited in the public debate — the 4% to 9% share of national electricity demand, the 176 terawatt-hours consumed annually, and the 627 million gallons of water used every day — are all corroborated by credible primary sources, including the Department of Energy and data compiled by Pew Research. This is not a media exaggeration, but a measurable, documented reality.

What sets this heatwave apart from previous ones is that it arrives at the exact moment the AI industry is announcing wave after wave of massive investment, creating an unavoidable collision between technological ambition and the physical limits of existing infrastructure.

What other Western countries are watching closely

Canada and Europe facing the same dilemma

The phenomenon isn't exclusively American: cities like Edmonton, cited directly as a population comparison for a hyperscale data center in Al Jazeera's reporting, show that Canada and other Western countries are closely watching this American trajectory to avoid repeating the same energy-planning mistakes.

Western governments know the competition with China over artificial intelligence won't be won with better models alone, but also with enough energy capacity to run them without triggering a local social crisis. It's a strategic factor few commentators still fully account for in their analysis of the global tech race.

Conclusion: a warning sign it would be unwise to ignore

An energy bill that can no longer be hidden

The July 2026 heatwave didn't invent anything new: it simply made visible, over a few days of extreme heat, a structural tension experts had already been documenting for months. Between PJM Interconnection's request for data centers to switch to backup generators and the call for a moratorium from officials as different as Greg Abbott and Bernie Sanders, the message converges: the expansion of AI data centers now has to reckon with real physical limits.

What to watch in the coming months

The real question is no longer whether data centers consume large amounts of electricity and water — that reality is now established and verified — but how regulators, tech companies, and local communities will negotiate a workable compromise before the next heatwave. The West has every interest in running this technological race with clear eyes rather than haste, since a local energy failure would only benefit its strategic rivals.

By Maxime Marquette, columnist

Columnist's transparency note

Who I am and how I work

I am a columnist and analyst, not an electrical systems engineer or a hydrologist. My job is to gather reliable sources, check their consistency, and formulate a well-argued opinion, not to claim a technical expertise I don't have. On this particular story, I rely mainly on Al Jazeera's reporting from July 3, 2026, which itself cites the Department of Energy, Pew Research, and industry experts.

My acknowledged biases and my limits

I believe the West needs to stay at the forefront of the technology race against China, which colors how I interpret the facts presented here. I don't claim to know every technical solution available for reducing data centers' water and electricity consumption, and I acknowledge this story will keep evolving rapidly in the months ahead.

Sources

Primary sources

Al Jazeera, US heatwave raises alarms over AI data centre energy demands — July 3, 2026

The New York Times, Electric grid data centers PJM — July 2, 2026

Secondary sources

CNN, Heat data centers electricity generators — July 2, 2026

Politico, Data center boom collides with record heat testing power grid — July 1, 2026

Pew Research Center, data on the geographic distribution of data centers — 2026

U.S. Department of Energy, projections on AI-driven electricity demand — 2026