Who Will Power the Data Center Boom?

The backlash against America's data-center boom is becoming difficult to ignore.

Communities across the country are asking reasonable questions about the enormous facilities appearing in their backyards. How much electricity will they consume? Who will pay for the new power plants, transmission lines and substations needed to support them? How much water will they use? And will families who never asked for a data center nearby eventually see the consequences on their utility bills?

These concerns are no longer confined to environmental activists or communities traditionally hostile to development. This week, the House passed the Ratepayer Protection Act by an extraordinary 417 to 3 vote, directing state utility regulators to consider whether large electricity users such as data centers should bear the additional infrastructure costs created by their demands.

The concern is understandable.

Artificial intelligence requires staggering amounts of electricity. The International Energy Agency says electricity consumption from AI-focused data centers increased 50 percent in 2025 alone. It projects total data-center electricity consumption will roughly double by the end of the decade, with AI-focused facilities growing even faster.

In the United States, the consequences will be particularly significant. The IEA estimates that data centers could account for nearly half of the growth in American electricity demand through 2030.

Then there is water.

The massive computing systems behind artificial intelligence generate tremendous amounts of heat, and many data centers rely heavily on water for cooling. In Texas, where data-center development has exploded, state officials are now demanding greater disclosure of water consumption and threatening penalties for facilities that fail to comply.

People have every right to ask whether the AI revolution is going to arrive with higher electric bills, strained water supplies and an increasingly fragile power grid.

But those questions should lead us toward a larger one.

What if the extraordinary demands of artificial intelligence are exposing a weakness America needed to confront anyway?

For decades, electricity demand in the United States barely grew. Our energy infrastructure increasingly reflected that reality. We learned to conserve, regulate and manage scarcity while the construction of major new power projects became slower, more expensive and politically difficult.

AI has suddenly changed the equation.

The computing infrastructure now being built requires enormous quantities of reliable electricity around the clock. And the scale of that demand is forcing technology companies to think about energy in ways Silicon Valley largely never had to before.

One of the more interesting examples is unfolding in Utah.

Valar Atomics is a young American nuclear company developing small, high-temperature reactors designed to produce reliable power close to where that power is actually needed. In June, its Ward 250 reactor achieved criticality. Shortly afterward, Valar used electricity from the reactor to power NVIDIA's Blackwell AI technology in a proof-of-concept demonstration.

Now Valar and NVIDIA are exploring something considerably more ambitious: a 30-megawatt AI facility powered directly by advanced nuclear energy.

The implications go beyond electricity.

Valar's reactor is cooled with helium rather than water. NVIDIA, meanwhile, has developed a closed-loop liquid-cooling system for its newest data-center architecture. The company says the technology can reduce facility cooling-water consumption from roughly 2.6 million gallons per megawatt per year to near zero.

There are important caveats. A demonstration is not the same thing as commercially deploying reactors at scale. Advanced nuclear companies still have to prove they can build reliably, navigate regulatory requirements and produce electricity at competitive prices. And reducing water consumption inside a data center does not eliminate water use throughout the semiconductor and electricity supply chains.

Still, something important is happening here.

Two of the most difficult objections to the AI buildout, enormous electricity demand and enormous water consumption, are becoming engineering problems that companies have an enormous financial incentive to solve.

That is a much healthier way to think about the data-center debate.

The United States is entering an era in which electricity abundance may once again be a strategic advantage. AI is one reason. Advanced manufacturing is another. Semiconductor fabrication, electrification and the reshoring of industrial production all require dependable power. America's ability to build things increasingly depends upon its ability to generate electricity.

For years, much of our energy debate assumed that reducing consumption was itself a worthy objective. Efficiency certainly matters. Waste is expensive, and technological improvements that accomplish more with less should be welcomed.

But conservation cannot be America's primary energy strategy in an era of rapidly expanding technological demand.

We need more power.

That should mean taking another serious look at nuclear energy, including the generation of smaller reactors now attempting to move from experimental projects into commercial deployment. It should mean faster permitting for new generation and transmission. It should also mean expecting the companies creating extraordinary new electricity demand to participate in building the infrastructure necessary to supply it, rather than simply passing those costs along to existing ratepayers.

There is nothing inevitable about today's data centers remaining as resource-intensive as they are now.

The history of technology is filled with machines that became smaller, faster, cheaper and more efficient because the economics demanded it. AI hardware is already becoming more efficient. Cooling technology is improving. Energy companies now have some of the largest and wealthiest customers in the world actively searching for new sources of reliable power.

That combination creates an opportunity.

The data-center backlash is revealing something larger than a dispute over where server farms should be built. It is exposing the consequences of an America that has spent too long thinking about electricity as a scarce resource to be carefully allocated instead of an essential resource whose supply can be dramatically expanded.

AI may make that old approach impossible to sustain.

Good.

If America's technological ambitions require vastly more electricity, we should build vastly more electricity. If data centers consume too much water, we should develop technologies that consume less. If the existing grid cannot accommodate the next generation of American industry, we should expand the grid and find ways to generate power outside it.

The companies that solve those problems will not merely make artificial intelligence possible. They could leave America with something much more valuable: a new generation of abundant, reliable energy infrastructure capable of powering whatever comes next.

Perhaps the greatest contribution of the AI boom will turn out to be something few people expected.

It may force America to remember how to build.