The Data Center Rebellion

For most of my career, I have worked in and around data centers.

To me, they have never seemed particularly mysterious.

Rows of racks. Servers. Storage. Networking. Power distribution. Cooling systems. Generators. UPS systems. Fiber. Security. An enormous amount of engineering designed around one objective: keep the digital infrastructure running.

So I was genuinely surprised the first time I came across a local news story about residents protesting the construction of a data center.

I didn’t think much of it.

Then recently, I saw my own nephew at an anti-data-center protest.

That got my attention.

My first reaction was confusion. Why would anyone protest a data center?

I have spent much of my professional life looking at these facilities as infrastructure. They represent investment, construction, connectivity, computing capacity and economic activity. They are the physical buildings behind services that most of us use every day without thinking about where those services actually live.

But rather than immediately dismiss the protesters, I wanted to understand their argument.

And some of their concerns are legitimate.

Communities across the country are asking increasingly serious questions about electricity prices, grid capacity, water consumption, noise and environmental impact as enormous new data-center campuses are proposed.

Those questions deserve answers.

But after looking more closely at the debate, I believe we are beginning to confuse two very different things:

Problems caused by poorly planned infrastructure and problems inherent to data centers themselves.

That distinction matters.

Because America’s response to the data-center boom shouldn’t be to stop building the infrastructure of the digital economy.

It should be to build it better.

The Data Center Suddenly Became the Factory Next Door

For decades, most people barely noticed data centers.

They were anonymous buildings sitting in industrial parks.

Then AI changed the scale of the industry.

The amount of electricity consumed by U.S. data centers reached roughly 176 terawatt-hours in 2023, about 4.4% of total U.S. electricity consumption. Lawrence Berkeley National Laboratory’s latest projections suggest data centers could account for between 9.5% and 15.3% of U.S. electricity consumption by 2030.

That is an extraordinary amount of infrastructure.

And communities are understandably noticing.

But there is another reason these facilities are becoming so large:

We keep asking them to do more.

Every ChatGPT prompt has to execute somewhere.

Every cloud-hosted application runs somewhere.

Every streamed movie, enterprise database, AI inference request, online game, financial transaction, social-media feed and cloud backup ultimately depends on physical computing infrastructure.

“The cloud” has always been a misleading abstraction.

The cloud is somebody else’s data center.

And increasingly, AI requires a lot of them.

Your Electric Bill Is a Legitimate Concern

One of the strongest arguments against massive data-center developments is electricity.

A hyperscale facility can represent an extraordinary new electrical load. Utilities may need additional substations, transformers, transmission infrastructure and generation capacity to serve it.

Residents reasonably ask:

Why should my electric bill increase so a trillion-dollar technology company can build an AI campus down the road?

It shouldn’t.

This is where I agree with the protesters.

But I disagree with the conclusion that sometimes follows.

The solution isn’t necessarily to reject the data center.

The solution is to properly allocate its infrastructure costs.

The Department of Energy has explicitly identified fair allocation of electricity-system costs as one of the central questions surrounding large data-center loads. Regulators are examining special tariffs and agreements designed to prevent the financial risks associated with these projects from being shifted onto ordinary customers.

In March 2026, several of America’s largest AI and cloud companies went even further, signing a federal Ratepayer Protection Pledge under which they agreed to build, bring or buy generation and cover the power-delivery infrastructure upgrades required by their data centers rather than passing those expenses to households.

That principle makes sense to me.

If your facility requires a new substation, transmission upgrade or dedicated generation capacity, those costs should be reflected in the economics of your project.

Make the data center pay for the infrastructure the data center requires.

That is a regulatory problem with a regulatory solution.

It isn’t an argument against computing.

The Grid Problem Is Bigger Than Data Centers

There is also an uncomfortable reality hiding underneath this debate.

America needs a better electrical grid regardless.

The Department of Energy’s 2026 transmission study describes a grid confronting rapidly increasing demand not only from data centers, but from advanced manufacturing, large industrial loads, electrification and broader economic growth.

In other words, data centers aren’t creating every weakness in America’s electrical infrastructure.

They are exposing weaknesses that already exist.

If one large industrial customer can overwhelm a regional grid, perhaps the question shouldn’t only be:

“Why is that customer using so much electricity?”

We should also ask:

“Why haven’t we built enough generation and transmission infrastructure for a growing twenty-first-century economy?”

We wouldn’t respond to an interstate becoming congested by deciding that economic activity must stop.

We would eventually have to build more transportation capacity.

Compute infrastructure deserves similar thinking.

The Water Argument Is More Complicated

Water deserves more nuance.

I initially viewed the argument skeptically because many modern computing systems use closed-loop liquid cooling. The coolant circulating through servers can remain inside a sealed loop rather than constantly consuming fresh water.

But that isn’t the whole system.

A closed server-side cooling loop still has to reject its heat somewhere. Some facilities ultimately transfer that heat into cooling towers, where evaporation consumes water. DOE documentation specifically describes this architecture and acknowledges that cooling towers can require substantial makeup water.

Other designs can significantly reduce direct water consumption.

So asking whether “data centers use too much water” is almost the wrong question.

We should ask:

What cooling architecture is this particular facility using, where is it being built, and what water resources are available there?

A water-intensive cooling design might make little sense in a drought-prone region.

The same design could create substantially less concern somewhere with abundant water resources.

This is fundamentally a siting and engineering problem.

Municipalities should be able to require developers to demonstrate sustainable water availability, publish realistic consumption estimates, use reclaimed water where practical and adopt lower-water cooling technologies when local conditions demand them.

What doesn’t make sense is treating every data center as environmentally identical.

They aren’t.

And Yes, They Can Be Loud

The noise complaints also shouldn’t be mocked.

A data center contains enormous mechanical systems. Cooling equipment, fans, pumps, transformers and backup generators can produce persistent sound.

If someone bought a quiet suburban home and suddenly finds themselves listening to industrial equipment every night, telling them they simply don’t understand technology isn’t much of an answer.

But again, this is why zoning exists.

Industrial infrastructure belongs in locations capable of supporting industrial infrastructure.

Cities approving these developments should establish enforceable acoustic limits, appropriate setbacks, sound barriers and operating requirements before construction begins.

If existing zoning allows an enormous facility to sit unreasonably close to residential neighborhoods, then perhaps the zoning ordinance deserves another look.

That is local governance doing what local governance is supposed to do.

The answer doesn’t have to be banning the underlying technology.

Climate Change Is an Energy Problem Before It Is a Data-Center Problem

Then there is carbon.

Data centers consume electricity.

A lot of it.

But a server does not know whether the electrons powering it originated from coal, natural gas, nuclear, hydroelectric, solar or wind generation.

The carbon intensity of computing is therefore deeply connected to the carbon intensity of the electrical system supplying it.

That doesn’t absolve data-center operators of responsibility. Their enormous purchasing power can influence what generation gets built, and companies making environmental commitments should be held accountable for them.

But if our underlying grid remains carbon-intensive, blaming the newest large electricity consumer does not solve the fundamental problem.

The larger challenge is producing enormous quantities of reliable, affordable and increasingly low-carbon electricity.

Because data centers will not be the last source of increasing electricity demand.

Factories need electricity.

Electric vehicles need electricity.

Electrified heating needs electricity.

Semiconductor fabs need electricity.

AI needs electricity.

An advanced economy should not make energy scarcity its industrial policy.

It should figure out how to produce more energy.

But What Do We Get in Return?

This is where I think the data-center debate becomes too focused on what these facilities consume and not enough on what they enable.

The direct employment argument is actually weaker than advocates sometimes suggest.

Data centers can employ large numbers of construction workers while being built, but once operating, highly automated facilities may employ far fewer permanent workers than a comparably sized traditional factory.

We shouldn’t pretend otherwise.

Their economic value comes from something larger.

They represent massive capital investment. They can expand local tax bases. They drive demand for electrical and fiber infrastructure. They support construction, engineering, maintenance and specialized technical ecosystems.

More importantly, they create compute capacity.

And compute is becoming an industrial resource.

AI training requires compute.

AI inference requires compute.

Scientific simulation requires compute.

Modern financial systems require compute.

Cloud computing requires compute.

Defense systems increasingly require compute.

Biotechnology, autonomous systems, advanced manufacturing and materials research increasingly depend on compute.

Data centers aren’t merely warehouses containing computers anymore.

They are becoming factories for computation.

Compute Is Becoming Strategic Infrastructure

This is the part of the debate that concerns me most.

The United States is simultaneously discussing AI leadership while communities across the country are attempting to stop the physical infrastructure required to run AI.

Those positions eventually collide.

You cannot lead the world in artificial intelligence without processors.

Processors require power.

They require cooling.

They require networking.

And ultimately, they require buildings somewhere.

The Department of Energy now explicitly discusses growing data-center and AI demand in the context of American energy security, industrial capacity and national competitiveness.

That isn’t hyperbole.

Imagine a future in which the United States designs the world’s best AI accelerators but cannot connect enough electrical capacity to operate them.

Or American companies develop frontier models but increasingly have to locate their computing infrastructure elsewhere because domestic projects cannot get approved.

Or scientific institutions wait for scarce compute while competitors overseas continue expanding capacity.

At that point, permitting isn’t just a local development question.

It becomes a strategic constraint.

We Shouldn’t Ignore Communities

None of this means communities should surrender control over what gets built around them.

Quite the opposite.

If a company wants to build a multibillion-dollar computing campus, communities should negotiate aggressively.

Protect residential ratepayers.

Require infrastructure upgrades.

Set water-consumption standards.

Enforce noise limits.

Demand responsible zoning.

Protect groundwater.

Require transparent environmental assessments.

Negotiate tax agreements that actually benefit the community.

And make developers demonstrate that the infrastructure supporting their facilities is resilient.

Those aren’t anti-technology positions.

They are competent governance.

But there is an enormous difference between saying:

“You can build here, but you must pay the true cost of your infrastructure.”

and saying:

“We shouldn’t build these facilities at all.”

The first position can produce better infrastructure.

The second could eventually produce technological dependence.

America Has Been Here Before

Every industrial era has required infrastructure that communities initially struggled to accommodate.

Railroads changed towns.

Factories changed cities.

Highways displaced neighborhoods.

Power plants transformed landscapes.

Transmission lines crossed thousands of miles of private land.

Telecommunications towers appeared on skylines.

Some of those projects caused genuine harm, and history gives communities good reason to scrutinize promises of economic development.

We should learn from those mistakes.

But we should not conclude from them that infrastructure itself is optional.

The digital economy has physical requirements just like the industrial economy did.

The difference is that much of the infrastructure was previously invisible to consumers.

AI is making it visible.

The Data Center Rebellion Is Asking the Right Questions—and Sometimes Reaching the Wrong Conclusion

Seeing my nephew at that protest changed my perspective.

Not because it convinced me that data centers are bad.

It made me realize that people living beside these projects are asking questions the technology industry cannot afford to dismiss.

Who pays for the electricity infrastructure?

Where does the water come from?

How much noise will the facility produce?

What does the community actually receive in return?

Those are fair questions.

And the technology industry should have good answers.

But I also think we need to ask one more:

What happens if America stops building the infrastructure required for the next generation of computing?

That cost won’t appear immediately on anyone’s electric bill.

It will appear slowly.

In lost investment.

In constrained AI capacity.

In research performed somewhere else.

In manufacturing advantages surrendered to competitors.

In companies choosing countries capable of supplying the power and infrastructure they need.

And eventually, perhaps, in technological capabilities that other nations possess and we do not.

Data centers should pay their own way.

They should respect the communities that host them.

They should use water responsibly.

They should meet reasonable environmental and noise standards.

And municipalities should absolutely hold developers accountable.

But America should be extremely careful about allowing legitimate local infrastructure problems to evolve into a national rejection of data-center development.

We are no longer merely deciding where to put server racks.

We are deciding where the computational infrastructure of the twenty-first century will be built.

If America decides it doesn’t want that infrastructure in its backyard, someone else will be more than happy to build it in theirs.