Americans are arguing about data centers.
They’re arguing about water.
They’re arguing about electricity.
They’re arguing about noise, taxes, transmission lines, and whether giant server farms belong in their communities.
The Pentagon is having an entirely different conversation.
It is talking about warfare.
A recently released forecast from Data Center Frontier may be one of the most revealing documents I’ve read this year. Not because it predicts some technological breakthrough, but because it quietly acknowledges something the public is only beginning to understand.
The age of artificial intelligence is becoming an age of infrastructure.
And infrastructure has consequences.
For years, most people viewed data centers as invisible parts of the internet. Large warehouses full of servers that somehow kept email, Netflix, Facebook, Amazon, and cloud services running.
The industry itself is now saying that’s no longer an accurate description.
The report argues that what we casually call “data centers” are increasingly becoming two separate industries.
The first is the traditional data center sector: cloud computing, storage, networking, enterprise applications, and digital services.
The second is something new.
The report calls them AI factories.
That may be the most important phrase in the entire document.
Factories produce things.
Steel mills produce steel.
Shipyards produce warships.
Aircraft plants produce fighters.
AI factories produce intelligence.
Not human intelligence.
Machine intelligence.
And producing it requires staggering amounts of electricity.
The report repeatedly returns to a single conclusion: power is now more important than land.
For decades, developers found a piece of property and then figured out how to get electricity to it.
Today they are looking for electricity first and then finding land nearby.
That single shift explains much of what is happening across the United States.
The private sector is racing to build AI infrastructure at a pace rarely seen in modern industrial development.
Amazon.
Microsoft.
Google.
Meta.
Oracle.
OpenAI and its partners.
Dozens of developers.
Hundreds of proposed projects.
Billions upon billions of dollars flowing into AI infrastructure.
The report notes that some of these AI campuses are planning for hundreds of megawatts of continuous power consumption. Some are approaching the scale of major industrial facilities. Some are being designed around their own dedicated generation sources.
And increasingly, those generation sources may include natural gas plants, fuel cells, batteries, microgrids, and eventually nuclear power.
That last point deserves attention.
The report treats nuclear power almost as a planning assumption.
Not because reactors are being built tomorrow.
Not because Small Modular Reactors are ready to appear next month.
But because developers are already selecting sites, structuring partnerships, acquiring land, and making long-term decisions based on the expectation that future AI infrastructure will require dedicated nuclear generation.
Think about how remarkable that is.
Just a few years ago, technology companies talked about software.
Today, they are discussing reactors.
The conversation has shifted from applications to power plants.
From coding to generation capacity.
From servers to industrial infrastructure.
The report even argues that utilities are no longer simply service providers.
They are becoming co-architects of projects.
Power availability determines whether projects move forward.
Power availability determines where they are built.
Power availability determines how large they become.
Power availability determines whether financing is approved.
Power availability determines almost everything.
That’s not just happening in the private sector.
The same realities are beginning to appear in military planning.
Defense Secretary Pete Hegseth has repeatedly spoken about creating an AI-first military. The services are increasingly exploring artificial intelligence for intelligence analysis, logistics, maintenance, cyber operations, real-time weather forecasting, planning, sensor fusion, autonomous systems, and decision support.
But an AI-first military is about far more than replacing paperwork with algorithms.
It promises faster intelligence analysis, real-time battlefield awareness, predictive logistics, autonomous systems operating at machine speed, cyber defenses that react in seconds instead of hours, and command systems capable of processing more information than any human staff could absorb. Supporters argue that such capabilities could provide American forces with a decisive advantage against near-peer adversaries such as China and Russia. Critics worry about increasing dependence on algorithms, vulnerabilities to cyber attack, and the possibility of machines influencing life-and-death decisions.
Regardless of where one stands in that debate, both sides agree on one thing: none of it happens without computing power.
Massive amounts of computing power.
This requires massive amounts of electricity.
All of that requires computing power.
Massive amounts of computing power.
This requires massive amounts of electricity.
The report doesn’t discuss military bases specifically, but its conclusions point in an obvious direction.
Military installations already possess many of the things AI factories require.
Large secure land holdings.
Established utility connections.
Substations.
Transmission access.
Security perimeters.
Federal ownership.
Communications infrastructure.
Environmental review pathways.
Long-term planning horizons.
In many ways, they are uniquely positioned to host this kind of infrastructure.
Whether they will host it remains an open question.
Whether surrounding communities will welcome it is another.
One of the most striking sections of the report discusses public opposition.
The industry openly acknowledges that AI factories are losing the luxury of invisibility.
For years, most people never thought about data centers.
Now they do.
Utilities are discussing them.
County commissions are discussing them.
State legislatures are discussing them.
Environmental groups are discussing them.
Neighborhood associations are discussing them.
The public is discussing them.
The larger the project, the brighter the spotlight becomes.
That’s already visible across the country.
Communities have demanded moratoriums.
Projects have been delayed.
Projects have been downsized.
Some proposals have been withdrawn altogether.
Questions about water, power, noise, emissions, land use, grid reliability, and economic benefits are becoming routine parts of public hearings.
The industry sees this.
The report sees this.
Investors see this.
And increasingly, government sees it too.
What nobody knows yet is how that growing opposition may affect military installations if large AI campuses begin appearing behind base gates.
Will communities view military AI infrastructure differently than private-sector AI infrastructure?
Will federal projects face the same political resistance?
Will national security arguments override local concerns?
Will military facilities become preferred sites because they already possess security and infrastructure advantages?
Nobody knows.
The report doesn’t claim to know.
Neither do I.
What it does suggest is that America is entering a new phase.
For years, we talked about AI as software.
Now we’re discovering AI is also steel.
Concrete.
Transmission lines.
Cooling systems.
Substations.
Power plants.
And perhaps someday reactors.
The report contains a simple phrase that captures the moment perfectly:
“The frontier has moved inward.”
The challenge is no longer inventing the technology.
The challenge is building the infrastructure necessary to support it.
History offers a useful reminder.
Great powers have always depended on industrial foundations.
The British Empire ran on coal.
The Second World War ran on oil.
The Cold War ran on nuclear deterrence.
The emerging AI era appears increasingly likely to run on electricity.
Lots of it.
Which means the real story may not be the artificial intelligence itself.
The real story may be the infrastructure race being built underneath it.
And whether that race eventually runs through America’s military bases remains one of the most important unanswered questions of the decade. At the same time, the AI age may run on electricity. But in much of the American West, it also may be constrained by water.
Hegseth’s vision of an AI-first military may ultimately succeed or fail for reasons no algorithm can predict. But one lesson from the Data Center Frontier report is already becoming clear. The future of military artificial intelligence may depend less on software engineers and more on electricians, utility planners, transmission engineers, and power and water producers.
Wars have always depended on logistics.
The AI age may simply be replacing fuel depots with substations.