There is something strange happening in America’s industrial landscape.
Communities are being asked to welcome enormous facilities that can require extraordinary amounts of electricity, substantial cooling infrastructure, new substations, transmission upgrades and, depending on the cooling design, significant water resources.
What are these facilities producing?
Not gasoline.
Not steel.
Not medicine.
Not electricity.
Not food.
They’re producing computing power.
More specifically, many of the newest projects are being built to support artificial intelligence.
And that raises a question that would have sounded ridiculous a decade ago:
How much of America’s electricity, water and industrial infrastructure should we dedicate to making AI more powerful?
This isn’t an argument that data centers should disappear. Modern society depends heavily on digital infrastructure, and artificial intelligence could produce enormous economic and technological benefits.
But when a massive AI campus arrives in a community and demands electricity on an industrial scale, potentially needs water for cooling, and triggers construction of new energy infrastructure, residents have every right to ask:
Is this really the best use of those resources?
AI Doesn’t Live in the Cloud—It Lives in an Industrial Facility
The technology industry has spent years teaching people to think about computing as something almost invisible.
Everything happens “in the cloud.”
But the cloud has foundations.
It has transformers.
It has switchgear.
It has generators.
It has chillers.
It has pumps.
It has cooling towers in some designs.
It has miles of cable.
It has piping.
And increasingly, it has an enormous appetite for electricity.
A modern AI data center is not simply an office building containing computers.
At sufficient scale, it begins looking like an industrial energy facility with computers sitting in the middle of it.
The servers may be digital.
Everything keeping those servers alive is physical.
The Electricity Has to Come From Somewhere
Every AI query ultimately requires physical equipment to perform calculations.
Training sophisticated AI models can require enormous computing resources.
Operating those models for millions of users requires additional computing.
As AI adoption expands, companies need more servers.
More servers require more electricity.
And electricity does not appear because somebody plugged another rack into the wall.
It has to be generated.
Natural gas turbines may have to run.
Nuclear reactors may provide part of the supply.
Solar farms may be constructed.
Wind farms may contribute.
Battery storage may be added.
Transmission systems may need expansion.
Substations may have to be built.
Transformers have to be manufactured.
The artificial intelligence revolution is therefore becoming an energy infrastructure revolution.
Here’s the Uncomfortable Question
Electricity is not an unlimited resource.
Neither is transmission capacity.
Neither is water.
When one industry begins demanding enormous quantities of infrastructure, society eventually has to decide how that industry fits among competing priorities.
Hospitals need electricity.
Factories need electricity.
Homes need electricity.
Water-treatment plants need electricity.
Refineries need electricity.
Manufacturing plants need electricity.
Transportation is becoming increasingly electrified.
Then along comes an enormous AI data center potentially requesting hundreds of megawatts—or more across a large campus.
So the controversial question becomes unavoidable:
Should making AI models smarter receive the same infrastructure priority as manufacturing physical products people directly need?
There is no easy answer.
But pretending the tradeoff doesn’t exist isn’t an answer either.
What Exactly Are We Getting in Return?
When a steel mill consumes enormous quantities of energy, the output is obvious.
Steel.
When a refinery consumes energy, society receives fuels and chemical feedstocks.
When a manufacturing plant consumes electricity, physical products leave the building.
When a hospital consumes electricity, patients receive medical care.
The output of an AI data center is harder to see.
Algorithms become more capable.
Companies receive computing capacity.
Businesses automate tasks.
Consumers generate images, analyze information, write software, search documents and use digital assistants.
Some of those applications could become extraordinarily valuable.
AI may accelerate scientific research, engineering, medicine, manufacturing and productivity.
But some computing is also used for much less consequential purposes.
And that’s where the argument becomes uncomfortable.
Should a Power Plant Be Built So AI Can Generate Memes?
That sounds intentionally provocative.
Because it is.
But underneath the exaggeration is a legitimate policy question.
Not every computation has equal social value.
The same data center capable of supporting cancer research can also generate millions of novelty images, advertising variations, social-media content and disposable digital entertainment.
Electricity doesn’t know the difference.
A megawatt is a megawatt.
Cooling equipment doesn’t know whether the GPU is analyzing a protein structure or generating somebody’s cartoon profile picture.
Infrastructure simply supplies the demand.
As AI becomes embedded into increasingly trivial applications, society may eventually have to confront whether unlimited computing consumption should automatically be treated as economically desirable.
But Who Gets to Decide What’s “Wasteful”?
Here is where the anti-data-center argument runs into trouble.
Who decides?
Government?
Utilities?
Politicians?
Local residents?
Technology companies?
A regulator might consider generating an AI video wasteful.
The person building a business with that technology may consider it economically transformative.
People once criticized the electricity consumed by television, air conditioning, personal computers and the internet.
Technologies that initially appear frivolous can become fundamental infrastructure.
Artificial intelligence could follow the same trajectory.
Trying to centrally determine which calculations deserve electricity could create problems far worse than the electricity consumption itself.
That doesn’t eliminate the resource question.
It makes it harder.
Then There’s Water
Electricity receives most of the attention, but water has become another controversial part of data center development.
Servers produce heat.
That heat has to go somewhere.
Data centers use different cooling technologies, and their direct water requirements vary substantially depending on design, climate and operating strategy.
Some facilities rely heavily on air-based or closed-loop systems.
Others use evaporative cooling methods that can consume significant amounts of water.
The important point is that not every data center has the same water footprint.
But when a large project does require substantial water consumption, the location matters enormously.
Using water in a water-rich region is one thing.
Building a water-intensive facility in a drought-prone area is another.
What Happens When the Community Needs the Same Water?
This is where the conversation can become heated very quickly.
Imagine a community being asked to conserve water.
Residents face watering restrictions.
Farmers worry about groundwater.
Population continues growing.
Then a multibillion-dollar technology company proposes an enormous data center campus.
Residents are naturally going to ask:
Why are we being told to conserve while one industrial customer receives enough water to cool thousands of servers?
Developers may respond that modern facilities can use reclaimed water, closed-loop cooling, air cooling or other technologies designed to reduce potable-water consumption.
Those solutions matter.
But communities should demand specifics rather than slogans.
How much water will actually be consumed?
Where will it come from?
What happens during drought conditions?
How will future expansion affect demand?
Those are reasonable questions.
AI Companies Are Rich Enough to Solve This
This may be the most controversial argument of all.
Some of the companies driving the AI boom are among the wealthiest corporations in human history.
If they require enormous amounts of electricity and water, why should communities accept the cheapest infrastructure solution available?
Why shouldn’t AI companies be expected to build better infrastructure?
If water is scarce, invest in cooling systems that minimize water consumption.
If electricity supply is constrained, finance new generation.
If transmission capacity is insufficient, help pay for expansion.
If backup generation creates emissions concerns, invest in cleaner alternatives where technically practical.
If the grid needs batteries, build batteries.
If nuclear power is needed, help finance nuclear power.
If natural gas generation is required, pay for the infrastructure necessary to support it.
The companies demanding extraordinary resources have extraordinary financial resources.
Make them use them.
Don’t Ask a Working Family to Subsidize a GPU Farm
This is where the electricity and water debates collide with economics.
Imagine a household struggling with utility bills.
Now imagine that household being told billions of dollars in infrastructure investment is necessary because enormous AI campuses are connecting to the regional grid.
The obvious question becomes:
Why should that family subsidize infrastructure primarily required by companies worth hundreds of billions—or even trillions—of dollars?
Large customers can absolutely provide economic benefits to utilities and communities.
Their electricity purchases can create revenue.
Their investments can expand infrastructure.
Their taxes can support local governments.
But those benefits depend heavily on how contracts, utility rates, infrastructure costs and tax incentives are structured.
A bad deal can socialize costs while privatizing profits.
A good deal can make the technology company pay for the infrastructure its growth requires.
That difference matters.
And Then There Are the Tax Breaks
The controversy gets even sharper when public incentives enter the picture.
A community can find itself offering tax advantages to attract a facility that simultaneously requires enormous amounts of electricity, land and infrastructure.
Supporters argue that incentives attract investment that might otherwise go somewhere else.
Critics ask whether communities are competing against each other to give increasingly favorable deals to companies that already possess enormous financial resources.
The argument shouldn’t simply be:
“Data centers are good.”
Or:
“Data centers are bad.”
The better question is:
What is the community receiving compared with what it is giving up?
The Permanent Job Numbers Matter
Construction can employ thousands of people.
That matters enormously to the skilled trades.
But once construction finishes, a highly automated data center may employ considerably fewer permanent workers than were required to build it.
That doesn’t make the construction employment meaningless.
Construction workers build temporary projects for a living.
But communities evaluating water allocations, electrical infrastructure and tax incentives should distinguish between temporary construction employment and permanent operational employment.
A $10 billion facility can represent enormous investment without becoming a $10 billion permanent payroll.
That difference should be understood before incentives are approved—not afterward.
Now Look at the Other Side: AI Could Build America’s Next Energy System
Here is where this story takes an unexpected turn.
What if AI’s enormous electricity appetite becomes exactly what America’s aging electrical infrastructure needed?
Utilities have long faced difficulty financing enormous projects without confidence that customers will exist to buy the electricity.
Data centers provide something energy developers love:
Large, concentrated, long-term demand.
That demand can support new generation.
It can justify transmission investment.
It can accelerate substation construction.
It can create markets for new technologies.
Suddenly, AI’s energy appetite stops looking purely like a problem.
It starts looking like a financing mechanism for America’s next energy buildout.
AI Could Help Bring Nuclear Power Back
Nuclear power becomes particularly interesting in this discussion.
Data centers want enormous amounts of reliable electricity around the clock.
Nuclear plants produce large quantities of continuous electricity.
Those characteristics fit together unusually well.
If technology companies become willing to support nuclear development financially, AI demand could contribute to reactor restarts, life extensions and potentially future nuclear construction.
That could create major opportunities for the industrial workforce.
Boilermakers.
Pipefitters.
Welders.
Millwrights.
Electricians.
Instrumentation technicians.
NDT inspectors.
Radiation protection technicians.
Crane operators.
Riggers.
Engineers.
Operators.
The machines threatening some white-collar jobs could ironically help create enormous amounts of blue-collar infrastructure work.
Natural Gas Could Become AI’s Unexpected Partner
Natural gas presents another contradiction.
The technology industry often emphasizes clean-energy goals.
But AI facilities need reliable power continuously.
Solar and wind can contribute enormous amounts of electricity, but their output varies with weather and time.
Battery storage can help bridge those periods.
Dispatchable generation remains valuable when continuous reliability is required.
That can increase interest in natural gas generation.
And natural gas generation requires an entire industrial ecosystem.
Gas production.
Pipelines.
Compressor stations.
Processing facilities.
Power plants.
Turbines.
Pumps.
Valves.
Instrumentation.
Maintenance.
A technology that feels completely digital may ultimately create more work for pipeline welders, pipefitters and millwrights.
That’s quite a twist.
AI Could Become One of the Biggest Skilled-Trades Employers Without Employing the Trades Directly
The technology companies themselves may never employ most of these craftsmen.
Their contractors will.
Utilities will.
Power producers will.
Pipeline companies will.
Electrical contractors will.
Mechanical contractors will.
Engineering and construction firms will.
The AI industry’s economic footprint can therefore extend far beyond the people working inside data centers.
An electrician may never touch a server while spending ten years building infrastructure required because those servers exist.
A pipeline welder may never enter a data center while welding the natural gas infrastructure feeding a power plant supplying one.
A millwright may never write a line of code while aligning a turbine generating electricity for AI.
That’s how interconnected modern industry has become.
Water Could Create Its Own Skilled-Trades Boom
Reducing water consumption also requires infrastructure.
Advanced cooling systems need mechanical equipment.
Water-recycling systems require pumps, piping, valves, filtration and controls.
Wastewater treatment requires operators and technicians.
Cooling plants require maintenance.
Environmental requirements create inspection and engineering work.
If communities demand better water management from data center developers, the solution doesn’t necessarily mean stopping construction.
It can mean building more sophisticated infrastructure.
And sophisticated infrastructure creates skilled-trade work.
Maybe the Problem Isn’t Data Centers—It’s Cheap Resources
For decades, industries have located projects partly according to the cost and availability of resources.
Cheap electricity attracts energy-intensive manufacturing.
Abundant natural gas attracts chemical plants.
Transportation infrastructure attracts distribution centers.
Water availability influences manufacturing and agriculture.
Data centers are no different.
If electricity or water is genuinely scarce in a particular region, prices and infrastructure requirements should reflect that scarcity.
The controversial alternative is pretending resources are abundant, offering incentives anyway, and then asking everybody else to absorb the consequences.
Data centers should compete honestly for resources.
If the economics still work, build them.
Communities Should Stop Acting Desperate
The AI boom has created intense competition between states and municipalities.
Everybody wants the next giant project announcement.
Politicians want ribbon cuttings.
Economic-development agencies want billion-dollar investment headlines.
Contractors want construction.
Workers want jobs.
Utilities want customers.
But communities possess something technology companies desperately need:
Land, electricity, water and infrastructure.
That gives communities negotiating power.
Instead of asking, “What incentives do we need to offer so you’ll build here?”
Perhaps the question should sometimes be:
“What are you going to build for us if we allow you to consume resources at this scale?”
That completely changes the negotiation.
Make the Deal Benefit Everyone
A responsible data center agreement could potentially deliver far more than a building full of servers.
Developers can help fund substations.
They can support transmission expansion.
They can invest in new generation.
They can use reclaimed water where practical.
They can finance water infrastructure.
They can contribute to apprenticeship programs.
They can hire local contractors.
They can support skilled-trades training.
They can structure electricity agreements to reduce the risk of costs shifting onto residential customers.
That turns a potentially extractive relationship into an infrastructure partnership.
The Skilled Trades Should Be at the Center of This Conversation
Too much discussion about artificial intelligence focuses exclusively on software engineers and technology executives.
But AI’s physical expansion increasingly depends on people who may never write code.
Electricians.
Pipefitters.
Welders.
Millwrights.
Boilermakers.
Ironworkers.
Scaffold builders.
Industrial insulators.
Instrumentation technicians.
HVAC technicians.
Sheet metal workers.
Linemen.
Equipment operators.
Riggers.
Crane operators.
NDT technicians.
Commissioning specialists.
Power plant operators.
These workers are building the physical world required for the digital one.
And if America decides to expand electricity generation, transmission, cooling and water infrastructure to support AI, their role becomes even larger.
So, Should AI Be Allowed to Consume This Much?
Yes—but not without conditions.
Artificial intelligence could become enormously valuable.
Data centers are now essential infrastructure.
America has economic and strategic reasons to remain competitive in advanced computing.
Stopping data center development entirely would create its own economic consequences.
But “AI is important” should not become a magic phrase that ends every discussion about resources.
If a project consumes extraordinary electricity, it should carry an appropriate share of the cost of expanding electricity supply.
If it requires significant water in a constrained region, developers should demonstrate how that demand will be managed.
If communities provide major incentives, residents deserve transparency about permanent employment and economic benefits.
And if technology companies need an unprecedented industrial buildout to make artificial intelligence possible, those companies should help pay for that buildout.
AI Can Have the Power—But It Should Pay the Industrial Bill
America does not necessarily have to choose between artificial intelligence and affordable electricity.
It does not necessarily have to choose between data centers and water security.
And it does not have to choose between technological progress and industrial workers.
But somebody has to pay for the infrastructure connecting all three.
The easiest solution would be to quietly spread those costs around until nobody knows exactly who paid for what.
The better solution is much simpler.
If AI needs another substation, help build it.
If AI needs another transmission line, help pay for it.
If AI needs another power plant, support the generation required to serve it.
If AI needs millions of gallons of water, prove that the supply is sustainable—or invest in technology that dramatically reduces the requirement.
And if communities are going to dedicate land, electricity, water and infrastructure to one of the richest industries the world has ever created, those communities should receive something substantial in return.
Artificial intelligence may very well transform the future.
But the future still runs through transformers, turbines, pumps, pipes, transmission lines and cooling systems.
AI can have the power.
Just don’t send everybody else the bill.
