ABE NEWS | September 3, 2026
America has spent the past several years racing to build the physical infrastructure required for artificial intelligence.
Now communities are beginning to ask a different question:
Who is paying for it?
Across the United States, enormous AI data centres are colliding with growing concerns over electricity prices, water consumption, land use, tax incentives and the relatively small number of permanent jobs some facilities create.
And nowhere is that collision becoming more politically important than Texas.
Less than a year ago, Texas Governor Greg Abbott described his state as the “epicenter of AI development.”
Texas had everything the AI industry wanted.
Land.
Energy.
Business-friendly regulations.
A growing technology sector.
And political leaders eager to attract billions of dollars of investment.
But the political environment has changed dramatically.
Abbott is now pushing tougher restrictions on data centres, demanding that developers pay for the electricity infrastructure they require, disclose their water and power needs and reduce their impact on surrounding communities.
He has also called for outdated data-centre tax incentives to be phased out.
And Texas has effectively paused approvals for new data-centre grid connections while regulators audit proposed projects.
The change is not occurring in isolation.
Republican candidates in at least five U.S. states, including Michigan, Wisconsin and Pennsylvania, have recently adopted tougher positions toward data-centre development or proposed restrictions.
Democrats have also increasingly targeted the industry.
What began as scattered local opposition is developing into something considerably larger:
a political backlash against the physical infrastructure powering the AI revolution.
And it could become one of the biggest obstacles facing America’s AI boom.
THE AI REVOLUTION NEEDS AN EXTRAORDINARY AMOUNT OF ELECTRICITY
Artificial intelligence can feel almost weightless.
A user opens an application.
Types a question.
Receives an answer seconds later.
But behind that interaction sits enormous physical infrastructure.
Servers.
Graphics processors.
Networking equipment.
Cooling systems.
Power substations.
Transmission lines.
Backup generators.
Water systems.
And enormous buildings filled with computing equipment.
Training and operating increasingly powerful AI systems requires vast amounts of electricity.
As companies including Microsoft, Google, Amazon, Meta and OpenAI race to expand their computing capacity, the demand for data centres has exploded.
That has created a new problem for American electricity grids.
Data-centre developers are requesting astonishing amounts of power.
A Reuters review of utility and grid data found that requests from very large electricity users — mostly proposed data centres — now exceed 700 gigawatts across parts of the Midwest, Mid-Atlantic and South.
For perspective, that is more than ten times estimates of the electricity currently consumed by all U.S. data centres.
Texas provides perhaps the clearest example.
Since 2023, requests from data centres and other enormous electricity users seeking connections to the Texas grid have jumped from approximately 48 gigawatts to more than 474 gigawatts.
That does not mean Texas is about to consume 474 gigawatts of additional electricity.
In fact, that is part of the problem.
Nobody knows how much of it is real.
THE “GHOST DEMAND” PROBLEM
The AI boom has created a scramble for access to electricity.
Developers know that securing power can determine whether a data-centre project succeeds.
Land near large grid connections has become enormously valuable.
Companies therefore have an incentive to reserve electricity capacity early.
But some developers may submit requests to multiple utilities.
Others may secure potential connections before they have customers.
Some projects may not have adequate financing.
Others may never receive permits.
And some may simply never be built.
That creates what regulators increasingly describe as “ghost demand.”
On paper, utilities see enormous future electricity requirements.
In reality, some portion may disappear.
That makes planning extremely difficult.
If utilities underestimate real demand, electricity shortages and grid reliability problems could follow.
If they overestimate it, billions of dollars could be spent constructing power plants, transmission lines and substations for data centres that never arrive.
And ordinary electricity customers could ultimately help pay for infrastructure that was unnecessary.
Texas has decided it wants to know which projects are real.
TEXAS HAS HIT THE BRAKES
Governor Abbott has ordered the Public Utility Commission of Texas and grid operator ERCOT to scrutinize proposed data-centre projects.
Developers are being asked to provide much more information.
Who actually owns the project?
How much electricity will it require?
Where will that electricity come from?
Does the company plan to build on-site generation?
How much water will it consume?
What taxpayer incentives is it receiving?
How will it affect nearby communities?
Texas officials argue that regulators cannot guarantee grid reliability using incomplete information.
And the consequences are significant.
Projects that fail to comply with the audit process can have their approvals denied.
The process has effectively frozen approvals for new data-centre grid connections while the state determines which projects are credible.
That is an extraordinary development for one of America’s largest data-centre markets.
DATA CENTRES MAY HAVE TO “PAY THEIR OWN WAY”
Abbott’s emerging position can be summarized in a simple phrase:
Data centres should pay their own way.
In June, he directed Texas regulators to ensure that data-centre operators fully fund the electrical infrastructure required to serve their facilities.
The goal is to prevent those costs from being transferred to residential customers.
He has also called for data centres to provide or contribute additional electricity capacity rather than simply adding enormous new demand to the grid.
Water is another concern.
Texas wants developers to use water-efficient cooling systems and increase water reuse.
Projects must also disclose expected water consumption and identify their water sources.
And Abbott wants Texas to phase out what he describes as outdated taxpayer-funded incentives for the industry.
Texas currently provides qualifying data centres with an exemption from the state’s 6.25% sales and use tax on certain equipment necessary to operate the facilities.
Those incentives were designed to attract investment.
Now political leaders are questioning whether some of them remain necessary when AI companies are already desperate to build infrastructure.
THE POLITICS HAVE FLIPPED
The speed of the political transformation is remarkable.
Texas spent years cultivating a reputation as one of America’s most business-friendly states.
Technology companies moved operations there.
Elon Musk relocated businesses.
Oracle moved its headquarters to Texas.
Huge data-centre projects followed.
OpenAI’s flagship Stargate campus in Abilene became one of the symbols of America’s AI infrastructure race.
For politicians, these projects once represented investment, technological leadership and economic growth.
But communities increasingly see another side.
Electricity demand.
Water consumption.
Noise.
Light pollution.
Industrial development.
Tax incentives.
And questions about how many permanent jobs communities actually receive in exchange.
An August Texas Politics Project poll cited by Reuters found that only 30% of respondents said they would support having a data centre in their own community.
That is the political danger facing the industry.
Americans may support artificial intelligence.
They may support competing with China.
They may support technology investment.
But that support can change when the infrastructure appears next door.
THE BACKLASH IS BECOMING BIPARTISAN
Data-centre opposition is becoming unusual in another way.
It is crossing America’s enormous political divide.
Democrats have increasingly criticized projects over environmental impact, utility costs and corporate subsidies.
Now Republicans are joining them.
Republican candidates in at least five states have recently toughened their positions toward data-centre development.
Texas Attorney General Ken Paxton has also adopted a harder stance, presenting proposals intended to maintain American AI competitiveness while limiting what his campaign describes as the negative effects of data centres.
That places some Republicans in tension with President Donald Trump’s aggressive push for American AI expansion.
Trump has argued that the United States must build enormous AI infrastructure to compete with China.
The strategic logic is straightforward.
The country with the greatest computing capacity may have a major advantage in developing the most powerful AI systems.
That means America needs chips.
Power.
Data centres.
Transmission infrastructure.
And enormous capital investment.
But politicians running for office are encountering another reality:
their voters live beside the infrastructure.
AI IS CREATING A NEW POLITICAL DIVIDE
The traditional political argument around technology was often relatively straightforward.
Republicans emphasized investment, deregulation and economic growth.
Democrats were generally more willing to regulate large technology companies.
Data centres are scrambling those lines.
A Republican politician can support American dominance in artificial intelligence while simultaneously opposing a giant data centre in a rural community.
A Democrat can support clean-energy investment while opposing an electricity-intensive AI facility.
A community can support technology companies while refusing to subsidize their infrastructure.
The political question is therefore shifting.
It is no longer simply:
Should America build AI?
Increasingly, it is:
Where should America build AI — and who should pay for it?
BIG TECH IS RESPONDING WITH POLITICAL MONEY
The technology industry understands the danger.
Big Tech companies and executives are pouring enormous sums into the 2026 midterm elections.
Reuters estimates that technology interests are spending more than $300 million across the election cycle as companies attempt to influence policy and defend the infrastructure needed for AI expansion.
In Texas alone, political groups connected to major AI interests have spent millions.
Super PACs with ties to Anthropic have spent more than $2.6 million in Texas.
Groups linked to OpenAI co-founder Greg Brockman have spent more than $4.8 million across the state.
Meta has contributed approximately $1.4 million to a super PAC focused on Texas state candidates.
The battle over AI infrastructure is therefore moving from engineering departments and utility commissions into elections.
That is a major change.
ELECTRICITY MAY BECOME AI’S BIGGEST BOTTLENECK
For the first several years of the generative-AI boom, one resource dominated discussion:
chips.
Nvidia GPUs were scarce.
Companies competed to secure them.
Cloud providers spent billions acquiring them.
But as semiconductor production expands, another constraint is becoming increasingly important.
Electricity.
A data centre can possess thousands of advanced GPUs.
Without enough power, those chips are useless.
That means AI companies increasingly need something technology companies historically didn’t worry much about:
power plants.
Transmission lines.
Transformers.
Substations.
Natural gas.
Nuclear energy.
Renewables.
Energy storage.
And regulatory approval.
The AI race is therefore becoming an energy race.
THERE STILL IS REAL DEMAND
The “ghost demand” problem should not create the impression that America’s data-centre boom is imaginary.
It isn’t.
Even after utilities introduce tougher financial requirements and remove speculative projects, enormous amounts of legitimate demand remain.
Exelon recently reduced its estimate of high-probability data-centre demand by about 40%, to approximately 11 gigawatts, after imposing stricter collateral requirements.
In Ohio, AEP Ohio’s data-centre power pipeline fell by more than half after rules introduced connection-study fees and other requirements.
But what remained was still substantial.
That is the paradox.
Some projected demand is exaggerated.
Yet real demand is still enormous enough to strain electricity systems.
The challenge for regulators is determining which is which.
CONSUMERS ARE WORRIED ABOUT THEIR BILLS
The politics become particularly dangerous when electricity prices enter the discussion.
Grid operators and utilities must invest enormous amounts of money to serve new demand.
Someone ultimately pays for those investments.
Consumer advocates fear households and small businesses could shoulder costs created by enormous technology companies.
The PJM Interconnection — America’s largest electricity grid, covering 13 states and Washington, D.C. — provides a warning.
Growth in existing and forecast data-centre electricity demand contributed to an estimated $29.4 billion increase in capacity costs across roughly the past four auctions used to determine those prices, according to the grid’s independent market monitor.
That does not mean data centres alone determine every household electricity bill.
Many factors influence consumer prices.
But when people begin associating higher electricity costs with AI infrastructure, the political equation changes quickly.
AI stops being an abstract technology story.
It becomes a household affordability story.
WATER COULD BECOME THE NEXT BATTLE
Electricity isn’t the only resource creating concern.
Data centres also require cooling.
Many cooling systems use substantial amounts of water.
That can create tension in areas already dealing with drought or limited water supplies.
Texas has explicitly incorporated water into its emerging data-centre standards.
Developers may be required to explain where their water will come from, how much they expect to consume and whether they can reuse it.
This matters particularly in rapidly growing communities.
Residents may accept industrial development when resources are abundant.
The political reaction can be very different when a community is being asked to conserve water while an enormous data centre is approved nearby.
THE JOBS QUESTION
Data-centre developers frequently emphasize economic development.
And construction can absolutely create significant employment.
Building enormous facilities requires engineers, electricians, construction workers, equipment suppliers and many other specialists.
Data centres also create permanent technical and operational jobs.
But once construction ends, the number of employees required to operate an enormous facility can be modest relative to the project’s size and electricity consumption.
That creates another political calculation.
Communities ask:
How many jobs are we receiving?
How much tax revenue?
How much electricity is being consumed?
How much water?
How much public infrastructure is required?
And how much subsidy did the company receive?
If voters conclude that the trade-off is poor, opposition becomes much easier to organize.
THIS COULD CHANGE WHERE AI GETS BUILT
The backlash doesn’t necessarily mean America will stop constructing data centres.
Far from it.
Hundreds of billions of dollars are already flowing into AI infrastructure.
Demand for computing continues rising.
The more likely outcome is that where and how data centres are built will change.
Developers may increasingly favour locations with abundant electricity.
They may build their own power generation.
They may sign long-term agreements with nuclear plants.
They may use more water-efficient cooling technology.
They may pay larger upfront deposits for grid connections.
They may lose some tax incentives.
And they may need to engage communities long before construction begins.
The era of announcing a giant data centre and assuming communities will automatically celebrate may be ending.
AI COMPANIES NOW HAVE A SOCIAL-LICENCE PROBLEM
Technology companies frequently talk about regulatory approval.
But there is another form of approval businesses need.
Community acceptance.
Sometimes called a social licence to operate, it describes whether people living around an industrial project believe its benefits justify its costs.
Oil pipelines understand this.
Mining companies understand this.
Power plants understand this.
Factories understand this.
Silicon Valley is beginning to understand it too.
AI has become physical infrastructure.
And physical infrastructure creates neighbours.
Those neighbours vote.
🔴 THE ABE NEWS TAKE
For years, the AI race looked like a competition between technology companies.
OpenAI versus Google.
Nvidia versus AMD.
Microsoft versus Amazon.
America versus China.
But another player has entered the competition:
the public.
And the public can slow things down.
The AI industry has spent extraordinary amounts of money solving technical problems.
How do we train larger models?
How do we manufacture more advanced chips?
How do we make inference cheaper?
How do we build smarter agents?
Now it faces a different category of problem.
How do you convince a community that an enormous industrial facility consuming vast quantities of electricity belongs next door?
That cannot be solved with a better algorithm.
Texas shows why this matters.
The state was practically designed for the AI infrastructure boom.
Huge.
Energy-rich.
Business-friendly.
Technology-friendly.
Growing rapidly.
Yet even there, political support is weakening.
That should concern every AI company planning hundreds of billions of dollars in infrastructure spending.
The industry has assumed that if demand for AI continues growing, infrastructure will follow.
But infrastructure requires land.
Power.
Water.
Transmission lines.
Permits.
Political support.
And increasingly, community approval.
Those resources are not infinite.
The 700-gigawatt queue of proposed large electricity users demonstrates how chaotic the boom has become.
Some projects are real.
Some may be speculative.
Some may be duplicates.
Some may disappear.
Texas’s 474-gigawatt queue does not represent 474 gigawatts of guaranteed future consumption.
But that’s precisely the point.
The AI boom has grown so quickly that governments don’t even know which pieces of the proposed infrastructure are real.
That is not sustainable.
The next phase will therefore be different.
Developers will have to prove they have financing.
Prove they have customers.
Pay for grid infrastructure.
Find electricity.
Secure water.
Reduce environmental impacts.
And convince communities that the economic benefits justify the costs.
That could make AI infrastructure more expensive.
It could delay projects.
It could eliminate weaker developers.
But it could also produce a healthier industry.
The strongest projects will survive.
Speculative projects may disappear.
Communities may receive better protections.
And technology companies may finally be forced to account for the true physical cost of artificial intelligence.
Because there is a reality underneath every AI model that Silicon Valley can no longer ignore:
The cloud isn’t actually in the sky.
It is sitting in enormous buildings.
Connected to power grids.
Drawing electricity.
Using water.
Occupying land.
And living beside real communities.
The AI revolution may have begun on computer screens.
Its next battle is happening on the ground.
ABE NEWS
Business. Money. Style. The News.
Understand More. Think Bigger.
