Politicians Are Fighting the Wrong War Over US Data Centers

Politicians Are Fighting the Wrong War Over US Data Centers

Every election cycle needs a convenient villain. Ahead of the US midterms, Washington has found its favorite punching bag—the sprawling, concrete-and-steel warehouses powering artificial intelligence.

Mainstream coverage frames the surge in data center construction as a binary collision between reckless Big Tech expansion and helpless local communities. Read any standard news brief on the subject and you get the exact same lazy script. On one side, tech giants demanding infinite megawatts and billions of gallons of water. On the other, angry voters watching their utility bills climb and politicians promising to slam the brakes on zoning permits to save the power grid.

That narrative is completely backwards.

Politicians aren't protecting ratepayers by blocking data centers. They are guaranteeing higher electricity prices, slower grid modernization, and the quiet offshoring of America’s most critical industrial infrastructure. If you actually want lower power bills and a cleaner electrical grid, you should be begging hyperscalers to build in your backyard—provided you know how to make them pay for the privilege.

The Grid Isn't Dying Because of AI

Let’s look at the fundamental lie driving this midterm panic: the idea that America’s electrical grid was a pristine, well-oiled machine until ChatGPT showed up and broke it.

Anyone who has actually worked in utility interconnection or high-voltage infrastructure knows the truth. The US power grid was already held together with duct tape, regulatory inertia, and fifty-year-old transformers long before large language models existed. For nearly two decades, US electricity demand stayed essentially flat. Utilities got comfortable. They deferred massive transmission upgrades because rate regulators wouldn't approve multi-billion-dollar capital expenditures without guaranteed load growth to pay for them.

Then came the twin shocks of domestic manufacturing reshoring and AI compute clusters. Suddenly, utilities face real load growth for the first time since the late twentieth century.

When local lawmakers blame data centers for grid strain, they are confusing the diagnostic test with the disease. A 300-megawatt data center campus doesn't break a healthy grid; it exposes decades of underinvestment by regional transmission organizations like PJM Interconnection and ERCOT. I have watched utility executives sit in closed-door planning meetings and privately admit that hyperscale demand is the only financial mechanism capable of funding the high-voltage transmission lines their territories desperately needed ten years ago.

Why Ratepayers Actually Win When Big Tech Moves In

The standard political talking point claims that residential customers subsidize the power lines built for massive server farms. When regulators fall asleep at the wheel, that happens. When regulators actually do their jobs, the exact opposite occurs.

Electricity generation and transmission are classic high-fixed-cost industries. Building a substation or a high-voltage transmission corridor costs roughly the same whether it serves ten thousand homes or ten thousand homes plus a massive data center. When a data center signs a long-term power purchase agreement (PPA) with strict minimum demand charges, it absorbs a massive share of those fixed infrastructure costs.

Per kilowatt-hour, spreading fixed grid maintenance across a massive, predictable, 24/7 industrial buyer lowers the unit cost for everyone else on the system. The problem isn't the data center. The problem is archaic utility tariff structures that fail to ring-fence capital upgrades. Banning the facility throws away the exact revenue stream needed to modernize local distribution networks without gouging households.

The Water Myth That Refuses to Die

If power grid panic doesn't work on voters, politicians pivot to water. You have seen the headlines claiming a single AI query pours a bottle of water onto the desert floor. It makes for great campaign ad copy and terrible engineering analysis.

Early-generation data centers relied heavily on evaporative cooling towers. They took potable municipal water, ran it through chillers, and evaporated it into the atmosphere. In drought-stricken regions like Phoenix or parts of Northern Virginia, that practice deserved every ounce of scrutiny it received.

What the political pundits deliberately ignore is that modern AI infrastructure has already abandoned that model.

High-density AI racks—packing 100 kilowatts or more per cabinet—cannot be cooled efficiently with traditional air handlers and evaporative towers anyway. The physics simply don't work. The industry has violently shifted toward closed-loop liquid cooling and direct-to-chip cold plates.

In a closed-loop system, the fluid circulates continuously in a sealed circuit, rejecting heat through dry coolers or chillers that consume virtually zero water after the initial fill. Think of it like the radiator in your car. You don't pour a gallon of fresh water into your engine every time you drive to the grocery store.

When a congressional candidate stands at a podium in 2026 railing against data centers draining local aquifers, they are fighting a war against 2015 hardware. Today’s hyperscale builds use less water per megawatt than a mid-sized golf course or a single alfalfa farm—yet nobody is running midterm attack ads against irrigated hay.

Stop Asking How to Stop Data Centers

Voters and journalists keep asking the wrong questions. Look at the standard public debate:

  • "How do we stop tech companies from raising our electric bills?"
  • "Should states pass moratoriums on new data center construction?"
  • "How much power will AI take away from regular homes?"

Every single one of those questions assumes a zero-sum scarcity mindset. They treat electrons like a fixed pie where a server rack eating a slice means a family freezes in the dark.

Power generation isn't a fixed pie. It is an elastic market governed by capital allocation and permitting speed. The real question voters should ask their representatives is brutal and simple: Why does it take seven years to permit a high-voltage transmission line in the United States, while a data center can be built in eighteen months?

That mismatch—not AI—is the entire crisis.

We have created a regulatory environment where private capital can build computing infrastructure at lightning speed, while public utility commissions and federal environmental reviews trap new power generation and transmission in endless bureaucratic purgatory. FERC (the Federal Energy Regulatory Commission) has spent years trying to unclog interconnection queues where hundreds of gigawatts of wind, solar, battery storage, and natural gas projects sit waiting for approval.

Politicians love passing local moratoriums on data centers because it lets them look tough on Big Tech without having to do the hard, unglamorous work of reforming federal permitting laws or overhauling regional utility monopolies. It is pure political theater.

The Uncomfortable Truth About Nuclear and Natural Gas

If you want to watch a politician sweat, ask them how they plan to power domestic AI without natural gas or nuclear energy.

For the past five years, corporate sustainability teams at Microsoft, Google, and Amazon promised voters and shareholders they would run their operations on 100 percent renewable energy. On paper, they achieved this by buying unbundled Renewable Energy Certificates (RECs). They would build a data center in Virginia running on coal and gas power, then buy wind credits from a farm in Texas to claim carbon neutrality.

Physics finally caught up with accounting.

An AI training cluster running tens of thousands of GPUs does not care if the wind is blowing in West Texas. It requires baseload power—uninterrupted, 24/7, ultra-reliable electricity with zero voltage fluctuations. Solar panels stop producing at dusk. Wind turbines stall during high-pressure weather systems. Battery storage works brilliantly for smoothing out four-hour evening peaks, not for powering a 500-megawatt compute campus through a three-day winter freeze.

This physical reality forced the tech giants to drop the greenwashing and start buying real baseload generation. That is why we are seeing unprecedented deals to restart retired nuclear reactors like Three Mile Island Unit 1, co-locate data centers directly behind the meter at existing nuclear plants, and sign long-term contracts for combined-cycle natural gas plants equipped with carbon capture.

Yet during election season, both political parties weaponize this shift. One side attacks tech companies for touching natural gas; the other side attacks them for consuming power that supposedly belongs to traditional manufacturing. Both sides ignore the obvious: behind-the-meter power agreements actually remove massive loads from the public grid while injecting billions of private dollars into keeping aging nuclear fleets financially viable.

What Actually Happens When You Ban Data Centers

Imagine a scenario where populist lawmakers get their wish. Swing states pass sweeping moratoriums on new data center zoning, slap punitive energy taxes on compute facilities, and starve hyperscalers of grid interconnections.

Does AI development stop? Does energy demand vanish?

Not for a second. Capital simply flows to the path of least resistance.

First, domestic data center construction retreats entirely into states with massive, deregulated energy surpluses and minimal permitting friction—places like Texas, North Dakota, and rural parts of the Midwest with abundant natural gas and wind resources. States that pass moratoriums don't save the planet; they simply forfeit hundreds of millions of dollars in local property tax revenue that usually funds public schools, roads, and emergency services without requiring a single new school bus or sewer hookup.

Second, and far more dangerously, compute capacity moves offshore. Sovereign wealth funds in the Middle East and state-backed energy monopolies in Asia are actively courting American tech giants with cheap, state-subsidized power, streamlined permitting, and guaranteed water access.

Once the physical infrastructure—and the massive datasets trained on it—migrates outside US borders, American regulators lose jurisdiction over privacy, national security, and AI safety standards. Local politicians bragging about blocking a server farm in their county are inadvertently accelerating the hollowing out of US technological sovereignty.

The Playbook That Actually Works

I have sat across the table from municipal planners, utility regulators, and hyperscale site-selection teams. The communities that get fleeced by Big Tech are the ones that negotiate from a place of either blind desperation or ideological hostility. The winners treat hyperscalers like heavy industrial counterparties and extract ironclad concessions.

If lawmakers actually want to protect voters instead of farming outrage for campaign commercials, they should throw out the moratoriums and enforce four non-negotiable rules:

  1. Mandatory Ratepayer Ring-Fencing: Require large-load customers (anything over 25 megawatts) to sign dedicated large-load tariffs. If a substation or transmission line must be built to serve a data center, the developer pays 100 percent of the upfront capital cost and guarantees minimum monthly payments for 15 to 20 years—even if they cancel the project or leave early. Zero stranded asset risk for residential customers.
  2. Zero-Evaporation Water Mandates: Outlaw open-loop evaporative cooling for new commercial compute facilities in any watershed experiencing moderate-to-severe drought stress. Force closed-loop liquid cooling or air-cooled chillers as a baseline zoning requirement.
  3. Grid-Interactive Demand Response: Stop letting data centers sit on the grid as passive, inflexible sponges. Require facilities to integrate their massive backup battery banks and on-site generation with the local utility. During peak summer heatwaves or winter polar vortexes, the utility gets contractual authority to curtail non-critical AI training workloads or draw power back from the data center’s batteries to stabilize residential neighborhoods.
  4. Co-Located Generation Requirements: Fast-track zoning approvals only for projects that bring their own net-new power generation to the table—whether that means funding new utility-scale solar-plus-storage, uprating existing nuclear reactors, or building clean thermal generation.

Admitting the Trade-Offs

Let’s be brutally honest about the downsides of this approach, because pretending there are none is just another form of political dishonesty.

Even with strict ring-fencing and closed-loop cooling, building massive industrial infrastructure next to residential suburbs creates real friction. Data centers are loud when diesel backup generators undergo monthly testing. They are visually ugly—massive, windowless concrete boxes surrounded by chain-link fences and high-voltage switchyards. During the two-year construction phase, local roads take a beating from heavy cranes and concrete trucks, while permanent operational jobs remain relatively low compared to a traditional auto plant or hospital.

Pretending those nuisances don't exist insults the intelligence of local residents. Communities deserve generous community benefit agreements, strict decibel limits on generator testing, architectural screening, and massive property tax allocations to compensate for living next to industrial infrastructure.

What they don't deserve is being lied to by politicians who claim that killing data center projects will magically fix a broken utility model.

The US power grid needs trillions of dollars in upgrades over the next two decades to handle electric vehicles, domestic manufacturing, extreme weather, and digital infrastructure. You can either force everyday families to pay for those upgrades through higher monthly utility bills, or you can force the wealthiest tech corporations in human history to foot the bill in exchange for grid access.

Washington is currently choosing to chase the cash away and stick you with the tab.

LC

Lin Cole

With a passion for uncovering the truth, Lin Cole has spent years reporting on complex issues across business, technology, and global affairs.