The whole country ostensibly wants America to win the artificial intelligence (AI) race. A striking number, however, would prefer someone else’s town to host the data centers, power plants, transmission lines, and cooling systems required to run it.
Adam Smith knew the type. In “The Theory of Moral Sentiments,” he warned against the “man of system,” who imagines society as a chessboard and believes he can move human beings as easily as pieces.
Today’s man of system has a data-center plan. Governors, legislators, regulators, and activists increasingly speak as though they can determine where enormous new electricity loads will locate, which power sources will serve them, how their owners will bargain with utilities, what labor terms they will accept, and how much support they will provide their communities—all while preserving low rates, grid reliability, environmental goals, and America’s lead in AI.
The chess pieces, it turns out, have a motion of their own.
In Pennsylvania, Gov. Josh Shapiro unveiled the Governor’s Responsible Infrastructure Development Standards, or GRID Standards, as the terms developers must meet to receive faster permitting, tax incentives, and coordinated state support. The Pennsylvania House voted 134-68 in June to codify them. When the Senate did not act, Shapiro told the Pittsburgh Business Times that he would consider executive action.
In Texas, Gov. Greg Abbott ordered regulators earlier this month to freeze approvals for data centers seeking grid connections until the state completes an audit. New York has imposed a one-year statewide moratorium on permits for the largest facilities.
The political pressure is clear. A July Quinnipiac poll found that 74% of Pennsylvania voters opposed an AI data center in their community. Republicans, Democrats, and independents agreed. The New York Times called the opposition perhaps “the most bipartisan issue since beer.”
Some of that resistance is exaggerated, emotional, and plainly hostile to growth. Other objections deserve a serious answer. Data centers are large industrial facilities, and residents are entitled to ask about noise, land use, water, air emissions, electricity bills, tax abatements, and the integrity of local decision-making. The secrecy surrounding some projects has deepened public distrust. A credible free-market case for data centers must acknowledge legitimate costs and concede that hyperscalers—companies that operate enormous networks of data centers—do not always strike defensible bargains.
Yet policymakers commit a grave error when they treat a project’s costs as grounds to stop an industry. That response treats scarcity as evidence of market failure, assumes public officials can identify the correct technical response in advance, and interrupts the decentralized adjustments already underway.
It also mistakes the visible building for the demand it serves. Demand for cloud computing, cybersecurity, medical research, financial services, logistics, streaming, and AI persists after a government prohibits a data center. The facility simply goes elsewhere and takes its investment, infrastructure, tax base, and accumulated knowledge with it.
Austrian economics focuses on how dispersed knowledge, prices, and entrepreneurial experimentation help people adapt to scarcity. Through that lens, the relevant question is whether moratoria and prescriptive mandates improve the process by which firms, utilities, communities, and consumers reconcile rapidly growing demand with limited supplies of electricity, water, land, labor, and capital. In reality, such policies obstruct that process.
The Grid’s Preexisting Condition
The data-center debate emerged amid longstanding grid problems. U.S. electricity demand remained comparatively flat for roughly two decades, and utility forecasts, investment patterns, and regulatory institutions adapted accordingly. Meanwhile, policy-driven plant retirements subordinated reliability to an environmental ideology. Slow permitting, transmission constraints, equipment shortages, inflation, flawed market rules, and interconnection backlogs further limited how quickly suppliers could add capacity. AI and cloud computing then produced a large, geographically concentrated surge in demand.
The popular account starts with the latest events. Data centers arrive, electricity prices rise, and officials decide to restrain the new demand. A Misesian account, named for Austrian economist Ludwig von Mises, starts earlier. One government intervention disrupts the market process and produces consequences that prompt another intervention. Government action reduces the electricity supply or raises production costs. Increased demand exposes those constraints. Officials blame data centers, the most visible new users, for all resulting price increases and then propose freezes or quotas that make new supply even harder to finance.
Regulation did not cause every grid problem. Still, the newest electricity user cannot explain the current scarcity by itself. PJM Interconnection (PJM), the country’s largest regional grid operator, recently held a capacity auction, in which electricity suppliers receive commitments to ensure that enough power will be available during periods of peak demand. The auction produced billions of dollars in additional charges. The New York Times reported a $6.3 billion increase affecting 13 states and the District of Columbia, with much of its coverage focused on data-center demand.
The evidence is serious, though incomplete. A detailed 2026 analysis by Energy and Environmental Economics (E3) estimated that peak-load growth, driven primarily by data centers, accounted for roughly half of the increase in PJM’s capacity price between 2024-25 and 2025-26. Retirements, program changes, supply constraints, and market-design problems accounted for the other half. PJM lost more than 9 gigawatts of generating capacity between 2023 and 2025, and utilities have announced further retirements. Its interconnection process, which connects new generators to the grid, has struggled to bring replacement capacity online.
Blaming data centers for the entire increase is like blaming a heavy truck for exposing the weakness of a neglected bridge. Years of neglect weakened the bridge. The truck revealed the cost.
A moratorium on data centers therefore amounts to political rationing masquerading as prudence. Moratoria leave the grid’s problems untouched. They build no generators, shorten no transmission-permitting timelines, manufacture no transformers, and make no improvements to PJM’s interconnection process. At most, they suppress one category of demand while leaving the supply shortage intact. If that demand relocates, the jurisdiction loses the associated investment while the national infrastructure problem remains.
Hayek Walks Into a Data Center
Friedrich A. Hayek’s “The Use of Knowledge in Society” begins with a simple claim. No single mind possesses the knowledge needed to coordinate an economy. That knowledge is dispersed among individuals and embedded in the “particular circumstances of time and place.” Prices communicate it without requiring anyone to understand every distant cause.
Electricity markets illustrate Hayek’s point unusually well. A decision to locate a data center in Luzerne County in northeastern Pennsylvania, rural Texas, northern Virginia, or a site no one has considered depends on thousands of facts. They include available generation, substation capacity, transmission congestion, fiber routes, cooling methods, water conditions, land prices, construction labor, regulatory timelines, requirements for uninterrupted service, future computing power per rack, and customers’ tolerance for delay. Officials cannot know many of these facts. Others have yet to emerge because investment and experimentation will create them.
A higher electricity price, longer interconnection wait, or more expensive water right conveys useful information. It may prompt a developer to reduce consumption, finance new supply, shift computing to off-peak hours, change its technology or location, or abandon a project that cannot justify its resource use. Developers choose among these responses, and their choices generate new information. Scarcity calls for economizing and building. Bans silence the signal.
A moratorium suppresses that discovery. “Wait” reveals nothing about which projects are genuine, which can cover their costs, which can shift their electricity use, or which sites have adequate capacity. It replaces countless project-by-project decisions with a single political decree.
Texas demonstrates the problem. Regulators reported a queue of roughly 1,800 proposed projects representing 474 gigawatts—more than five times the state’s record peak demand. Data centers accounted for about 90%. Many applications are plainly speculative. Only 28 of 377 companies responded to a state survey requesting details about their plans. Regulators cannot protect reliability using fictional projects.
The queue presents a screening and information problem. Treating every application alike fails to solve it. Regulators can require binding demand forecasts and deposits that developers forfeit when they reserve scarce queue capacity without making progress. They can demand collateral, minimum payments, construction milestones, and evidence that developers control their proposed sites. They can assign the costs of dedicated substations and transmission upgrades to the customers that cause them. They can also remove applications that miss their deadlines. These mechanisms separate serious projects from placeholders. A categorical freeze obscures that distinction.
Israel Kirzner extends Hayek’s argument by describing competition as a process of entrepreneurial discovery. Markets allocate known resources among known uses. They also allow entrepreneurs to notice overlooked opportunities, test judgments under uncertainty, and reveal information through profit and loss. That process becomes especially valuable when inputs, production methods, and products are changing simultaneously, as they are in AI infrastructure.
No governor knows the optimal power mix for data centers in 2035. Neither does any hyperscaler. Institutions determine who can test forecasts and who bears the cost of error. Firms can try different combinations, copy successful approaches, and pay for their mistakes. A mandate imposes one political forecast and distributes the cost of its errors among everyone subject to it.
Hayek’s fatal conceit appears here in industrial form. Intelligence and good intentions cannot replace a process that generates knowledge no one possesses at the outset.
Make the Hyperscalers Pay the Electric Bill
The strongest objection to data centers concerns electricity. A large, steady load may require new generation, substations, and transmission. If a project shrinks, stalls, or disappears, other customers could be stuck paying for facilities built to serve it. Poor cost allocation could leave residential ratepayers subsidizing some of the world’s largest companies.
Yet the evidence does not show that data centers inevitably raise residential rates. E3 found no quantitative evidence of systematic historical subsidies. The Data Center Coalition funded the report, and E3 cautioned that the limited research cannot predict future outcomes. Its analysis of four Amazon facilities found that they produced an average annual utility surplus of about $3.4 million. An Electric Power Research Institute analysis also found that greater electricity sales correlated with modestly lower retail rates between 2015 and 2024 because large customers helped spread fixed costs across more kilowatt-hours.
Recent examples support that finding. Georgia regulators approved a plan expected to reduce a typical residential bill by about $50 a year, with much of the reduction attributed to revenue from new large customers under a tariff designed to protect other ratepayers. Indiana Michigan Power has cited data-center growth in support of a base-rate reduction. In Louisiana, Entergy projects that its agreements to serve Meta’s Hyperion campus will provide customers with approximately $2.65 billion in benefits over 20 years.
These figures remain projections from utilities and regulators. They establish no universal rule. PJM shows that rapid demand growth can raise capacity prices when suppliers cannot add power quickly enough. Data centers can lower average costs when their payments cover the added expenses and their revenue grows faster than total system costs. They can raise costs when supply remains constrained or utilities shift expenses to other customers. Policy should create terms that favor the former result.
Utilities are already developing those terms. According to E3, utilities established at least 38 specialized large-load tariffs between 2018 and early 2026, including 30 in 2025 and 2026. These rate schedules use longer contracts, minimum bills, take-or-pay provisions that require payment for reserved power even if the customer does not use it, upfront construction contributions, credit support, exit fees, phased increases in electricity use, and cost-allocation reviews. American Electric Power Ohio, for example, developed a tariff intended to protect other customers from speculative projects and abandoned assets.
These mechanisms apply a straightforward principle known as cost causation. A data center should pay the additional costs it creates. If it receives continuous reliability, grid balancing, transmission, or backup service, it should pay for those services even when it generates some electricity on-site. If the project finances assets that benefit the broader grid, its contract should also assign those benefits accurately.
Cost causation cannot settle every dispute. Regulators must still determine how to measure shared expenses, and utilities retain incentives to overbuild or shift risk to captive customers. Transparent tariffs, auditable assumptions, and enforceable contracts allow regulators and utilities to revise their approach as evidence accumulates. A moratorium offers no similar means of correction. It turns a difficult pricing problem into a political veto.
Every Gallon Has an Address
Water gives the data-center backlash its most vivid imagery. Computers generate enormous amounts of heat, and many facilities use water to keep their servers cool. A server farm drawing millions of gallons near irrigated fields or residential wells naturally alarms neighbors.
Assessing the strain requires more than a single annual figure. A facility may withdraw water from a river or aquifer and return much of it, while consumption measures the portion lost to evaporation or otherwise removed from the local supply. Data centers also have an indirect water footprint because many power plants use water to generate the electricity they consume.
Timing and location matter just as much. Heavy use on the hottest day of the year can strain a local water system even when annual use appears modest, and ample water nationwide cannot replenish an overdrawn aquifer in a particular community.
U.S. data centers directly used roughly 17 billion gallons of water in 2023. That figure sounds enormous in isolation, though it represents a small share of national use and reveals little about any particular project. Evaporative cooling in an arid watershed may create a serious local problem. Air cooling, reclaimed wastewater, and recirculating systems in a water-rich region create different costs and risks. Efforts to reduce direct water use may also increase electricity demand or construction costs.
Officials should assess water use by watershed and during periods of peak demand. Comparable industrial users should report how much water they withdraw and consume. Prices should reflect local scarcity, authorities should enforce water rights, and developers should pay for dedicated infrastructure. Technology-neutral standards can identify a measurable local harm while allowing developers to choose among reclaimed water, dry cooling, recirculation, and other methods.
The American Enterprise Institute’s (AEI) review of the backlash makes a concession that strengthens the free-market case. National totals can appear reassuring even when local peak demand strains water systems. Research cited by AEI projects substantial additional water-capacity needs through 2030 if cooling efficiency remains unchanged. The same research finds that efficiency gains could sharply reduce those needs. Governments may reasonably require reporting and coordination, as well as “pipe-neutral” development that adds enough capacity or conservation to offset a project’s demands. Blanket bans disregard these differences.
Critics often overstate the water objection, but water remains scarce and costly in many communities. Authorities should measure, price, and address that scarcity where it occurs. A statewide moratorium based on a national statistic treats unlike communities identically and blocks technologies that could render yesterday’s estimates obsolete.
Spontaneous Order, Now With Turbines
Markets are already answering the claim that they cannot supply enough power for AI.
Two years ago, economist Lynne Kiesling framed the choice as “make or buy.” Would data-center operators continue purchasing electricity from utilities, or would high prices, delays, and reliability concerns lead them to generate their own? Her recent follow-up describes a third option. Firms can make, buy, or ally with utilities, power producers, equipment manufacturers, and investors. These arrangements are changing the commercial relationship between computing and electricity.
Some operators continue to buy power under ordinary or specialized utility tariffs. Others sign long-term power-purchase agreements, which commit them to buy electricity from a generator for a set period. Microsoft entered a 20-year agreement supporting the restart of Three Mile Island Unit 1 in Pennsylvania, while Meta signed a long-term agreement associated with the Clinton nuclear plant in Illinois. Google and Amazon are backing advanced nuclear-reactor projects. Other firms are developing power projects with gas-turbine manufacturers, utilities, and infrastructure investors. Some place generation “behind the meter,” meaning on the customer’s side of the utility connection, or next to an existing power plant.
Each choice assigns risk, control, financing, permitting duties, and technical responsibility differently. Delays also impose a price. A firm racing to deploy scarce computing capacity may lose more during a multiyear wait for a grid connection than it would spend on costlier on-site generation. One company may accept the complexity of a nuclear agreement to secure long-term reliability. Another may choose natural gas because turbines can begin operating sooner. A third may stay with a utility because producing power would distract from its main business.
Pennsylvania offers several examples. Prime Data Centers has proposed a 450-megawatt facility beside a gas plant in Hanover Township and says it can operate behind the meter. Amazon has discussed a possible data center at the former Homer City coal site in Indiana County, where developers envision a combined natural gas-generation and computing complex. GE Vernova is investing nearly $166 million in western Pennsylvania facilities that manufacture high-voltage and gas-turbine equipment. Data-center demand is prompting investment throughout the electricity supply chain.
These projects still warrant scrutiny. A behind-the-meter facility may remain dependent on the grid for synchronization, balancing, backup power, or other reliability services. Only a facility capable of operating independently during a grid outage avoids that dependence. The Hanover Township project has yet to resolve every water and permitting issue. Homer City still lacks a final customer. GE Vernova’s investment also receives state support. Each project carries costs, uncertainties, and unresolved questions.
They also provide evidence of discovery. Firms are testing different combinations because no one knows which model will prove most efficient, and the answer will likely vary by project. Experience will show which factors favor utility service, long-term contracts, partnerships, colocation, or on-site generation.
The Amazon-Talen colocation agreement at Pennsylvania’s Susquehanna nuclear plant shows how complicated that experimentation can become. Colocation places a large electricity user beside a generator, potentially allowing the facility to obtain power without moving all of it across the wider grid. Federal regulators rejected an amended interconnection agreement in 2024. They later directed PJM to develop clearer rules for arrangements that pair large data-center loads with power plants. Physical proximity does not erase the costs of shared grid services. The rules must assign those costs to the customers that incur them.
Law and contracts establish the terms for this experimentation, though no regulator, utility, or firm planned the full pattern. Many parties responding to prices and local constraints have developed specialized tariffs, nuclear restarts, joint ventures, flexible electricity use, colocation agreements, and on-site generation. Austrian economists call this spontaneous order, a system of coordination that emerges through many decentralized decisions rather than a central plan. A moratorium halts that experimentation when firms and regulators have the most to learn.
GRID Comes With Strings Attached
In a recent piece for National Law Review, I argued that Pennsylvania correctly recognized what AI data centers could bring. The projects could attract private investment, create demand for new power plants and electrical equipment, put former industrial sites back to productive use, and generate jobs and tax revenue. Pennsylvania’s energy resources, manufacturing base, and available industrial sites make it a plausible beneficiary. GRID’s error lies in assuming the state can prescribe the energy sources, labor arrangements, and community commitments that developers should use to produce those gains. As the backlash grows, the distinction between enforcing neutral rules and dictating production choices becomes more urgent.
GRID combines sound principles with central planning. Data centers should pay the costs attributable to their electricity use. Disclosure, early community engagement, water planning, and enforcement of generally applicable environmental laws are also defensible. GRID adds a prescribed “build, bring, or buy” framework, an increasing share of “clean firm” power that can operate regardless of weather, solar-ready construction, specified labor and compensation rules, community-benefit commitments, and continuing administrative oversight.
For now, GRID remains voluntary. A developer may reject its standards and proceed under ordinary permitting requirements and generally applicable laws, though it will lose access to tax incentives, expedited permitting, and coordinated state support. Pennsylvania can use those government-controlled benefits to steer investment toward its preferred production model. The House vote to codify the standards and Shapiro’s contemplated executive action show how readily an optional program can supply the blueprint for regulation.
Economist Murray Rothbard called this arrangement “triangular intervention.” The term describes government requiring private parties to conduct an exchange on terms they would not otherwise choose. Under GRID, a developer gains access to state support only after accepting politically selected obligations governing energy, labor, reporting, and community benefits.
Compliance costs are only part of the problem. The deeper difficulty concerns knowledge. Officials cannot know in advance which technologies, contracts, or operating practices will best address a project’s demands. Every prescribed input or quota prevents firms from testing at least one possible approach.
A data center might discover that shifting computing tasks that can tolerate delay to off-peak hours reduces grid costs more than dedicated generation would. A technology-neutral tariff can reward that choice by charging the customer for the costs it causes. A mandate tied to a specified technology or quota cannot accommodate the innovation until officials amend or waive it. Incumbent firms will usually handle that process more easily than new entrants. A policy advertised as discipline for Big Tech may end up protecting it from competition.
Economist James Buchanan’s principle of generality offers a better standard. Rules should group regulated parties according to relevant characteristics. A 450-megawatt chemical plant and a 450-megawatt data center may create similar risks for the grid. A large beverage plant may pose a greater threat to local water supplies than a server facility that uses dry cooling. Regulation should reflect electricity use, emissions, water withdrawals, credit risk, noise, and land-use effects. Officials’ views of the customer’s business model have no bearing on those harms.
Several reforms meet that standard. Texas can require every large electricity user to substantiate its place in the interconnection queue and demonstrate its financial capacity. Pennsylvania can require any customer responsible for a network upgrade to pay the added cost. Local governments can apply neutral rules governing noise, setbacks, traffic, light, wastewater, and emergency services. A special permission system aimed at one politically unpopular industry, by contrast, would replace general rules with administrative favoritism.
The Sweetheart Deal Comes Due
Louisiana’s agreement with Meta is often presented as proof that data-center growth can benefit ratepayers. Entergy says the project will support seven new combined-cycle gas plants, more than 5,200 megawatts of generation, major transmission upgrades, batteries, nuclear uprates, and solar development. The utility projects $2.65 billion in customer savings over 20 years, including $120 million for vulnerable customers and $140 million for energy efficiency. The project has already produced construction activity, local sales-tax revenue, training commitments, and infrastructure spending.
Those claims deserve attention, along with attribution. Parties to the deal made the projections, and most of the promised benefits have yet to materialize. A New York Times investigation also documented secret negotiations, nondisclosure agreements, potentially enormous tax breaks, the use of public land, political conflicts, and contractual risks that could fall on other customers if the assumptions prove wrong. The investigation found genuine local gains as well as rising rents and community disruption.
Austrian economics and public choice theory can account for both. A project this large can create substantial value, finance new electricity supply, and spread the grid’s fixed costs across more customers. Yet government control over tax privileges, public land, and regulatory approvals creates opportunities for rent seeking, in which firms pursue political favors instead of competing on equal legal terms. It can also foster favoritism, secrecy, and the transfer of private risks to the public. Corporate welfare deserves no free-market defense, and opposition to moratoria provides none.
Louisiana also shows how one intervention can prompt another. Officials grant a favored project customized tax and regulatory treatment, while some risks may fall on ratepayers or taxpayers. Secrecy and favoritism then provoke public anger. Rather than withdrawing the privileges and applying general rules, officials impose new restrictions on later projects. Privilege breeds backlash, and backlash breeds control. Alternating between political favors for selected firms and restrictions on an entire industry produces an interventionist spiral in miniature.
What the Moratorium Doesn’t Show
Frédéric Bastiat’s distinction between the seen and the unseen helps explain why data-center politics favors prohibition. Residents can see a proposed building, transmission line, cooling plume, or higher utility charge. The costs of rejection are harder to observe. They include investments canceled, suppliers that never expand, workers who never receive training, power plants that remain uneconomic, and innovations that scarce computing capacity makes more expensive or delays.
Critics often point out that a completed data center employs relatively few permanent workers compared with a manufacturing plant occupying similar acreage. That comparison is fair, though server-room employment captures only part of the economic activity. Data centers also require construction trades, electrical equipment, fiber connections, security, maintenance, engineering, and new power generation. They add tax revenue and demand for suppliers. PricewaterhouseCoopers’ (PwC) 2026 report estimates that the broader U.S. data-center industry supported about 5.5 million direct, indirect, and induced jobs and contributed roughly $927 billion to gross domestic product (GDP) in 2024.
Those figures require context. They do not represent 5.5 million new jobs or $927 billion in new output caused solely by data centers. PwC used a broad definition that included workers who support data-center functions and counted indirect jobs at supplier firms and induced jobs supported by workers’ spending. The report estimated roughly 1 million direct jobs and $302 billion in direct GDP. The Data Center Coalition commissioned the study, which measures activity associated with the industry rather than the benefits guaranteed by any proposed facility.
Even with those qualifications, the political imbalance remains. The benefits are dispersed among future workers, suppliers, taxpayers, and consumers, while nearby residents bear concentrated and immediate costs. Public choice theory predicts that organized residents who attend a hearing will command more attention than people who might benefit years later and may not know that the project would affect them. The residents’ concerns are real. So are the forgone benefits.
Moratoria can also reduce competition. They protect existing data centers against new entrants, increase the value of scarce computing capacity, and favor companies large enough to wait, lobby, litigate, or move elsewhere. Smaller developers and innovative energy suppliers have fewer resources to endure indefinite pauses and negotiate customized political deals. Policymakers worried that AI will concentrate power among a few technology companies should consider who benefits when the government restricts new computing and electricity supply. A policy advertised as a restraint on Big Tech may become Big Tech’s best moat.
A moratorium also imposes costs during the pause. A data-center project requires developers to coordinate land, fiber connections, interconnection rights, equipment orders, construction schedules, and customer commitments. Delay one part, and the entire arrangement may collapse. Investors compare states and communities, then move when another location offers greater certainty. A community may reject the facility while continuing to consume services produced elsewhere and paying a share of regional grid costs. It will forgo the investment and tax revenue, along with its ability to influence the project’s design.
Meter Readers, Not Master Planners
Government has a legitimate role in data-center development, one compatible with dispersed knowledge, property rights, and the rule of law.
Utilities should adopt transparent large-load tariffs based on cost causation. Developers should fund dedicated upgrades, post collateral for speculative capacity reservations, make minimum payments for infrastructure built to serve them, and pay exit charges if they abandon a project. Behind-the-meter facilities should also pay for any backup power, real-time balancing, and other grid-stabilization services they use.
State and local governments should enforce neutral, measurable rules governing noise, setbacks, traffic, emissions, wastewater, lighting, and emergency services. Water standards should reflect conditions within the affected watershed and apply equally to comparable industrial users. Community review should begin early, disclose material commitments, and produce decisions within predictable timelines. A hearing should establish site-specific facts rather than become a referendum on AI itself.
Government has a role to play in policing cost shifting. Developers should decide which power sources best meet their needs. Policymakers can help by removing barriers to new generation and transmission, improving interconnection queues, and permitting on-site and colocated generation subject to safety, environmental, and cost-allocation rules.
Performance standards can set measurable requirements for reliability, emissions, water use, and incremental costs while allowing firms to test different ways of meeting them. Technology mandates and politically chosen quotas foreclose those choices. Pennsylvania’s clean-firm percentages illustrate the problem. They prescribe a particular input instead of charging for a measurable effect. A better rule would require each project to cover its added grid costs and meet defined reliability and environmental standards while allowing the developer to choose its technology.
Governments should also abandon selective tax favors as a tool of social planning. Hayek’s distinction between general rules and discretionary commands in “The Road to Serfdom” offers a better framework. Government should announce general, prospective rules that allow firms to make their own plans. Bargaining over individualized privileges and obligations after a favored applicant arrives produces uncertainty and favoritism.
If a data-center tax exemption cannot be justified under a neutral tax system, lawmakers should repeal it. Retaining the preference and conditioning it on an expanding list of energy, labor, and community-benefit mandates compounds the original distortion. Tax neutrality provides a cleaner remedy for corporate welfare than industrial planning does.
Officials also need honest baselines. Electricity rates reflect fuel costs, inflation, plant retirements, reliability spending, transmission constraints, grid upgrades, market rules, and demand from many sources. Data centers may raise costs in one location and spread fixed costs among more customers in another. Water may be abundant over an entire year yet scarce during the hottest week.
Regulators who attribute every increase to the most visible new customer conceal other causes. Industry advocates who deny genuine local burdens sacrifice their credibility. Sound rules begin with candor from both.
The Chess Pieces Keep Moving
The data-center backlash is understandable. AI is advancing quickly, its infrastructure is physically immense, and secret negotiations and rosy projections have given communities ample reason for skepticism. Political unease, though, supplies none of the technical knowledge needed to design the next generation of computing and electricity.
Firms are already financing new generation, negotiating long-term power agreements, developing specialized tariffs, restarting nuclear plants, testing advanced reactors, building gas-fired generation, shifting computing workloads, improving cooling systems, and inventing new contracts. No state plan anticipated the full range of responses. Some projects will fail, and some deals will deserve rejection. Profit, loss, and experience reveal which arrangements work and who should bear their costs.
Government should protect property rights, enforce contracts, require truthful disclosure, charge accurately for public services, and make developers pay the costs they create. Policy should reflect the limits of official knowledge through general rules, technological neutrality, open entry, transparent agreements, and predictable decisions.
Infrastructure anxiety calls for more investment guided by prices, property rights, contracts, and competition. Those institutions allow firms, utilities, and communities to discover solutions that none can fully anticipate alone.
A community may reject a data center, but demand for computing will persist and attract investment elsewhere. The chess pieces will keep moving.
