Introduction: The City Ballot That Became a National Referendum
On the evening of June 2, 2026, in a modest mid-century city hall ten miles east of downtown Los Angeles, along the eastbound corridor of the I-10 freeway, the residents of Monterey Park did something that no American municipality had ever done before. By a margin of 10,321 votes to 1,362 — a landslide of 88.34 percent — they approved Measure NDC, a ballot initiative that permanently prohibits data centers within the city limits, amending the city’s General Plan so that the ban can only be reversed by a future vote of the people themselves.[1][2] It was the first time in American history that a municipality banned data centers through a direct ballot initiative, and it happened not in a rural county fighting a distant corporation, but in a dense, diverse, working-class and middle-class city in the heart of the nation’s second-largest metropolitan area.[3]
The proximate cause was a single project: a proposed conversion of an aging office property at 1977 Saturn Avenue into a quarter-million-square-foot artificial intelligence data center, advanced by the Australian investment firm HMC StratCap through its DigiCo platform.[2] Residents learned that the facility would have consumed roughly three times as much electricity as the entire city of Monterey Park currently uses. They organized under the banner “No Data Center Monterey Park,” filled council chambers, planted yard signs, and pressed their elected officials until the City Council imposed a development moratorium extending through January 2027 and placed the permanent prohibition on the June ballot.[1][4] When the votes were counted, the co-founder of the resident group, Steven Kung, summarized the outcome in five words:
“We had a victory landslide.”
— Steven Kung, co-founder, No Data Center Monterey Park [4]
Taken alone, Monterey Park might read as a local land-use curiosity — one small city, one withdrawn project, one emphatic election. But it did not happen alone, and that is the argument of this paper. Six weeks later, on Saturday, July 18, 2026, coordinated demonstrations against data center development took place in at least 125 locations — by the organizers’ own count, 142 communities across 42 states — in the first nationally synchronized day of protest against the physical infrastructure of artificial intelligence.[8][9][10] The demonstrations were organized not by a progressive environmental coalition but by HumansFirst, an advocacy group chaired by Amy Kremer, a co-founder of the Tea Party movement, and they drew participants across the partisan spectrum who share a common set of grievances: rising electricity bills, water demand, noise, land conversion, and the sense that enormous decisions are being made about their communities without their consent.[8][10] The organizers’ own framing was explicitly electoral:
“Americans deserve a voice before Big Tech reshapes their communities.”
— HumansFirst, statement following the July 18, 2026 national day of protest [10]
And then, four days after the protests, the electric grid itself intervened in the argument. On the morning of July 22, 2026, a transmission line outage in northern Virginia — home to the largest concentration of data centers on Earth — triggered the automatic protection systems of a massive block of data center load. The facilities’ control systems unplugged from the grid and transferred to backup power in milliseconds, and the sudden loss of demand produced a voltage disturbance that sensors registered from Washington, D.C. to Chicago. Residents reported flickering lights and appliances groaning; the grid required roughly ten minutes to fully stabilize, an eternity in a system where disturbances are normally corrected in milliseconds.[5] The event was not an isolated malfunction. The North American Electric Reliability Corporation’s 2026 State of Reliability report had already documented multiple incidents in 2025 in which more than a gigawatt of data center demand disconnected within moments of a transmission disturbance — including a February 2025 event that shed roughly 1,800 megawatts — and on May 4, 2026, NERC issued only the third Level 3 Essential Actions Alert in its history, devoted entirely to large computational loads.[6][59] NERC’s director of reliability assessment, John Moura, has stated the problem plainly:
“the grid is not designed to withstand the loss of 1,500MW data centers”
— John Moura, Director of Reliability Assessment and System Analysis, NERC [7]
A city ballot, a national day of protest, and a continental-scale grid disturbance — all within seven weeks of one another, all traceable to the same underlying phenomenon — mark the arrival of something new in American politics. This paper gives that phenomenon a name: Permit Populism.
Defining Permit Populism
Permit Populism is the political mobilization of ordinary residents, across partisan lines, around the ordinarily obscure machinery of infrastructure approval — zoning hearings, conditional-use permits, utility rate cases, interconnection queues, tax-abatement votes — as the primary arena in which the costs and benefits of the artificial intelligence economy are contested. It is populist in the classic sense: it pits a broad, self-identified community of ordinary people (ratepayers, homeowners, well-owners, parents of schoolchildren) against a concentrated set of powerful institutions (hyperscale technology companies, monopoly utilities, and the state and local officials perceived as accommodating them). And it is permit-centered because the permit — the discrete, local, legally contestable moment of authorization — is the one point in the entire trillion-dollar AI build-out where an ordinary citizen retains genuine leverage.
The term is deliberately chosen over more familiar labels. This is not classic NIMBYism, because the movement’s demands increasingly concern cost allocation and enforceability rather than pure exclusion; many of the most consequential policy responses of 2026 — Pennsylvania’s GRID Standards, Virginia’s consumption tax, Texas’s Senate Bill 6 — accept continued construction while restructuring who pays.[16][27][22] It is not simply a “techlash,” because the same voters who oppose an unconditioned data center next door often support artificial intelligence as a technology and welcome projects that arrive with binding commitments. And it is not a partisan movement: a Gallup poll conducted in March 2026 found roughly 70 percent of Americans opposed to new AI data centers in their own neighborhoods, a June 2026 Reuters/Ipsos survey found that only 14 percent of Americans would support a data center in their community, and one 2026 survey found Republicans somewhat more favorable to local data center development than Democrats — a partisan inversion of the usual environmental-siting politics.[9][10] A Marquette University Law School poll of Wisconsin voters found the share who believe data center costs outweigh their benefits rose from 55 percent to 70 percent within six months.[49]
Why is “Permit Populism” the right framework for the 2026 midterms specifically? Three reasons. First, the election calendar and the infrastructure calendar have collided: the November 3, 2026 federal midterm elections arrive at precisely the moment when the largest capital deployment in corporate history — roughly $725 billion in planned 2026 capital expenditure by Microsoft, Amazon, Alphabet, and Meta alone, up approximately 77 percent from about $410 billion in 2025 — is being converted from earnings-call announcements into physical substations, transmission corridors, diesel generator farms, and water withdrawals in specific congressional districts.[53][54] Second, electricity affordability has become the most legible face of the broader cost-of-living anxiety that both parties agree will dominate the cycle, and data centers have become the most legible face of electricity costs — fairly or not. As Jefferies analysts told investors in early 2026, data centers are emerging as the “villain” of the affordability narrative, and every candidate this cycle will be forced to take a position for or against them.[47] Third, the permitting arena scrambles conventional partisan coalitions in ways that make it electorally decisive in exactly the kinds of districts that decide control of Congress: suburban swing districts where a substation siting or a 21.7 percent annual rate increase, as Pennsylvania experienced in 2025, can move votes regardless of party registration.[48]
As Professor Dan Cassino of Fairleigh Dickinson University, who directs the university’s statewide poll, observed of the electoral environment:
“electricity prices are right now the most clear example of problems of affordability”
— Dan Cassino, Professor of Political Science, Fairleigh Dickinson University [46]
Thesis and Research Questions
The thesis of this paper is that the durable political divide emerging around AI infrastructure is not pro-AI versus anti-AI, but presumed consent versus conditional consent — and that the jurisdictions, parties, and candidates who master the grammar of conditional consent will convert Permit Populism from a threat into a governing coalition. The paper pursues five research questions, which double as the five questions every candidate in the 2026 cycle must be able to answer for any proposed facility in their district: Who pays? Who powers? Who benefits? Who decides? And who can stop it? These questions organize the entire manuscript. Sections 1 and 2 establish why data centers have become electoral infrastructure and develop the five questions in full. Sections 3 through 5 examine the underlying disputes over money, jobs, and environmental burdens. Section 6 analyzes seven governors as seven distinct political models for managing the collision. Sections 7 and 8 translate the analysis into district-level strategy and a model incentive framework. Section 9 synthesizes the findings into a five-pillar (plus one) compact that both parties could adopt, before the paper confronts the strongest counterarguments to its own thesis and concludes.
One important premise deserves emphasis at the outset, because it defines the analytical space in which the entire argument operates. On March 4, 2026, the White House convened seven of the largest AI companies and hyperscalers — Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI — to sign the Ratepayer Protection Pledge, under which the signatories committed to build, bring, or buy the new generation resources their data centers require, to cover the cost of all power-delivery infrastructure upgrades needed to connect them, to negotiate separate rate structures with utilities and state governments, and to keep paying for the power and infrastructure brought online for their facilities whether they use the electricity or not.[11][12] Whatever one concludes about the pledge’s enforceability — and Section 2 will conclude that its lack of binding mechanisms is its central weakness[13][14] — its very existence establishes a bipartisan premise of enormous consequence: AI expansion and ratepayer protection are no longer officially presented as opposing goals, by either party, at any level of government. The debate is no longer whether data centers should pay their own way. The debate is how that principle is defined, verified, and enforced. That is precisely the terrain on which Permit Populism operates.
Section 1: Why Data Centers Are Becoming Electoral Infrastructure
For most of the internet era, data centers were politically invisible. The “cloud” was a metaphor that did its ideological work well: computing seemed to happen nowhere in particular, and the warehouses that actually performed it were modest in scale, sited quietly in industrial parks, and cheap to serve. Artificial intelligence has destroyed that invisibility, for a reason that is ultimately thermodynamic rather than political. Training and serving large models requires computation at a density and continuity that older facilities never approached, and computation at that density is, in physical terms, the conversion of electricity into heat. The International Energy Agency estimates that global data center electricity consumption grew 17 percent in 2025 alone — more than five times the growth rate of overall global electricity demand — and projects that it will roughly double from 485 terawatt-hours in 2025 to about 950 terawatt-hours by 2030, approaching 3 percent of global electricity demand; consumption by AI-focused facilities is projected to triple over the same period.[50] In the United States, the IEA projects that data centers will account for nearly half of all electricity demand growth between now and 2030, by which point the country will consume more electricity for data processing than for the production of aluminum, steel, cement, chemicals, and all other energy-intensive goods combined.[52] The IEA’s executive director, Fatih Birol, has compressed the strategic point into six words:
“there is no AI without energy”
— Fatih Birol, Executive Director, International Energy Agency [51]
From digital abstraction to physical land use, then, is the first transformation: a hyperscale AI campus is now a heavy industrial facility by any traditional measure — a thousand or more acres, a gigawatt or more of demand, hundreds of diesel backup generators, its own substations, and water requirements that in some designs rival small municipalities. Physical facilities of that character have always been political objects in American local government. What is new is the speed, scale, and concentration of the build-out. Data center construction starts reached a record 25.2 billion dollars in January 2026 alone, the highest monthly total ever recorded.[55] The Electric Reliability Council of Texas reported in June 2026 that it was tracking more than 438,000 megawatts of proposed large-load connection requests — nearly 89 percent of them data centers — against an all-time system peak demand of 85,508 megawatts; the queue of aspirations is more than five times the size of the largest load the Texas grid has ever actually served.[19]
The second transformation is fiscal and financial visibility. The hyperscalers’ own earnings disclosures through the first quarter of 2026 have made the scale of the build-out impossible to ignore, and have simultaneously made the companies’ capacity to pay impossible to deny. In the first calendar quarter of 2026, Amazon reported 44.2 billion dollars of capital expenditure in a single quarter; Alphabet reported 35.67 billion, more than doubling year over year, with a Google Cloud contract backlog exceeding 460 billion dollars; Microsoft reported 30.88 billion in its fiscal third quarter, up 84 percent; and Meta raised its full-year 2026 capital expenditure guidance to between 125 and 145 billion dollars, citing higher component and data center costs.[53] Amazon’s roughly 200-billion-dollar 2026 program has been described as the largest single-year corporate investment commitment in history, and Goldman Sachs now projects a combined 5.3 trillion dollars of capital expenditure by the four largest hyperscalers between fiscal 2025 and fiscal 2030.[54] When a voter facing a 20 percent electricity rate increase reads that four companies plan to spend three-quarters of a trillion dollars in a single year on the facilities driving new demand, the political conclusion — they can afford to pay their own way — writes itself.
Third, and most decisive for electoral politics, is the political significance of the permitting chokepoint. Substations, water agreements, road improvements, and tax abatements all share one feature: they require affirmative public authorization, usually from bodies that meet in public, near the affected residents, on recorded votes. In a national political economy where most decisions of comparable magnitude — monetary policy, trade policy, platform governance — are made in venues far beyond any citizen’s reach, the county zoning hearing is a rare site of genuine, immediate, individually exercisable public power. Permitting concentrates leverage. It converts diffuse anxiety about artificial intelligence into a specific, winnable, local campaign with a date, a docket number, and a roll-call vote. That is why the movement’s energy flows there, and why research firm Data Center Watch estimates that local and state opposition blocked or delayed roughly 98 billion dollars in data center projects in a single recent quarter, with more than 300 data-center-related bills introduced across state legislatures in the first half of 2026.[9][56]
Fourth, data center policy is a coalition-crossing issue, and conventional partisan categories are insufficient to describe it. Consider the coalition that assembled in 2026 alone: a Tea Party co-founder organizing a national day of protest;[10] a Democratic governor of Virginia signing the nation’s first per-kilowatt-hour tax on data center electricity consumption;[27] the Republican governor of Texas directing his utility regulators to ensure that data centers pay all of their own infrastructure costs;[21] a Democratic governor of Arizona freezing data center tax exemptions;[33] a Republican White House convening hyperscalers to pledge ratepayer protection;[11] and Democratic and Republican members of Congress jointly advancing a Ratepayer Protection Act at an April 29, 2026 House Energy and Commerce subcommittee hearing.[14] Brookings survey research finds that while Democrats are somewhat more likely than Republicans to view data centers as “mostly bad” for household energy costs — 44 percent versus 33 percent — concern is substantial across both parties, and the issue’s valence flips depending on whether it is framed as technology policy (where Republicans are friendlier) or corporate cost-shifting (where both parties converge on hostility).[45] An issue with that structure does not sort by party. It sorts by proximity — to a project, to a substation, to a rate increase — which is to say, it sorts by district. That is what makes it electoral infrastructure in the most literal sense: it is embedded in the geography of representation itself, and it will be litigated district by district on November 3, 2026.
Section 2: The Five Questions Every Candidate Must Answer
Permit Populism can be reduced to five questions. They are simple enough to be asked at a town hall and rigorous enough to organize a regulatory proceeding. A candidate — for county commission, for public utility commission, for governor, for Congress — who can answer all five for any facility proposed in their jurisdiction is prepared for the 2026 cycle. A candidate who cannot answer them will have the answers supplied by opponents. This section develops each question in full; the sections that follow deepen the underlying disputes.
Question 1 — Who Pays?
The first question is the question of cost allocation, and it is the gravitational center of the entire controversy. A hyperscale campus does not merely buy electricity; it causes a cascade of capital expenditure across the energy system, and every dollar of that cascade must ultimately land on someone. The honest inventory is long. It begins with generation and capacity costs: the new power plants, or the extended lives of old ones, required to serve a load that may run at 80 to 90 percent utilization around the clock. It continues through the delivery system: new substations, high-voltage transmission, distribution upgrades, network reinforcements, and the interconnection studies that precede them. It includes backup generation and fuel infrastructure — the hundreds of diesel or gas generators and fuel storage systems that a single campus can require for reliability. It extends beyond the electrical system entirely, to water pipelines and treatment capacity, road improvements, and the fire, hazardous-materials, and emergency-medical capabilities that a facility of industrial scale demands of its host community. It includes the quieter fiscal costs: sales and property tax exemptions, and the public revenue foregone through them. And it includes two categories that are almost always omitted from groundbreaking press releases: stranded assets — the infrastructure built for a project that is delayed, downsized, or cancelled after the concrete is poured — and decommissioning, the eventual cost of safely retiring a facility whose technological useful life may be far shorter than the life of the transmission built to serve it.
The default machinery of American utility regulation, left to itself, tends to socialize much of this cascade. Regulated utilities recover approved capital investments from their entire ratepayer base; absent deliberate countermeasures, the substation built for a single hyperscaler enters the same rate base as the poles and wires serving every household. The Harvard Electricity Law Initiative’s influential 2025 study, Extracting Profits from the Public, documented the specific mechanisms — confidential special contracts reviewed in opaque proceedings, mismatches between federal transmission cost allocation and state ratemaking formulas, and co-location arrangements that remove existing generation from regional markets — through which data center energy costs migrate onto captive residential customers, and warned that utilities have affirmative profit incentives to let that migration occur.[41] Its lead author, Ari Peskoe, put the accountability problem succinctly after his team reviewed fifty regulatory proceedings across the country and concluded it was all but impossible to verify utilities’ assurances that no cost-shifting was occurring:
“we can’t just take the monopolist utility’s word for it”
— Ari Peskoe, Director, Electricity Law Initiative, Harvard Law School [42]
It is against this backdrop that the March 2026 Ratepayer Protection Pledge must be read. The pledge’s substantive commitments track the cost-causation inventory above with surprising fidelity: participating hyperscalers agree to provide or purchase the generation they require, to pay for the delivery-system upgrades their facilities cause, to negotiate separate utility rate structures rather than sheltering inside general service classes, and — critically — to continue paying for infrastructure brought online for them even when actual usage falls below expectations, which is the pledge’s answer to the stranded-asset problem.[11] The commitments were published in the Federal Register as a presidential proclamation declaring that they effectuate national policy.[12] The weakness is equally clear: the pledge is voluntary, contains no binding enforcement mechanism, no independent auditing, no penalties for noncompliance, and no defined methodology for determining what adequate cost coverage even means — and, as Ari Peskoe has emphasized, the tariff structures through which these costs actually flow are controlled by states and public utility commissions, not by the White House or the signatories.[13][44] Brookings scholars David Klaus and Mark MacCarthy have argued that the pledge’s terms must be converted into statute and tariff — and Congress has begun to move, with bipartisan support at the April 2026 House hearing for a Ratepayer Protection Act and a parallel Senate proposal conditioning priority grid interconnection on pledge-equivalent terms.[14] A Consumer Reports survey in May 2026 found that 75 percent of Americans were not confident the companies would keep their promises, and in June 2026 the Federal Energy Regulatory Commission issued show-cause orders to six regional grid operators requiring them to justify or reform their large-load interconnection rules.[15] The federal position thus gives the paper its bipartisan premise — AI expansion and ratepayer protection are officially compatible goals — while leaving the enforcement question squarely to the states and localities where Permit Populism lives.
Question 2 — Who Powers?
The second question goes beyond the familiar inquiry into whether a facility’s electricity is renewable, because a renewable-energy certificate purchased from a wind farm two thousand miles away does nothing for the physical grid serving the facility. The question of who powers a data center decomposes into a series of harder sub-questions. Is the generation genuinely additional — new capacity that would not exist but for the project — or is the facility simply absorbing existing supply and re-labeling it? Is the generation located where it can physically serve the facility, within the same deliverability area, or is it an accounting construct? Is it firm and dispatchable, available on a windless January night, or intermittent? Will the data center reduce its demand during grid emergencies — and is that commitment written into an enforceable tariff, or into a press release? Who receives power first during scarcity: the hospital, the household, or the training run? Will the facility’s arrival keep existing coal or gas plants open beyond their planned retirements, and will nuclear plants be restarted to serve it? Will ratepayers end up guaranteeing the economics of new generation through long-term contracts they never signed? And — a question that cuts in the developer’s favor — can the operator migrate workloads to other regions during stress, converting the facility from a reliability burden into a reliability resource?
Pennsylvania’s GRID program, examined at length in Section 6, offers the most developed state-level answer to date: developers seeking Commonwealth support must present an energy plan, must build, bring online, or buy the incremental capacity their demand requires and pay its full cost, must draw that capacity generally from the same deliverability area as the project, must meet escalating requirements for dispatchable clean firm generation located in-state — rising to 32 percent by 2035 — and must pay all interconnection, transmission, distribution, network-upgrade, and ancillary-service costs their demand causes in whole or in part.[16] The reliability dimension has acquired new urgency from the events described in the Introduction: NERC’s 2026 State of Reliability report and its May 2026 Level 3 alert establish that large computational loads are no longer passive consumers but dynamic grid actors whose synchronized protective behavior — a thousand facilities tripping to battery in the same four milliseconds — can itself destabilize the system, as the July 22, 2026 disturbance demonstrated from the Potomac to Lake Michigan.[6][5][59] “Who powers?” therefore now includes “who rides through?” — and the answer belongs in the interconnection agreement, not in the marketing material.
Question 3 — Who Benefits?
The third question requires disciplined disaggregation, because campaign rhetoric systematically conflates three categories of economic benefit that differ in duration, distribution, and reliability: temporary construction employment, permanent operating employment, and indirect regional development. The Indiana Meta project, developed fully in Section 4, illustrates the distinction with unusual clarity. The announced investment exceeds 10 billion dollars across a 1,500-acre campus of thirteen buildings; construction employment is expected to exceed 4,000 workers at peak; and permanent operating employment is expected to be approximately 300 positions.[23][24] The company has additionally committed more than 120 million dollars to local water, road, and utility infrastructure and pledged annual community payments tied to completed project phases, including one million dollars per year for twenty years to a local energy-cost assistance fund.[25][23]
Those numbers do not automatically make the project beneficial or unfavorable. What they establish is the correct analytical question, which every candidate should be able to recite: what is the public cost per durable job, and what other measurable benefits compensate for a relatively small permanent workforce? Benefits, properly evaluated, run through a checklist far broader than a jobs number: permanent payroll and its multiplier; construction wage standards and whether prevailing wages apply; registered apprenticeships that convert a temporary boom into lasting workforce capacity; local procurement commitments; school and municipal revenue net of abatements; infrastructure improvements with public value beyond the fence line; community investment funds; university partnerships; resilience improvements — such as grid hardening or water system upgrades — available to other customers; and, above all, whether each of these commitments is binding and clawback-secured or merely discretionary. Georgia Tech researchers studying host counties have found that employment and income gains are real but tilt toward metropolitan rather than rural hosts, and that local electricity prices rise on the order of 5 percent after a facility begins operations — a reminder that benefit and burden can arrive in the same envelope.[57]
Question 4 — Who Decides?
The fourth question exposes the jurisdictional fragmentation that makes AI infrastructure uniquely difficult to govern — and uniquely productive of populist grievance. A single project can implicate seven distinct decision-making arenas. County or municipal governments control zoning, land use, setbacks, and noise ordinances. State utility commissions control retail rates, cost recovery, and the approval of special contracts. Regional grid operators and reliability authorities assess interconnection and system impacts. State governments award tax incentives and, increasingly, set statewide standards. Federal agencies govern environmental reviews, interstate transmission, public lands, and — since the 2026 proclamation — declared national priorities.[12] Utilities negotiate confidential commercial arrangements whose terms even regulators struggle to scrutinize.[41] And developers frequently operate through special-purpose subsidiaries and code names that conceal the ultimate user until after key approvals are secured.
The consequences of this fragmentation are perverse in both directions. A project may be locally approved but electrically unviable, stranded for years in an interconnection queue no county commissioner controls. Or it may be supported by a governor, blessed by a utility, and financed by a hyperscaler — and rejected by the residents who were consulted last, as Monterey Park demonstrated. Permit Populism intensifies precisely where responsibility is fragmented but power is concentrated: when no single official can be held accountable for the whole decision, voters rationally treat every official as accountable for all of it, and the ballot box becomes the appellate court of last resort. The Harvard researchers’ observation about the sheer scale of what these fragmented processes are being asked to absorb captures the institutional mismatch:
“energy demand for entire cities regularly materializing out of thin air”
— Eliza Martin, Legal Fellow, Harvard Law School Environmental and Energy Law Program [43]
Question 5 — Who Can Stop It?
The fifth question is the one that distinguishes participation from power. Public engagement is incomplete — is, in fact, a ritual — unless a community can enforce the conditions under which approval was granted. The instruments of enforcement form a graduated arsenal. At the front end: permit denial; temporary moratoria of the kind Monterey Park adopted while it deliberated;[1] and conditional-use permits that embed operating requirements into the authorization itself. During and after construction: judicial review; utility or grid-operator interconnection rejection on reliability grounds; incentive clawbacks when employment or investment targets are missed; revocation for materially inaccurate disclosures — including concealment of the ultimate user; operating limits on hours, noise, or generator testing; water and noise enforcement with escalating penalties; financial assurance and decommissioning bonds that survive the developer’s corporate structure; and mandatory re-approval when a project materially expands its acreage, megawatts, generator count, or water use beyond what was originally authorized.
The crucial analytical point — and the point on which this paper departs from both the industry and the movement’s maximalists — is that the strongest policy is not necessarily an unconditional local veto. An absolute veto is a blunt instrument that invites the pathologies examined in the counterarguments section: indiscriminate exclusion, jurisdiction-shopping toward the weakest regulators, and the loss of genuinely beneficial projects. The strongest policy is a system in which government retains the demonstrated, credible authority to pause, modify, or revoke benefits when the developer fails to satisfy measurable commitments — a system, in other words, in which the answer to “who can stop it?” is “the public, through law, at any point where the bargain is broken.” Enforceable exit, not categorical exclusion, is the mature form of Permit Populism.
The Civic Load Test
The five questions can be operationalized as a single instrument, which this paper calls the Civic Load Test: before any major approval or incentive, a jurisdiction scores a proposed facility on cost responsibility (Who pays?), power adequacy and reliability citizenship (Who powers?), verified net benefit (Who benefits?), procedural legitimacy (Who decides?), and enforceable exit (Who can stop it?). A project that passes all five earns expedited, predictable treatment; a project that fails any one has identified exactly what must change. Two brief hypotheticals show the test at work.
Hypothetical Campus A — “Meridian Ridge.” A 600-megawatt campus on former agricultural land near a mid-sized county seat. The developer, a named hyperscaler, files a public energy plan committing to contract for new in-region generation including a firm dispatchable component; accepts a minimum-demand tariff paying for reserved capacity regardless of utilization; publishes peak demand, projected water use from reclaimed sources, and closed-loop cooling specifications; signs a community-benefit agreement with prevailing wages, apprenticeship ratios, a school-revenue floor, and funding for fire and hazmat training; accepts an expansion trigger requiring renewed approval beyond 700 megawatts; and posts a parent-company guarantee plus a decommissioning bond. Meridian Ridge passes the Civic Load Test. The correct policy response is speed: consolidated permitting on a defined timeline. Conditional consent has been earned.
Hypothetical Campus B — “Project Cobalt.” A subsidiary with no disclosed parent seeks rezoning of 900 acres, declining to identify the end user, the peak load, or the water source, citing competitive sensitivity. It requests the maximum statutory tax exemption, offers a non-binding letter promising “hundreds of jobs” without distinguishing construction from operations, proposes to take service under a general industrial tariff, and resists any expansion trigger. Project Cobalt fails the test on every element — not because a data center is unwelcome, but because nothing about the proposal is verifiable or enforceable. The correct response is neither approval nor a permanent ban, but a conditioned refusal that specifies exactly what disclosure and commitments would convert refusal into approval. The test thus disciplines both sides: it denies the developer presumed consent, and it denies the opposition costless absolutism.
Section 3: Ratepayers, Taxpayers, and Shareholders
Every dispute catalogued in this paper ultimately reduces to a contest among three legal identities that most Americans inhabit simultaneously without ever distinguishing them. The ratepayer finances regulated utility investments through monthly bills, under rate structures set by state commissions. The taxpayer finances incentives, public infrastructure, and public services through the general fisc, under budgets set by legislatures. The shareholder receives corporate returns and, in the classical theory of the firm, bears commercial risk in exchange. The core argument of this section is a single sentence: commercial risks should not migrate automatically from shareholders to households. The entire architecture of responsible data center policy — the Ratepayer Protection Pledge, Pennsylvania’s GRID Standards, Texas Senate Bill 6, Virginia’s new rate class — can be understood as a set of dams built across the channels through which that migration occurs.[11][16][22][30]
The channels themselves deserve enumeration, because each corresponds to a specific policy instrument. Special rate classes isolate very large loads into their own tariff so that the costs of serving them are recovered from them; Virginia’s State Corporation Commission created such a class in late 2025 for facilities of 25 megawatts and above, setting transmission and distribution demand charges at 85 percent and generation demand charges at 60 percent — meaning those customers must pay for the great majority of the capacity reserved for them whether they use it or not.[30] Minimum demand charges and take-or-pay electricity structures perform the same function contractually: the developer pays for reserved infrastructure even when utilization disappoints, which is precisely the commitment the White House pledge extracts — payment for power and infrastructure “whether they use the electricity or not.”[11] Interconnection deposits and financial security requirements, of the kind ERCOT’s Batch Zero process now demands before it will even study a large-load request, filter speculative applications out of the queue and ensure that study and upgrade costs are borne by applicants rather than socialized.[19][20] Cost-causation principles — the oldest doctrine in utility ratemaking — hold that costs should be borne by those who cause them; the entire reform project is, in one sense, simply the rigorous application of a principle regulators have always professed. Tax-exemption opportunity costs must be counted honestly: Virginia’s data center sales tax exemption saved the industry an estimated 1.9 billion dollars in fiscal 2025 alone, and Pennsylvania’s equivalent is projected to cost 517 million dollars annually by 2030 — sums that are invisible on any electricity bill but entirely real to any school board.[29][18] Infrastructure prepayment shifts timing risk to the developer. Stranded-capacity provisions assign the cost of cancelled or downsized projects to those who proposed them. And corporate guarantees from the ultimate parent company close the final escape hatch: a thinly capitalized special-purpose subsidiary must not be permitted to dissolve and leave the public holding infrastructure built on its promises.
Intellectual honesty, however, requires this section to complicate its own villain narrative, because the empirical literature does not support the proposition that data centers have, to date, raised average American electricity rates. The most rigorous causal study yet produced — a 2026 working paper by Asa Watten and Geoffrey Blanford of the Electric Power Research Institute and John Bistline of Watershed, using an instrumental-variables design to isolate causation from correlation — finds that data centers caused average retail electricity rates to fall modestly in the United States between 2015 and 2024.[38] The mechanism is not mysterious once stated: the grid carries enormous fixed costs that exist regardless of throughput, and a large, steady, around-the-clock customer spreads those fixed costs across more kilowatt-hours, lowering the average cost recovered from everyone. As Watten summarized the counterfactual:
“about 6% higher without the data centers built from 2019 to 2024”
— Asa Watten, Electric Power Research Institute, on average residential retail rates [39]
The same body of research, however, carries a warning label that the industry’s advocates quote less often: the historical result depended on spare capacity. States that experienced rate decreases were, as Brattle Group principal Ryan Hledik has emphasized, states that had room on their systems to absorb large customers without major new investment; as that headroom is exhausted and new demand begins to require new generation, transmission, and distribution at today’s elevated construction costs, the effect can reverse.[39] Analysis by the consultancy E3 attributes roughly half of the dramatic increase between recent PJM capacity auctions to load growth — while attributing the other half to market design changes, plant retirements, and supply-side factors, and finding no evidence of historical cost-shifting from data centers to Virginia households in the world’s largest data center market.[40] A Federal Reserve Bank of Dallas working paper reaches the theoretically honest conclusion: the net effect of data center load on customer bills is ambiguous, determined by whether fixed-cost spreading outruns new fixed-cost creation.[58] And Brookings analysis warns that without protective rate structures, residential rates could rise 15 to 40 percent by 2030 in affected regions.[56]
The paper therefore refuses the easy claim that every data center raises bills — a claim that is empirically false for the past decade — and refuses equally the complacent claim that no data center ever will. The stronger, defensible formulation, which candidates should adopt verbatim, is this: rate effects depend on generation availability, rate design, infrastructure timing, customer concentration, and whether the developer remains financially responsible for capacity constructed on its behalf. Where those five variables are governed well, large loads can lower everyone’s bills; where they are governed badly, households insure the AI industry’s ambitions. Permit Populism, at its most sophisticated, is the political demand that they be governed well — and the Harvard Electricity Law Initiative’s finding that verification is nearly impossible under current confidentiality practices explains why the demand is expressed with such distrust.[41][42]
Section 4: Jobs, Automation, and the Local Employment Dispute
No claim in the data center debate is made more often, or examined less carefully, than the jobs claim. The characteristic pattern is now well documented: a construction boom of genuine scale and brevity, followed by an operating workforce of striking smallness relative to the land, power, and public money involved. The pattern is not a scandal; it is the technology. A hyperscale facility is, by design, one of the most automated industrial environments ever constructed — a building whose product is computation and whose permanent staff exists chiefly to maintain, secure, and administer machines that run themselves. The political dispute arises because the facility is nonetheless marketed to host communities in the rhetorical register of the twentieth-century factory, and voters increasingly know the difference.
The Indiana case study rewards close attention because it is, in most respects, a best-practice example rather than a cautionary tale — which makes the arithmetic it exposes all the more instructive. In February 2026, Meta and the State of Indiana formally unveiled the Lebanon campus in Boone County’s LEAP Innovation and Research District: an investment exceeding 10 billion dollars, thirteen buildings totaling four million square feet across 1,500 acres, one gigawatt of capacity, more than 4,000 construction jobs at peak, and approximately 300 permanent operating positions once complete.[23][24] Meta committed more than 120 million dollars to local infrastructure — including roughly 45 million dollars for public road improvements and more than 75 million dollars for water infrastructure — pledged to pay the full cost of the energy the campus uses so that consumers do not, committed community investments tied to each completed phase, and pledged one million dollars per year for twenty years to the Boone REMC Community Fund for direct energy-cost assistance to families.[23][25] Governor Mike Braun, celebrating the project, nonetheless articulated the constraint that defines the entire politics of the issue — growth, he said, cannot come
“at the expense of Hoosiers’ quality of life or their electric bills”
— Mike Braun, Governor of Indiana, at the Meta Lebanon announcement [24]
Now the arithmetic. Three hundred permanent positions across 1,500 acres is one durable job per five acres — a land-to-employment ratio that no factory, warehouse, or hospital would survive in an incentive review. Per megawatt, it is roughly 0.3 permanent jobs. Whether that represents a good public bargain cannot be answered by the jobs number alone; it depends on the denominator of public cost and the full numerator of public benefit. The correct evaluative frame, developed in Section 2, asks: permanent jobs per acre, per megawatt, and per incentive dollar; the wage level and durability of those jobs (data center operations roles are, in fairness, well paid and stable); the value of construction employment when — and only when — wage standards and apprenticeship requirements convert a transient boom into workforce capacity; the tax revenue that flows even from a lightly staffed facility, which can be transformative for a rural school district; and the infrastructure whose public value outlasts the project. Meta’s Lebanon commitments are strong on several of these dimensions precisely because Indiana negotiated phase-linked, published obligations rather than accepting aspirational totals.
Four disputes recur wherever the employment question is litigated, and candidates should be fluent in all of them. First, should AI facilities be judged like factories at all? The industry answers no — the correct comparison, it argues, is to capital-intensive utility infrastructure such as power plants, which are also lightly staffed and also indispensable. The answer matters because it determines the benchmark against which 300 jobs is judged. Second, union construction and prevailing wages: a four-thousand-worker peak is a genuine prize for the building trades, and jurisdictions that attach prevailing-wage and apprenticeship-ratio conditions convert it into lasting regional capacity; those that do not may import traveling crews and export the payroll. Third, local hiring versus imported technical labor: the permanent roles require specialized skills that small host communities rarely possess on day one, which is why workforce-development promises — like the Boone County Career Collaborative connecting three school districts to employers — must be evaluated as pipelines with enrollment numbers, not as press-release garnish.[25] Fourth, automation inside the facility will not stand still: the staffing assumptions of 2026 should be treated as a ceiling, not a floor, and incentive agreements should therefore key clawbacks to verified headcount rather than projected headcount.
The deepest reframing this paper urges is to treat public revenue as an alternative — and often superior — form of economic benefit. A facility that employs 300 people but reliably remits tens of millions of dollars annually to schools, counties, and energy-assistance funds may serve a community better than a facility employing 1,500 people behind an abatement that exempts it from the tax base for two decades. Virginia’s consumption tax, Arizona’s exemption pause, and Meta’s phase-linked community payments all represent, in different idioms, the same conclusion: where the durable jobs are few, the durable dollars must be many, and they must be binding.[27][33][23]
Section 5: Water, Noise, Land, and Emergency Services
If cost allocation is the intellectual center of Permit Populism, the sensory and environmental burdens are its emotional center. Electricity bills arrive monthly; the hum of a cooling plant arrives nightly. This section inventories the four burden categories that dominate local proceedings, and the disclosure architecture that responsible policy now attaches to them.
Water
Water disputes begin with a distinction that project marketing routinely blurs: annual withdrawal versus consumptive use. A facility may withdraw large volumes and return most of them, or withdraw less and evaporate nearly all of it; the watershed experiences the two very differently. The second distinction is potable versus reclaimed supply — whether the facility competes with households for treated drinking water or runs on effluent that would otherwise be discharged. Closed-loop cooling systems, which Meta specified for the Lebanon campus, dramatically reduce ongoing consumption at the price of somewhat higher electricity use, illustrating that water and energy burdens trade off against each other and must be disclosed together.[25] Watershed-restoration claims — pledges to be “water positive” by funding projects elsewhere — deserve the same scrutiny as renewable-energy accounting: restoration in a distant basin does not refill the aquifer under the neighbors’ wells. Mature water policy therefore requires drought triggers and curtailment schedules written into the permit, specifying exactly what the facility reduces, and in what order relative to other users, when scarcity arrives; and public disclosure of Water Usage Effectiveness, the industry’s own efficiency metric, so that performance can be tracked against promises. Pennsylvania’s GRID framework points the way, requiring qualifying projects to disclose estimated peak demand, annual water use and source, campus acreage, energy composition, and both Power Usage Effectiveness and Water Usage Effectiveness, while addressing traffic, noise, lighting, air quality, and emergency coordination through community-benefit commitments.[16] In water-constrained states the stakes are existential to the project pipeline itself: Arizona’s 2026 session, described in Section 6, turned substantially on the collision between data center growth and groundwater anxiety.[34]
Noise
Noise is the burden most likely to convert a supportive neighbor into an organized opponent, because it is continuous, involuntary, and — in its low-frequency components — notoriously difficult to escape indoors. The sources are cumulative: cooling systems whose fans run around the clock; transformers whose electromagnetic hum is tonal and penetrating; backup generators whose periodic testing produces the loudest single events a facility generates; and construction activity that can precede all of the above by two years or more. The cumulative low-frequency character of the problem matters technically, because conventional A-weighted noise limits under-measure exactly the frequencies that travel farthest and annoy most. Responsible permitting therefore specifies residential setbacks calibrated to modeled — and later measured — sound at the property line; continuous third-party monitoring with public dashboards rather than complaint-driven enforcement; generator testing windows confined to weekday daytime hours; and escalating penalties that make chronic exceedance more expensive than acoustic remediation. Noise was among the specific grievances that powered both the Monterey Park campaign and the July 18 national protests, and it appears by name in Pennsylvania’s community-benefit requirements and in the Texas governor’s 2026 directives.[4][8][16][21]
Land
Land disputes take five recurring forms. Agricultural conversion — the Lebanon campus occupies 1,500 acres of former Indiana farmland — raises both sentimental and food-system objections that resonate powerfully in rural districts.[24] Industrial clustering concentrates burdens: once a region acquires a reputation for available power and fiber, projects multiply until the cumulative footprint, traffic, and grid stress differ in kind from any single facility, as northern Virginia’s experience demonstrates. Transmission corridors extend the land question far beyond the campus, imposing easements on landowners who receive none of the project’s benefits — a classic recipe for cross-jurisdictional resentment. Visual impact — windowless buildings at warehouse scale, security fencing, substation infrastructure — transforms landscapes in ways zoning boards struggle to condition. And property-value disputes, though empirically contested, are politically decisive: the South Whitehall, Pennsylvania resident who asked a reporter whether a proposed facility would make it impossible to sell her house spoke for a constituency that polling suggests is enormous.[48]
Emergency Services
The least discussed burden category may impose the most acute unfunded mandate. A hyperscale campus presents hazards that small-town fire departments were never resourced to manage: fire suppression in high-voltage electrical environments; battery energy storage incidents, whose thermal-runaway chemistry requires specialized training and equipment; bulk fuel storage for generator fleets; hazardous materials in cooling and treatment systems; and the cyber-physical security profile of nationally critical infrastructure. The question a candidate must be able to answer is direct: who finances the specialized training, apparatus, and staffing that the facility’s risk profile demands? The defensible answer — embedded in the model rules of Section 8 — is that the project does, through binding service agreements negotiated before occupancy, not through after-the-fact appeals to a county budget that the facility’s own abatement may have shrunk. Emergency coordination is, notably, among the enumerated subjects of Pennsylvania’s required community-benefit commitments — evidence that at least one state has recognized the mandate and priced it into the bargain.[16]
Section 6: Seven Governors, Seven Political Models
If Permit Populism is the demand side of the new politics, the nation’s governors are its supply side — the laboratory in which distinct governing models are being tested in real time, eighteen months ahead of any federal statute. This section examines seven governors, chosen not for prominence alone but because each embodies a genuinely different answer to the question of how a state should meter its consent. Together they map the full strategic space available to candidates in 2026, and several of them are themselves plausible national candidates whose data center records will follow them.
Gretchen Whitmer (Michigan) — Energy Restoration
Michigan’s model connects industrial competitiveness to the physical restoration of firm generation. The centerpiece is Palisades: the 800-megawatt nuclear plant on Lake Michigan that shut down for decommissioning in May 2022 and is now, under Holtec International’s ownership, in the final stages of the first restart of a decommissioned nuclear plant in American history — supported by roughly 300 million dollars in state funding championed by the Whitmer administration and a 1.52 billion dollar federal loan, with two small modular reactors planned on the same campus to add a further 680 megawatts in the 2030s.[31][32] Whitmer’s 2026 State of the State framing claimed the milestone for Michigan as the first state to restart a nuclear power plant, and her broader argument is explicit: restored firm generation
“will lower energy costs, reaffirm Michigan’s clean energy leadership”
— Gretchen Whitmer, Governor of Michigan, on the Palisades restart and planned SMRs [31]
The analytical question the Michigan model poses: should states rehabilitate legacy generation for broad public reliability, or primarily to compete for new industrial and AI loads? The two purposes overlap but are not identical — capacity restored “for everyone” can be quietly committed to a single customer, and the restart’s public financing sharpens the question of who holds first claim on its output. The lesson, however, is one every state is now learning: energy supply can itself become an economic-development incentive, arguably the most valuable one a state can offer in a power-constrained decade.
Gavin Newsom (California) — Innovation with Consumer Constraints
California presents the most paradoxical model: the state is simultaneously the world’s AI software capital, an aggressive public-sector adopter of the technology, and — at the municipal level — the site of the nation’s first voter-enacted data center ban. In June 2026 Newsom announced a first-of-its-kind partnership making Anthropic’s Claude available to all state agencies, cities, and counties at a 50 percent discount, with free workforce training and technical support, building on his executive orders on generative AI in government.[36] His framing was carefully calibrated to the labor politics of the moment:
“AI should not replace the human work of government”
— Gavin Newsom, Governor of California, announcing the Anthropic partnership [37]
Yet the same month, Monterey Park’s Measure NDC passed with 88 percent of the vote, and the state’s high retail electricity rates make new large loads politically radioactive in ways the governor’s office cannot simply override.[2][3] California has advanced regional electricity-market coordination and public-sector AI adoption while remaining conspicuously cautious about the physical infrastructure layer. The analytical question: can a state lead in AI applications while remaining restrictive toward the physical infrastructure and utility costs required to sustain them — in effect, exporting its compute demand to Arizona, Texas, and Virginia? The lesson generalizes: model leadership does not automatically produce infrastructure abundance, and a jurisdiction can be simultaneously the greatest beneficiary of AI and the least willing host of its machinery.
Josh Shapiro (Pennsylvania) — Conditional State Support
Pennsylvania has produced the most complete template for what this paper calls conditional consent. The Governor’s Responsible Infrastructure Development (GRID) Standards, first sketched in Shapiro’s February 2026 budget address and released in full on May 27, 2026, condition all Commonwealth support — coordinated project assistance, permitting certainty, and access to state tax incentives — on a demanding, published checklist: an energy plan demonstrating how the project meets its demand without imposing costs on other ratepayers; an obligation to build, bring online, or buy incremental capacity and pay its full cost; a locational requirement that capacity generally come from the project’s own deliverability area; escalating percentages of in-state dispatchable clean firm generation, reaching 32 percent by 2035; full payment of all interconnection, transmission, distribution, network-upgrade, and ancillary-service costs the project causes; disclosure of peak demand, water use and source, acreage, energy composition, PUE and WUE; community-outreach plans with public hearings; and job, wage, and environmental commitments.[16] The administration is simultaneously moving to fold the state’s 517-million-dollar-a-year data center sales tax exemption into the GRID framework and to codify the standards in legislation, and has issued a planning toolkit to help municipalities evaluate proposals on their own terms.[18][16] Shapiro’s formulation is the purest one-sentence expression of the model:
“we have a responsibility to set strict accountability standards”
— Josh Shapiro, Governor of Pennsylvania, releasing the GRID Standards [17]
The analytical question: can accelerated permitting be exchanged for stronger obligations rather than weaker regulation? Pennsylvania’s wager is that it can — that developers value certainty and speed enough to accept binding standards as their price. The lesson: “fast permits for high standards” may prove more politically durable than either unrestricted subsidy or blanket prohibition, because it gives growth advocates, consumer advocates, labor, and environmental constituencies each something enforceable.
Greg Abbott (Texas) — Growth with Ratepayer Protection
Texas demonstrates that Permit Populism can emerge from within conservative, pro-business politics — indeed, that it may be most vivid there, because the gap between the state’s open-for-business identity and its new demands on the industry measures how much the politics have moved. The pressure is quantitative: ERCOT’s June 2026 materials report more than 438,000 megawatts of proposed large-load demand in the connection queue, nearly 90 percent from data centers, against an all-time peak of 85,508 megawatts.[19] The legislature acted first with Senate Bill 6 in June 2025, requiring large power users to help fund their own interconnection costs and empowering ERCOT to curtail new large loads during grid emergencies.[22] The Public Utility Commission then approved ERCOT’s Batch Zero framework — the first batch-study process for large loads at any American grid operator, requiring substantial financial security from applicants and studying projects of 75 megawatts and above collectively so the grid’s real capacity is allocated to real projects.[20][19] And in mid-2026 Abbott directed his regulators to go further: to require data centers to pay for all of their electric infrastructure costs so that no residential ratepayer is burdened, to examine water-efficient cooling requirements, to improve electricity and water reporting, to reconsider outdated incentives, and to address siting and noise — with the PUC and ERCOT ordered to respond by July 17, 2026.[21][22][35] The analytical question: can a state famous for rapid development preserve its pro-growth identity while imposing new obligations on its largest customers? Texas’s answer so far is that cost discipline is being reframed not as regulation but as the defense of the small ratepayer against the large one — a framing perfectly native to the state’s populist tradition. The lesson: Permit Populism is not a left-coded phenomenon; in Texas it wears boots.
Mike Braun (Indiana) — The Megadeal Model
Indiana represents the investment-attraction model in its most disciplined contemporary form, examined in detail in Section 4: the Meta Lebanon campus at more than 10 billion dollars and 1,500 acres, 4,000 construction jobs at peak, approximately 300 permanent positions, more than 120 million dollars in committed local infrastructure, full payment for the campus’s energy, and phase-linked community payments including twenty years of energy-assistance funding.[23][24][25] The analytical question: what combination of tax revenue, permanent jobs, community payments, and infrastructure improvements justifies a megaproject? Indiana’s implicit answer is that the justification must be assembled from binding components rather than asserted from the headline — and Braun’s own insistence that projects cannot come at the expense of quality of life or electric bills concedes the premise of the critics even while welcoming the investment.[24] The lesson: the headline investment number is not a sufficient measure of public return; a ten-billion-dollar announcement can coexist with a poor local bargain or a good one, and only the enforceable schedule of obligations distinguishes them.
Katie Hobbs (Arizona) — Water-Constrained Industrial Policy
Arizona is where the AI build-out meets hydrological arithmetic. The state hosts one of the nation’s largest data center markets — roughly 98 operating facilities with 86 more planned or under construction — while simultaneously anchoring the TSMC semiconductor complex and confronting structural water scarcity.[34] The 2026 legislative session made data centers its dominant subject, with the issue arising in more than eighty hearings and floor proceedings and more than fifty energy bills introduced; the budget Hobbs signed in June 2026 imposed a three-year moratorium on the state’s data center sales tax exemption, barring new applications from July 1, 2026 through June 30, 2029, after the governor had sought outright repeal.[34][35] Hobbs framed the pause as saving 57 million dollars redirected to childcare, rural hospitals, and food assistance, and paired it with new utility reporting requirements on data center grid connections — while pointedly declining to oppose the facilities themselves where water and grid conditions permit.[33][34] Her formulation is the sharpest single sentence any governor has produced on the subject:
“Arizona is a partner in progress. We are not a blank check.”
— Katie Hobbs, Governor of Arizona, on the data center tax-exemption moratorium [33]
The analytical question: how should a state prioritize semiconductor fabrication, data centers, housing, agriculture, and population growth when all compete for constrained water and electricity? Arizona’s emerging answer is triage by strategic value and employment intensity — protecting high-employment advanced manufacturing while withdrawing subsidy from lower-employment computing loads until their resource terms improve. The lesson: states may legitimately distinguish among industrial customers, and “strategically important manufacturing first, subsidized computing later” is a coherent, politically saleable ranking. The coda is instructive about revealed preferences: in the final two weeks before the moratorium took effect, developers filed 113 exemption applications — nearly as many as the program had received in its entire thirteen-year history — a stampede that says more about the subsidy’s value to the industry than any lobbying brief.[34]
Abigail Spanberger (Virginia) — Consumption-Based Accountability
Virginia, host to the world’s densest data center cluster, has executed the most consequential single policy pivot of 2026: from unconditional cluster expansion toward fiscal and energy accountability keyed to consumption. After a budget standoff that nearly produced the state’s first government shutdown — driven by a bipartisan Senate bloc seeking early repeal of the sales tax exemption that saved the industry 1.9 billion dollars in fiscal 2025 — lawmakers and Governor Spanberger reached a compromise she signed on June 30, 2026: the exemption survives, but the Commonwealth now levies the nation’s first Data Center Electricity Consumption Tax, 1.1 cents per kilowatt-hour on all electricity a data center consumes — utility-supplied, competitively supplied, or self-generated behind the meter — effective July 1, 2026, capped at 600 million dollars annually with excess refunded, and sunsetting June 30, 2028.[27][28][29] The tax rides atop the State Corporation Commission’s new dedicated rate class for loads of 25 megawatts and above, with demand charges structured so large customers pay for the great majority of capacity reserved for them, and atop strengthened backup-generator standards, expanded local assessment tools, and an enlarged environmental-quality role over noise and water.[30][28] Spanberger’s framing:
“We want data centers to pay their fair share”
— Abigail Spanberger, Governor of Virginia, on the first-of-its-kind consumption tax [26]
The analytical question: should data center taxation be based on property value, revenue, energy consumption, or environmental burden? Virginia’s answer — consumption — has a compelling logic for this industry in particular: it scales automatically with the facility’s actual draw on the shared system, it cannot be minimized by depreciation schedules or corporate structure, and by reaching behind-the-meter self-generation it closes the most obvious avoidance route.[28] For a 500-megawatt campus the levy approximates 48 million dollars a year — real money, yet visibly survivable for signatories of a 700-billion-dollar capital program.[53] The lesson: the world’s largest data center market has concluded that hosting the cluster and taxing its energy are compatible, and more than twenty-five states are now studying the template.[28]
The Seven Models Compared
Read together, the seven governors define a spectrum of consent-metering strategies rather than a partisan divide. Whitmer supplies power and calls it development; Newsom adopts the software while his municipalities reject the hardware; Shapiro sells speed for standards; Abbott polices cost-shifting in the name of the small ratepayer; Braun negotiates the megadeal’s components; Hobbs triages by water and employment; Spanberger meters the electrons and taxes them. Four of the seven are Democrats and three Republicans, yet the policy instruments interleave without regard to party — the strongest evidence in this paper that Permit Populism is a governing problem, not an ideological one. A summary table follows.
| Governor (State) | Political Model | Signature Instrument (2026) | Lesson |
| Gretchen Whitmer (MI) | Energy Restoration | Palisades nuclear restart; $300M state support; planned SMRs | Energy supply can itself become an economic-development incentive. |
| Gavin Newsom (CA) | Innovation with Consumer Constraints | Anthropic Claude partnership for all state and local agencies at 50% discount | Model leadership does not automatically produce infrastructure abundance. |
| Josh Shapiro (PA) | Conditional State Support | GRID Standards: full cost responsibility; in-region capacity; 32% clean firm by 2035 | Fast permits for high standards is more durable than subsidy or prohibition. |
| Greg Abbott (TX) | Growth with Ratepayer Protection | SB 6; ERCOT Batch Zero; directive that data centers pay all infrastructure costs | Permit Populism can emerge within conservative, pro-business politics. |
| Mike Braun (IN) | The Megadeal Model | Meta Lebanon: $10B+, 1 GW, 300 permanent jobs, $120M+ bound infrastructure commitments | The headline investment number is not a sufficient measure of public return. |
| Katie Hobbs (AZ) | Water-Constrained Industrial Policy | Three-year moratorium on data center tax exemption (2026–2029); utility reporting | States may distinguish strategic manufacturing from low-employment computing loads. |
| Abigail Spanberger (VA) | Consumption-Based Accountability | First statewide tax on data center electricity: $0.011/kWh, capped at $600M/yr | The largest market is moving from cluster expansion to fiscal and energy accountability. |
Section 7: Policy Emphasis by Political Geography
National frameworks win legislatures; district-calibrated emphasis wins elections. The five questions of Section 2 are universal, but the weight a credible candidate places on each varies systematically with district type — and the variance is substantive, not cosmetic, because different geographies genuinely bear different burdens and hold different leverage. This section translates the paper’s analysis into the six district archetypes that will decide the 2026 map, offering in each case the policy emphasis that is most credible — meaning most consistent with the district’s material interests and civic culture — rather than a slogan. The table below presents the full matrix; the paragraphs that follow elaborate its logic.
| District context | Most credible policy emphasis |
| Conservative or rural | Private-property protection; local authority; water security; no corporate cost shifting; disclosure of foreign ownership and true power requirements. |
| Progressive or urban | Environmental justice in siting; union labor and prevailing wages; emissions controls; ratepayer protection; transparency; enforceable community benefits. |
| Suburban or competitive | Electricity bills; noise; school revenue; setbacks; funded emergency services; independently verified employment claims. |
| Technology-centered | Faster approvals for projects meeting high published standards; clean firm generation; workforce pipelines; predictable statewide rules. |
| Energy-producing | New generation and transmission investment; host-community compensation; protection against boom-and-bust development via minimum-payment structures. |
| Water-constrained | Reclaimed water mandates; consumption caps with drought triggers; no tax incentives without a verified water budget. |
In conservative and rural districts, the credible emphasis flows from property and self-government: protection of private-property rights against transmission easements and industrial encroachment; defense of local decision authority against state preemption; water security for wells and agriculture; an absolute bar on corporate cost-shifting to rural cooperatives’ members; and disclosure of foreign ownership and true power requirements — a transparency demand with particular resonance where developers arrive behind anonymous subsidiaries. In progressive and urban districts, the same underlying framework is credibly emphasized as environmental justice in facility siting; union labor and prevailing wages on the construction boom; emissions controls on generator fleets; ratepayer protection framed as anti-monopoly consumer politics; and enforceable — not aspirational — community-benefit agreements. In suburban and competitive districts, where control of Congress will actually be decided, the emphasis is relentlessly kitchen-table: electricity bills first, then noise, school revenue, setbacks, emergency-service funding, and independently verified employment claims — the checklist of a homeowner who is neither pro- nor anti-technology but is emphatically pro-verification.[48]
In technology-centered districts, credibility runs the other way: candidates support faster, consolidated approvals for projects meeting high published standards — the Shapiro trade — plus investment in clean firm generation, workforce pipelines into operations careers, and predictable statewide rules that spare good projects the veto-by-attrition of serial local proceedings. In energy-producing districts, data center demand is opportunity: the emphasis is new generation and transmission investment, host-community compensation for the burdens of exporting power, and protection against boom-and-bust development through minimum-payment structures that keep paying when the load does not materialize. In water-constrained districts, the emphasis is reclaimed-water mandates, consumption caps with drought triggers, and a hard rule that no tax incentive issues without a verified water budget — Arizona’s politics, generalized.[34] Across all six geographies the universal message is identical, and it is the sentence this paper proposes as the movement’s credo: build what strengthens the community; condition what imposes risk; reject what hides its costs.
Section 8: Ten Rules for Responsible Incentive Agreements
The preceding sections diagnose; this one drafts. What follows is a model checklist of ten rules for any state or local incentive agreement with a data center developer — each rule stated as an obligation, each traceable to a failure mode documented earlier in the paper, and each already piloted somewhere in the 2026 policy landscape. Together they constitute the operating system of conditional consent.
Rule 1 — Name the ultimate user. No major incentive should be approved while the end user remains concealed behind a development entity or code name. Anonymity defeats every other safeguard: a community cannot assess creditworthiness, track-record, or parent-guarantee value for a counterparty it cannot identify. Disclosure at application, under seal if commercially necessary and publicly before final approval, is the minimum price of public money.
Rule 2 — Publish the complete power plan. The application must state peak demand, expected load factor, generation sources and their location and firmness, and interconnection requirements — the disclosure core of Pennsylvania’s GRID Standards.[16] A jurisdiction that cannot see the power plan is not approving a project; it is co-signing a mystery.
Rule 3 — Require full cost responsibility. The project funds all directly caused grid, water, road, and emergency-service investments, under cost-causation principles applied without exception — the shared substance of the White House pledge, Texas Senate Bill 6, and the Abbott directives.[11][22][21]
Rule 4 — Use minimum-payment obligations. Developers pay for reserved infrastructure even when utilization falls short of forecast — take-or-pay in structure, matching Virginia’s demand-charge floors and the pledge’s whether-they-use-it-or-not commitment.[30][11] This single rule neutralizes the stranded-asset risk that speculative interconnection queues would otherwise transfer to households.[19]
Rule 5 — Separate construction from permanent employment. Incentives contain distinct, verified targets and distinct clawbacks for construction employment and permanent operating employment, so that a 4,000-job construction peak can never be rhetorically laundered into the permanent-jobs column.[24]
Rule 6 — Mandate a community-benefit agreement. Binding commitments must cover labor standards, schools, emergency services, noise, roads, water, and public amenities, with third-party enforcement rights — the difference between Meta’s phase-linked, published Lebanon obligations and the unenforceable goodwill letter of the hypothetical Project Cobalt.[23][25]
Rule 7 — Require parent-company guarantees. A thinly capitalized project subsidiary must never be the public’s only counterparty. The guarantee of the ultimate parent — the entity whose earnings calls announce the capital program — must stand behind every obligation the subsidiary undertakes.[53]
Rule 8 — Create expansion triggers. Material increases in acreage, megawatts, generator count, or water use beyond the approved envelope require renewed approval. Without this rule, the first permit becomes a perpetual option on unlimited growth, and every safeguard above is diluted by accretion.
Rule 9 — Publish annual operating data. Electricity and water consumption, emissions, verified employment, tax payments, and community contributions are reported annually and audited — extending the disclosure logic of GRID’s PUE and WUE requirements and Arizona’s new utility reporting into the operating life of the facility.[16][33] Trust that cannot be verified will not survive an election cycle; the Consumer Reports finding that three-quarters of Americans doubt the industry’s promises is the measured cost of unverifiable commitments.[15]
Rule 10 — Fund eventual closure. Developers post a decommissioning bond or equivalent financial assurance at approval. Server generations turn over in a few years; buildings and substations last decades; the gap between those lifespans is a liability, and it must be pre-funded by the party that created it, not discovered by the county that inherits it.
The ten rules are deliberately symmetrical in their politics. They give consumer advocates enforceability, labor verified standards, environmental interests disclosure and mitigation, local governments continuing authority — and they give developers the one thing serial ad-hoc opposition can never provide: a published, stable, satisfiable definition of yes.
Section 9: What Have We Learned? The Permit Populism Compact
The analysis now assembles into its synthesis: a compact of six pillars — five drawn directly from the five questions, and a sixth that the events of 2025–2026 have made unavoidable — offered as a model that both political parties could adopt without abandoning their broader philosophies. Each pillar is stated as a standard a project either meets or does not.
Pillar 1 — Cost Responsibility. Data center customers pay the full incremental cost of serving their facilities: generation, transmission, substations, distribution, and dedicated infrastructure, under minimum-payment structures that survive under-utilization. This is the settled answer to Who Pays, now professed — whatever the enforcement gaps — by a Republican White House, Democratic and Republican governors, and the industry itself.[11][16][21][27]
Pillar 2 — Power Adequacy. Projects provide additional, deliverable generation in their own region with a firm dispatchable component; they participate in reliability planning; and they identify in advance which workloads can be curtailed during emergencies. This is the answer to Who Powers, and its locational and firmness requirements are already law-in-practice in Pennsylvania and policy-in-motion in Texas.[16][20]
Pillar 3 — Shared Benefit. Projects create measurable local value — employment with verified counts, tax revenue net of abatements, workforce development with enrollment numbers, community investment, and infrastructure improvements of public value — and the commitments are binding rather than discretionary. This is the answer to Who Benefits.[23][24]
Pillar 4 — Democratic Process. Residents receive timely notice, accessible project information including the identity of the ultimate user, genuine public hearings, and sufficient opportunity to influence major design decisions before approval — not after. This is the answer to Who Decides, and Monterey Park is the standing demonstration of what happens when it is skipped: the process reasserts itself at the ballot box, permanently.[1][3]
Pillar 5 — Enforceable Exit. Government retains the authority to modify permits, recover incentives, or suspend benefits when a project fails to satisfy binding commitments, secured by parent guarantees, expansion triggers, audited annual data, and decommissioning assurance. This is the answer to Who Can Stop It — enforceable exit rather than categorical exclusion.
Pillar 6 — Grid Citizenship. A pillar the original five-question framework did not require but 2026 has: large computational loads accept reliability obligations commensurate with their scale — disturbance ride-through capability, high-resolution operational data sharing, participation in the modeling and standards work NERC has now made mandatory in all but name, and compensation for system services their protective behavior consumes.[6][59] The July 22, 2026 disturbance settled the argument: a gigawatt-class facility is a grid actor, and citizenship in the grid, like citizenship anywhere, carries duties as well as rights.[5]
The Compact’s political architecture is its point. Pro-growth officials receive predictable and accelerated permitting for compliant projects. Consumer advocates receive cost protections with teeth. Labor receives enforceable employment standards. Environmental interests receive disclosure and mitigation. Local governments retain meaningful authority. Developers receive a stable approval pathway and a satisfiable definition of consent. No participant is asked to renounce a philosophy — only to accept that the era of presumed consent has ended.
Important Counterarguments — And Why They Sharpen Rather Than Defeat the Thesis
A credible paper must challenge its own thesis, and five counterarguments deserve unflinching statement in their strongest forms.
First: local opposition can become indiscriminate NIMBYism. Absolute local vetoes may block nationally important infrastructure, push projects toward jurisdictions with the weakest protections, and raise the cost of electricity and computing for everyone. The objection is real: Monterey Park’s permanent ban forecloses even a future project that would pass every test in this paper, and the campaign’s own organizers have announced their intent to pursue neighboring jurisdictions.[4] The answer is that this paper does not defend the unconditional veto; it defends conditional consent precisely because the veto is the pathology that emerges when no conditional pathway exists. Communities offered only presumed consent will manufacture absolute refusal. The Compact is the alternative to both.
Second: interconnection queues exaggerate actual demand. The 438-gigawatt ERCOT queue must not be read as a forecast of completed projects; developers file duplicative, speculative requests across multiple utilities, and grid operators themselves caution that the figures are inflated.[19][20] Correct — and the correct inference cuts both ways. Inflated queues counsel against panic, but they equally counsel against ratepayer-financed infrastructure sized to paper demand, which is exactly why Batch Zero’s financial-security screens and Rule 4’s minimum-payment obligations exist: they make the speculators, not the households, bear the cost of speculation.[19]
Third: data centers can benefit other ratepayers. The best causal evidence finds that data centers lowered average retail rates from 2015 to 2024 by spreading fixed system costs, and large customers can anchor new generation and contribute major tax revenue.[38][39] This paper accepts the finding without reservation — and notes that its authors condition the result on spare capacity and warn that future supply constraints can reverse it. The finding is not an argument against the Compact; it is an argument for it, because the Compact’s cost-causation machinery is what keeps the beneficial case beneficial as headroom disappears.[40][58]
Fourth: permanent job counts are not the only benefit. A low-employment facility may still generate substantial tax revenue, construction activity, technical ecosystems, and durable infrastructure.[57] Agreed — Section 4 argues exactly this, urging public revenue as an alternative currency of benefit. The demand is not that facilities employ thousands; it is that whatever benefit is claimed be measured, binding, and clawback-secured.
Fifth: overregulation can weaken national competitiveness. Long delays could impede American AI development relative to China and other rivals, and the federal government has declared the build-out a national priority.[12] This is the industry’s strongest card, and the answer must concede its premise while contesting its conclusion. Uncertainty, litigation, ballot bans, and community mistrust are themselves delay — the roughly 98 billion dollars in blocked or delayed projects in a single quarter was not caused by published standards but by their absence.[9] Predictable high standards are the fastest available route through a democracy that has begun paying attention. The pledge’s own signatories implicitly agree: they signed.[11]
The unified answer to all five counterarguments is the sentence on which this paper’s policy argument rests: Permit Populism is not a case for stopping infrastructure; it is a framework for making expansion politically durable.
Conclusion: From Permission to Legitimacy
Three larger arguments emerge from the analysis, and each extends beyond the 2026 cycle that occasioned it.
First: AI infrastructure is becoming a political institution. Data centers now influence energy planning, taxation, land use, water allocation, labor markets, reliability standards, and public investment; their political significance extends far beyond their property boundaries and will outlast any single election. An industry planning 725 billion dollars of annual capital deployment, drawing nearly half of national electricity demand growth, and capable — as July 22, 2026 demonstrated — of perturbing the grid from Washington to Chicago in ten milliseconds of protective logic, is not a land use.[53][52][5] It is a standing feature of the constitutional order of energy and territory, and it will be governed as one.
Second: speed and consent are not mutually exclusive. The paper’s survey of the 2026 policy landscape — GRID’s fast-permits-for-high-standards exchange, Batch Zero’s disciplined queue, Virginia’s tax-and-host compromise, the Compact’s six pillars — supports a conclusion that both the industry’s accelerationists and the movement’s absolutists resist: predictable standards accelerate responsible projects by reducing uncertainty, litigation, and community mistrust, while unpredictable permissiveness produces Measure NDC.[16][19][27][2] Legitimacy is not a tax on speed; in a democracy that has noticed, it is the precondition of speed.
Third: the durable divide is not pro-AI versus anti-AI. That framing, beloved of both boosters and catastrophists, describes almost no actual voter and no governor examined in this paper. The operative divides are four, and they will organize this policy area for a decade: presumed consent versus conditional consent; promotional benefits versus enforceable benefits; socialized costs versus cost responsibility; confidential decisions versus public legitimacy. Every instrument catalogued in this paper — the pledge, the standards, the taxes, the moratoria, the bonds, the ballots — sorts onto one side or the other of those four lines.
Which returns the argument, finally, to its name. This paper is called Permit Populism because the permit is where the abstraction of artificial intelligence becomes a neighbor, and populism is what happens when neighbors discover they were expected not to ask questions. The framework earns its place in the 2026 midterms because it is the only lens that explains the cycle’s otherwise incoherent facts: a Tea Party organizer and an environmental-justice coalition marching on the same Saturday; a Texas Republican and a Virginia Democrat converging on cost-causation; an 88 percent ballot margin in a Los Angeles suburb against the signature industry of California’s economy; and a White House proclamation adopting, nearly verbatim, the demands first raised at county zoning hearings.[10][21][26][2][12] Candidates who dismiss the movement as ignorance will lose to it; candidates who pander to it with unconditional vetoes will be abandoned by it when the costs arrive; candidates who master its five questions — who pays, who powers, who benefits, who decides, who can stop it — will find in it something rarer than a wedge issue: a governing mandate.
The communities that power artificial intelligence will not remain passive locations on a corporate infrastructure map; they are becoming political principals whose consent must be measured not by silence, but by the enforceable terms under which the future is permitted.
Footnotes / Endnotes:
[1] City of Monterey Park, “Monterey Park City Council Places Data Center Prohibition on June 2 Ballot, Extends Moratorium on Development,” March 5, 2026. https://www.montereypark.ca.gov/m/newsflash/Home/Detail/1306
[2] Fox Business, “California city votes to permanently ban data centers in first-of-its-kind measure,” June 10, 2026. https://www.foxbusiness.com/fox-news-tech/monterey-park-permanent-data-center-ban
[3] Evan Halper, The Washington Post, “Voters in Monterey Park, California, overwhelmingly back data center ban,” June 5, 2026. https://www.washingtonpost.com/business/2026/06/05/voters-monterey-park-california-overwhelmingly-back-data-center-ban/
[4] ABC7 Los Angeles, “Monterey Park voters approve Measure NDC, banning power-hungry data centers within city limits” (Steven Kung), June 4, 2026. https://abc7.com/post/monterey-park-voters-approve-measure-ndc-banning-power-hungry-data-centers-within-city-limits/19229466/
[5] Reuters (via AOL News), “Massive disconnect of power roils largest US electric grid,” July 22, 2026. https://www.aol.com/articles/massive-disconnect-power-roiled-largest-222641000.html
[6] Data Center Knowledge, “NERC Flags AI Data Center Grid Risks in Report” (2026 State of Reliability), July 2026. https://www.datacenterknowledge.com/energy-power-supply/nerc-flags-ai-data-center-grid-risks-in-report
[7] Quartz, “How AI data centers create cascading power outages” (John Moura, NERC), May 2026. https://qz.com/cascading-outages-data-centers-grid-load-shedding-risk-051326
[8] Jasmine Laws, Newsweek, “National day of protest against data centers on July 18 — map, list of cities,” July 9, 2026. https://www.newsweek.com/national-day-of-protest-against-data-centers-on-july-18map-list-of-cities-12150084
[9] Tech Times, “Anti-Data-Center Protests Hit 125 Cities in First National Mobilization” (Data Center Watch; Gallup; Reuters/Ipsos), July 18, 2026. https://www.techtimes.com/articles/320932/20260718/anti-data-center-protests-hit-125-cities-first-national-mobilization.htm
[10] YourNews, “Nationwide Data Center Backlash Draws 142 Protests Across 42 States” (Amy Kremer; HumansFirst statement), July 20, 2026. https://yournews.com/2026/07/20/7118445/nationwide-data-center-backlash-draws-142-protests-across-42-states/
[11] The White House, “Fact Sheet: President Donald J. Trump Advances Energy Affordability with the Ratepayer Protection Pledge,” March 4, 2026. https://www.whitehouse.gov/fact-sheets/2026/03/fact-sheet-president-donald-j-trump-advances-energy-affordability-with-the-ratepayer-protection-pledge/
[12] Proclamation 11014 of March 4, 2026, “Ratepayer Protection Pledge,” Federal Register, Vol. 91, No. 45 (Document 2026-04645), March 9, 2026. https://www.govinfo.gov/content/pkg/FR-2026-03-09/pdf/2026-04645.pdf
[13] POWER Magazine, “Hyperscalers Sign White House Pledge to Fund Data Center Power, Grid Upgrades,” March 6, 2026. https://www.powermag.com/hyperscalers-sign-white-house-pledge-to-fund-data-center-power-grid-upgrades/
[14] David M. Klaus and Mark MacCarthy, Brookings Institution, “The pledge to protect ratepayers from AI data center costs needs enforcement,” July 2026. https://www.brookings.edu/articles/the-pledge-to-protect-ratepayers-from-ai-data-center-costs-needs-enforcement/
[15] Tech Times, “AI Data Centers Are Raising Your Power Bill: White House Expands Pledge Amid Tariff Gap” (Consumer Reports surveys; FERC show-cause orders; Ari Peskoe), July 14, 2026. https://www.techtimes.com/articles/320405/20260714/ai-data-centers-are-raising-your-power-bill-white-house-expands-pledge-amid-tariff-gap.htm
[16] Commonwealth of Pennsylvania, Office of Governor Josh Shapiro, “Gov. Shapiro Releases Full GRID Standards to Protect Pennsylvanians,” May 27, 2026. https://www.pa.gov/governor/newsroom/2026-press-releases/gov-shapiro-releases-full-grid-standards-to-protect-pennsylvania
[17] Daily Energy Insider, “Pennsylvania governor unveils data center standards tied to energy, tax incentives” (Gov. Josh Shapiro), June 4, 2026. https://dailyenergyinsider.com/news/52490-pennsylvania-governor-unveils-data-center-standards-tied-to-energy-tax-incentives/
[18] Pennsylvania Capital-Star, “Shapiro pitches sustainability, transparency requirements for data center developers,” June 2026. https://penncapital-star.com/economy/shapiro-pitches-sustainability-transparency-requirements-for-data-center-developers/
[19] Electric Reliability Council of Texas (ERCOT), “New Batch Connection Process for Large Electricity Users,” Trending Topic explainer, June 18, 2026. https://www.ercot.com/files/docs/2026/06/18/ERCOT-Trending-Topic-New-Batch-Connection-Process-for-Large-Electricity-Users.pdf
[20] Robert Walton, Utility Dive, “Texas, facing 438 GW queue, approves initial large-load interconnection process,” June 22, 2026. https://www.utilitydive.com/news/texas-facing-438-gw-queue-approves-initial-large-load-interconnection-pro/823367/
[21] Texas Energy and Power Newsletter, “Gov. Abbott calls for data centers to lower costs: Texas Grid Roundup #95,” July 2026. https://www.texasenergyandpower.com/p/gov-abbott-calls-for-data-centers
[22] Texarkana Today, “Abbott spent a decade courting data centers. Now he wants to rein them in.” (SB 6; ERCOT CEO Pablo Vegas testimony), July 2026. https://txktoday.com/news/abbott-data-center-reversal-bowie-county/
[23] City of Lebanon, Indiana, “Meta Makes Lebanon Data Center Announcement Official,” February 11, 2026. https://lebanon.in.gov/2026/02/11/meta-makes-lebanon-data-center-announcement-official/
[24] Indiana Capital Chronicle, “Details on long-expected Meta data center campus unveiled” (Gov. Mike Braun; Rachel Peterson, Meta), February 11, 2026. https://indianacapitalchronicle.com/2026/02/11/details-on-long-expected-meta-data-center-campus-unveiled/
[25] Data Centre Magazine, “Meta Breaks Ground on 1GW Indiana Campus with US$10bn Boost,” February 12, 2026. https://datacentremagazine.com/news/meta-breaks-ground-on-1gw-indiana-campus-with-us-10bn-boost
[26] Office of the Governor of Virginia, press release on Virginia’s first-of-its-kind data center energy consumption tax (Gov. Abigail Spanberger, POLITICO Energy interview), July 6, 2026. https://www.governor.virginia.gov/newsroom/news-releases/2026/july-releases/name-1120725-en.html
[27] Data Center Knowledge, “Virginia Approves First-Ever Data Center Power Tax,” June 2026. https://www.datacenterknowledge.com/regulations/virginia-approves-first-ever-data-center-power-tax
[28] Kiplinger, “Virginia Approves First-of-Its-Kind Data Center Power Consumption Tax,” July 2026. https://www.kiplinger.com/taxes/virginia-approves-first-data-center-power-tax
[29] BDO USA, “Virginia Enacts Unprecedented Electricity Consumption Tax on Data Centers” (H.B. 30; Virginia Department of Taxation estimate), July 2026. https://www.bdo.com/insights/tax/virginia-enacts-unprecedented-electricity-consumption-tax-on-data-centers
[30] The American Prospect, “Data Centers Win-Lose in Virginia” (State Corporation Commission rate class), July 6, 2026. https://prospect.org/2026/07/06/data-centers-win-lose-in-virginia/
[31] Circle of Blue, Great Lakes News Collaborative, “Nuclear Power Plant Restart: A New Era Begins” (Gov. Gretchen Whitmer statement), May 13, 2026. https://www.circleofblue.org/2026/water-energy/a-nuclear-shift-buoyed-by-billions-and-the-waters-of-the-great-lakes/
[32] FOX 17 West Michigan, “Palisades Nuclear Plant another step closer to restarting,” March 30, 2026. https://www.fox17online.com/news/local-news/palisades-nuclear-plant-another-step-closer-to-restarting
[33] David Iversen, KTAR News, “Katie Hobbs: Data center freeze fills federal gaps” (Gov. Katie Hobbs press conference), July 7, 2026. https://ktar.com/arizona-politics-news/katie-hobbs-data-57m-saved/5886748/
[34] Arizona Capitol Times, “Data centers dominated 2026 session — lawmakers answered with 3-year tax incentive pause,” July 9, 2026. https://azcapitoltimes.com/news/2026/07/09/data-centers-dominated-2026-session-lawmakers-answered-with-3-year-tax-incentive-pause/
[35] MultiState, “State Data Center Policy: Governors Restrict Tax Exemptions” (Arizona HB 4168/SB 1861; Texas directives), June 22, 2026. https://www.multistate.us/insider/2026/6/22/state-data-center-policy-shifts-as-governors-impose-new-restrictions
[36] Office of Governor Gavin Newsom, “Governor Newsom announces a first-of-its-kind partnership, providing Anthropic tools to state agencies and improving services for Californians,” June 29, 2026. https://www.gov.ca.gov/2026/06/29/governor-newsom-announces-a-first-of-its-kind-partnership-providing-anthropic-tools-to-state-agencies-and-improving-services-for-californians/
[37] CBS Sacramento, “California signs deal to bring Claude AI tools to government workers” (Gov. Gavin Newsom statement), June 2026. https://www.cbsnews.com/sacramento/news/california-claude-deal-gavin-newsom/
[38] Asa Watten, Geoffrey Blanford, and John Bistline, “Have Data Centers Raised Your Electric Bill? Causal Evidence from the United States,” working paper, arXiv:2606.19777, June 2026. https://arxiv.org/abs/2606.19777
[39] Marketplace, “Data centers lowered electric bills in some places — for now” (Asa Watten, EPRI; Ryan Hledik, The Brattle Group), July 10, 2026. https://www.marketplace.org/story/2026/07/10/data-centers-lowered-electric-bills-in-some-places-for-now
[40] Energy and Environmental Economics (E3), “Understanding the Drivers of Rising Electricity Rates and the Role of Data Centers,” whitepaper, May 2026. https://www.ethree.com/electricity-rate-drivers-data-center-role-2026/
[41] Eliza Martin and Ari Peskoe, Harvard Electricity Law Initiative, “Extracting Profits from the Public: How Utility Ratepayers Are Paying for Big Tech’s Power” (coverage: Utility Dive), March 2025. https://www.utilitydive.com/news/utilities-subsidize-data-center-growth-ratepayer-cost-shif-harvard-peskoe/742001/
[42] 12News Phoenix, “Secret data center contracts prevent accountability, says Harvard researcher” (Ari Peskoe), April 2025. https://www.12news.com/article/news/local/arizona/who-pays-more-energy-arizona-homeowners-saw-cost-increase-major-companies-saw-prices-decrease/75-4f7c61b9-d447-4ac9-a5e1-7735b59a85ac
[43] Harvard Magazine, “How AI Could Be Raising Your Energy Bill” (Eliza Martin; Ari Peskoe), July–August 2025. https://www.harvardmagazine.com/2025/07/harvard-ai-increasing-energy-costs
[44] Harvard Salata Institute Climate Brief, “The data center boom is colliding with the grid’s hardest problems” (interview with Ari Peskoe), March 17, 2026. https://salatainstitute.harvard.edu/data-centers-ai-artificial-intelligence-grid-permitting-transmission-electricity-energy
[45] Brookings Institution, “How rising electric rates could affect the 2026 midterms,” April 16, 2026. https://www.brookings.edu/articles/how-rising-electric-rates-could-affect-the-2026-midterms/
[46] Marc Levy and Jesse Bedayn, Associated Press (via U.S. News & World Report), “Voters’ Anger at High Electricity Bills and Data Centers Looms Over 2026 Midterms” (Prof. Dan Cassino, Fairleigh Dickinson University), November 8, 2025. https://www.usnews.com/news/us/articles/2025-11-08/voters-anger-over-high-electricity-bills-and-data-centers-loom-over-2026-midterms
[47] USA Today (via AOL News), “Surging electricity rates put data centers on 2026 ballot. Here’s why.” (Jefferies analyst report), 2026. https://www.aol.com/articles/surging-electricity-rates-put-data-100820887.html
[48] CNBC, “AI data center backlash threatens Pennsylvania GOP incumbents in 2026 election” (EIA rate data; resident interviews), April 24, 2026. https://www.cnbc.com/2026/04/24/ai-data-centers-pennsylvania-republicans-2026-election.html
[49] Capacity Media, “What is the Ratepayer Protection Pledge, and why does it matter for data centres?” (Marquette University Law School poll), July 2026. https://capacityglobal.com/news/what-is-the-ratepayer-protection-pledge/
[50] International Energy Agency, “Key Questions on Energy and AI,” Executive Summary, 2026. https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary
[51] W.Media, “Data center power demand will double by 2030: IEA” (Fatih Birol, IEA Executive Director), April 20, 2026. https://w.media/data-center-power-demand-will-double-by-2030-iea/
[52] International Energy Agency, “AI is set to drive surging electricity demand from data centres while offering the potential to transform how the energy sector works” (Energy and AI special report), 2025. https://www.iea.org/news/ai-is-set-to-drive-surging-electricity-demand-from-data-centres-while-offering-the-potential-to-transform-how-the-energy-sector-works
[53] Yahoo Finance, “Hyperscalers Hit $700 Billion in 2026 AI Spending Plans” (Q1 2026 earnings: Amazon, Alphabet, Microsoft, Meta), May 1, 2026. https://finance.yahoo.com/sectors/technology/articles/hyperscalers-hit-700-billion-2026-111243744.html
[54] Quartz (via Yahoo Finance), “Meta, Microsoft, Amazon, and Alphabet are about to spend a shocking amount of money to dominate the AI era” (Goldman Sachs Research estimates), June 3, 2026. https://finance.yahoo.com/sectors/technology/article/meta-microsoft-amazon-and-alphabet-are-about-to-spend-a-shocking-amount-of-money-to-dominate-the-ai-era-115359575.html
[55] ConstructConnect, “White House, Tech Giants Sign Pledge to Shield Ratepayers from AI Data Center Power Costs” (March 2026 Data Center Report), March 10, 2026. https://news.constructconnect.com/white-house-tech-giants-sign-pledge-to-shield-ratepayers-from-ai-data-center-power-costs
[56] MLQ News, “White House Plans Expanded Ratepayer Pledge Bringing Utilities Into Data Center Cost Framework” (Brookings 15–40% estimate; 300+ state bills), July 2026. https://mlq.ai/news/white-house-plans-expanded-ratepayer-pledge-bringing-utilities-into-data-center-cost-framework/
[57] FOX 5 Atlanta, “Data center protest held at Georgia Capitol as part of nationwide demonstration” (Georgia Tech research on host-county effects), July 18, 2026. https://www.fox5atlanta.com/news/data-center-protest-held-georgia-capitol-part-nationwide-demonstration
[58] Federal Reserve Bank of Dallas, Working Paper 2606, “The Effect of Data Centers on Wholesale Electricity Markets,” 2026. https://www.dallasfed.org/~/media/documents/research/papers/2026/wp2606.pdf
[59] Utility Dive, “Sudden data center load losses prompt NERC alert, recommendations” (Level 3 Essential Actions Alert, May 4, 2026), April 21, 2026. https://www.utilitydive.com/news/data-center-load-disruptions-nerc-alert-recommendations/818036/
