The analysis exactly as it was written, reviewed and stored. The tabs before this one re-cut this same text into sections; the writer refuses to publish a re-cut that does not reproduce it character for character, so nothing here is added, dropped or reworded.
WHAT IS AT STAKE
Data centres became a policy issue because three measurable quantities grew large enough to reach state budgets, utility resource plans, and federal grid dockets: electricity, water, and forgone tax revenue. The most detailed federally sponsored estimate, the Lawrence Berkeley National Laboratory report to Congress of December 2024, finds U.S. data-centre electricity use reached 176 TWh in 2023, about 4.4 percent of national consumption, and projects a 2028 range of roughly 325 to 580 TWh. Those are the established facts of this issue, set out elsewhere on this page; this analysis does not re-argue them, it asks the legal question that sits on top of them.
That question is not whether a data centre violates anyone's constitutional rights. It is a question of structure: who has legal authority to decide how these facilities connect to the grid, who lawfully bears the cost of the transmission and generation they require, and whether a state may buy their construction with tax exemptions that other taxpayers cannot review. This is a federalism and public-utility-law problem, and the honest answer is that most of the binding law lives in the Federal Power Act's division between federal and state regulators, and in the rate cases and tax-expenditure evaluations that division produces, rather than in the Constitution's rights guarantees. This page keeps three things separate: what the law establishes, what is genuinely contested, and where StumpWatch believes the law should go.
HOW AUTHORITY OVER THE GRID IS DIVIDED
The Federal Power Act draws the line. Title 16 of the United States Code, Section 824(b)(1), provides that the Act applies to the transmission of electric energy in interstate commerce and to the sale of electric energy at wholesale in interstate commerce, but shall not apply to any other sale of electric energy, reserving retail sales, and the local distribution that serves them, to the states. In FERC v. Electric Power Supply Association, decided in 2016, the Court read that division directly. Justice Kagan, writing for a six-to-two majority, described the Federal Energy Regulatory Commission's authority over wholesale rates and over the rules and practices affecting those rates, and the parallel command that the Commission may not regulate either within-state wholesale sales or, more pertinent here, retail sales of electricity. That is a holding of the Court: federal jurisdiction over the sale of power stops at the wholesale line, and what happens on the retail side belongs to state public utility commissions.
The majority also fixed the reach of the word affecting. To keep the statute from swallowing everything a wholesale rule touches, the Court adopted the lower court's limiting construction, under which FERC's affecting jurisdiction extends only to rules and practices that directly affect wholesale rates. A federal regulation is not converted into forbidden retail regulation merely because, like every wholesale rule, it has downstream consequences for what retail customers ultimately pay. The line is drawn at what the regulation targets, not at what it incidentally moves.
WHAT FERC CONTROLS, AND WHERE ITS POWER STOPS
The wholesale line cuts both ways, and the Court has enforced it against the states as firmly as it read it for FERC. In Hughes v. Talen Energy Marketing, decided the same term, the Court, in an opinion by Justice Ginsburg, held that a Maryland program guaranteeing a generator a state-set price for capacity it sold into PJM's federal auction was preempted, because, in the Court's words, by adjusting an interstate wholesale rate, Maryland's program invades FERC's regulatory turf. A state may not set, or effectively reset, a wholesale rate FERC controls.
But Hughes is written to be narrow, and its own limiting paragraph is the part that matters most for data-centre policy. The Court said: Our holding is limited: We reject Maryland's program only because it disregards an interstate wholesale rate required by FERC. It went on to reserve, expressly, the permissibility of various other measures States might employ to encourage development of new or clean generation, including tax incentives, land grants, direct subsidies, construction of state-owned generation facilities, or re-regulation of the energy sector. So long as a state does not condition payment on a resource clearing the federal auction, Hughes leaves state authority over siting, subsidy, and taxation intact. Preemption reaches the wholesale rate; it does not reach the incentives states actually use to attract data centres.
WHO BEARS THE COST, AND WHO DECIDES
Because retail rate design is state work, the question of who pays for data-centre-driven grid upgrades is answered, in the first instance, in state commission dockets rather than in court. On the federal side, two things must be kept apart, because the earlier draft of this analysis ran them together and the merger leaned toward a tidier federal answer than the law supplies. The first is solid: the Federal Power Act's standard for the transactions and facilities FERC does control is that all rates and charges be just and reasonable, under Sections 824d and 824e, and a data centre's interconnection to jurisdictional transmission facilities, along with the practices of the transmission providers and regional operators FERC oversees, sits squarely inside that standard. The hook there is the transmission-in-interstate-commerce clause of Section 824(b)(1) and FERC's authority over jurisdictional transmission providers and regional transmission organisations, not the sale-at-wholesale head, because a data centre is a retail consumer of electricity, not a wholesale seller; calling FERC's reach over it wholesale jurisdiction is a category error the earlier draft made.
The second thing is not solid, and honesty requires saying so plainly even though it makes the federal toolbox look smaller. Whether co-located data-centre load, sited behind the meter with a generator, falls within FERC's jurisdiction at all is a live, contested question, not settled law. In its order of December 18, 2025 in Docket EL25-49-000, reported at 193 FERC 61,217, the Commission found PJM's tariff unjust and unreasonable under Section 206 for failing to address co-located load with sufficient clarity, and directed PJM to file tariff amendments. But that order is not final: it launched a paper hearing, with PJM's amendments due February 16, 2026 and briefing running through April 17, 2026, and it is subject to rehearing and judicial review. The rules are being written now, and a reader should not mistake a live Section 206 proceeding for established law. The earlier draft's placement of the co-located-load reach among settled holdings was an overstatement, and it ran in the platform's own favour by making the federal cost tools look more available than the cases support.
On the retail side, state commissions are the forum, and at least one has acted to head off cost-shifting directly. In its Opinion and Order of July 9, 2025 in Case No. 24-508-EL-ATA, entered in the docketing system of the Public Utilities Commission of Ohio and retrievable from dis.puc.state.oh.us, the Commission ordered AEP Ohio to create a data-centre-specific tariff, adopting one stipulation over competing ones to guard other customers against the cost-shifting risks of underused investments made to serve Ohio's growing data-centre industry. The tariff's terms, now read against the Commission's own order rather than secondary reporting, apply to new data-centre load of twenty-five megawatts or more and set a minimum demand charge that will not exceed eighty-five percent of contract capacity for the shortest-commitment customers, tiered down to eighty and seventy-five percent for longer terms and reached through a ramp of up to four years, with the longest term option running about twelve years and an exit fee equal to three years of minimum charges; the eighty-five-percent figure is a ceiling on the shortest term, not a flat universal take. That order is itself on appeal to the Supreme Court of Ohio, where the Ohio Manufacturers' Association challenged it in case 2025-1458, its notice of appeal filed November 3, 2025, and it has not been stayed.
THE TAX EXEMPTIONS AND THE COMMERCE CLAUSE
Two settled holdings frame the tax-incentive question, and they point in opposite directions. On one side, the dormant Commerce Clause forbids a state tax break that discriminates on its face against out-of-state commerce. In New Energy Company of Indiana v. Limbach, a unanimous Court, in an opinion by Justice Scalia, struck down an Ohio tax credit available only for ethanol produced in Ohio or in a State that reciprocated. The source of the rule is the opinion of the Court, not the Reporter's syllabus: the Ohio provision, Justice Scalia wrote, on its face appears to violate the cardinal requirement of nondiscrimination, and the Court reversed on account of this plain discrimination against products of out-of-state manufacture. The same opinion draws the line that matters most for data centres: Direct subsidization of domestic industry does not ordinarily run afoul of that prohibition; discriminatory taxation of out-of-state manufacturers does. The clause bars the discriminatory tax, not the subsidy, and the safe harbour is facial neutrality rather than the form the help takes. The most recent Supreme Court dormant-Commerce-Clause decision, National Pork Producers Council v. Ross, 598 U.S. 356 (2023), rejected, as the opinion of the Court, a free-standing extraterritoriality principle a challenger might otherwise invoke — and in a plurality portion that commanded no majority cast doubt on Pike balancing as well — while leaving intact this anti-discrimination rule, which the Court placed at the very core of the doctrine; the earlier draft left that decision unnamed, and naming it cuts toward caution, because it confirms that facial discrimination is the avenue that reliably triggers the clause.
On the other side stands the obstacle that makes the first hard to use, though it is narrower than it first appears. In DaimlerChrysler Corporation v. Cuno, the Court, in an opinion of the Court by Chief Justice Roberts, held that state taxpayers have no standing under Article III to challenge state tax or spending decisions simply by virtue of their status as taxpayers, and it vacated a lower-court ruling that had struck an Ohio investment-tax credit under the Commerce Clause, finding that the lower courts erred by considering their claims against it on the merits. The merits question the court of appeals had answered was left open, because the plaintiffs could not get through the courthouse door as taxpayers. But Cuno bars the taxpayer route, not the merits themselves: its holding is limited to standing asserted by virtue of taxpayer status. A plaintiff with a concrete, particularised injury of another kind faces no Cuno bar, and New Energy is the proof of it, because the plaintiff there was an out-of-state competitor who had standing and won. These holdings are read here from the opinions of the Court and verified against the opinion text, not from the Reporter's syllabus the earlier draft quoted.
WHAT IS GENUINELY CONTESTED
Whether the dormant Commerce Clause actually reaches the incentives states use to attract data centres is unsettled, and probably narrower than the argument around it suggests. The exemptions at issue, typically sales-and-use-tax relief on servers and equipment for any qualifying facility built in the state, are not written like the ethanol credit struck down in New Energy; they do not, on their face, tax out-of-state goods differently from in-state goods, and New Energy's own distinction between forbidden discriminatory taxation and ordinarily permissible subsidy places most of them on the permissible side. A challenger would also face a standing hurdle, but only on one route: Cuno closes the door to a plaintiff suing as a taxpayer, while a competitor able to show a concrete competitive injury may sue and reach the merits. The result is that a policy several state audits find fiscally negative may, at the same time, be largely beyond the reach of an ordinary taxpayer's constitutional challenge, though not immune from a competitor's. That is a genuine tension, not a settled rule, and this analysis does not claim that any particular data-centre exemption is unconstitutional. The strongest form of the counter-argument is simply that current law already permits, and does not compel, the fixes this page goes on to urge.
Who ultimately bears data-centre grid costs is contested, but the primary records are more precisely described as answering two different questions than as flatly contradicting each other. A study commissioned by Virginia's Joint Legislative Audit and Review Commission concluded that current rate structures make data centres pay their full cost of service, and the same body projected that a typical residential customer could nonetheless see generation and transmission costs rise by an estimated fourteen to thirty-seven dollars a month, in real terms, by 2040. Those findings are reconcilable: a customer class can pay its own allocated cost of service and still drive system-wide capacity and transmission building that raises everyone's bills. The genuine disagreement is not internal to the Virginia study but between it, with its full-cost-of-service finding, and the Ohio commission, which acted on the fear that data-centre investment would shift costs onto other customers and imposed minimum charges to prevent it. Both are primary records; the disagreement is the substance of what is contested, and it is fought out in rate cases rather than in constitutional litigation.
A narrower federal question is open in the same space, and it cuts against this page's own preferred answer as much as for it. The question is whether and how a very large load co-located behind the meter with a generator can take service without paying its share of the transmission system it relies on, and the Commission's December 18, 2025 finding in Docket EL25-49-000 that PJM's tariff was inadequate on exactly this point means the rules are being written now. But there is a prior question the federalism cases pose, and it runs the other way, because FERC may lack jurisdiction over the arrangement at all. The Federal Power Act applies, in the words of Section 824(b)(1), to the transmission of electric energy in interstate commerce and to the sale of electric energy at wholesale in interstate commerce, but shall not apply to any other sale, and the Supreme Court in FERC v. Electric Power Supply Association confined FERC's reach over practices affecting wholesale rates to those that directly affect the wholesale rate. A genuinely islanded load, taking power behind the meter from a co-located generator without a wholesale sale and without using the interstate transmission grid, may fall outside both heads, which would throw the cost question to the states rather than to FERC. The co-located configuration is designed precisely to avoid a wholesale sale, so the strongest objection to the federal half of the fixes below is that the federal tool may not reach the very facilities at issue.
WHERE THE LAW LEAVES GAPS
Several of the harms driving public concern have no comfortable federal legal handle at all, and saying so is more honest than manufacturing one. There is no federal requirement that a data centre report its facility-level electricity or water use; the Lawrence Berkeley figures are modelled estimates, and the report says so. Water withdrawal and consumption are governed, where they are governed, by state water law and local permitting under the states' police power; this analysis found no federal statute or holding that preempts or federalises that question, and none should be implied to exist. Local land use, meaning zoning, noise, setbacks, and the siting of the facilities themselves, is likewise ordinary municipal and state authority, and the primary legal record there is a patchwork of local decisions rather than a doctrine.
The structural gap is the one that ties the others together. The measures with real bite, a discriminatory-incentive challenge, a demand for facility-level disclosure, a fight over who pays, each run into a standing wall, an absent reporting duty, or a rate docket that most affected people never see. The Constitution does very little work here. What binding law exists is statutory and regulatory, and it is enforced, if at all, by legislatures, state commissions, and the occasional plaintiff with a concrete competitive injury.
WHERE STUMPWATCH BELIEVES THE LAW SHOULD GO
This section is StumpWatch's position, and it is a position rather than a description of existing law. We hold it whether or not a court would, and we mark it as a position so no reader mistakes it for a holding.
Because Cuno closes the ordinary taxpayer's route to federal court, and because most data-centre exemptions are not the facially discriminatory kind a competitor could readily strike, discriminatory or fiscally negative incentives are hard to reach through constitutional litigation even though they are not immune from it. That difficulty is a reason to locate accountability in the forum that grants the incentive in the first place. Georgia's evaluation is the model of what that looks like: an econometric analysis that estimated roughly seventy percent of data-centre construction activity would have occurred without the exemption, and a bottom-line finding that the incentive's fiscal impact on the state was negative. An incentive that cannot survive that kind of published, independent test should not be renewed, and the test should gate the incentive rather than arrive years after the money is spent.
On cost, the Federal Power Act's just-and-reasonable standard and the states' retained authority over retail rate design already supply tools; what is missing is the will to use cost-causation rate design before the infrastructure is built, not after. Ohio's minimum-take tariff and FERC's co-located-load proceeding are early examples of regulators making the load that causes a cost carry that cost. StumpWatch's position is that this should be the default: the party whose demand drives a transmission build should bear its cost, and a residential ratepayer should not subsidise a hyperscaler's grid through an averaged rate. One honest caveat belongs on the federal half of this: for a genuinely islanded co-located load that makes no wholesale sale and uses no interstate transmission, the FERC just-and-reasonable tool may simply be unavailable, and the cost question then falls to the states. The position does not depend on FERC reaching every configuration; where FERC cannot, the states' retained retail authority is the forum.
One disclosure is owed to the reader, because it cuts against this page. Every correction this analysis made to its own earlier draft ran in the same direction: the earlier draft had placed FERC's reach over co-located load among settled holdings, had labelled a retail-consumer interconnection dispute as wholesale jurisdiction, had omitted the objection that FERC may lack jurisdiction entirely, and had called incentives effectively unreviewable in federal court. Each of those made the federal cost tools and the accountability gap look more settled, and more favourable to the position argued here, than the law supports. Correcting them makes StumpWatch's case harder, not easier, and we make them anyway, because a position that needs the law overstated to stand is not one worth holding. None of this requires a new constitutional theory. It requires legislatures and commissions to use the authority the Federal Power Act and their own retail jurisdiction already leave them, and to publish the evidence when they do.