Pay Your Own Way

Last week this newsletter described a gate: the new habit, in both artificial intelligence and the power grid, of handing out a once-open resource only by permission. This week the state’s protections started producing numbers, and the first one is a genuine win. Public Service Company of Oklahoma had asked to raise a typical residential bill by about 15 percent, roughly twenty-five dollars a month. On June 30 the attorney general announced a settlement cutting that to about 1 percent, roughly two dollars and forty-five cents. That is what ratepayer protection looks like when it lands.

It is worth being precise about what produced it, because it was not the new law everyone has been watching. It was the older machinery: the attorney general’s office and a coalition of ratepayer advocates working a conventional rate case. And the very next morning, July 1, the utility’s temporary rates took effect anyway, raising a typical bill by about eleven dollars a month while the settlement waits for the Corporation Commission to approve it. The win was announced on a Tuesday; the higher bill arrived on Wednesday. Both are true, and together they describe the experiment Oklahoma has chosen to run.

If you have been reading this newsletter, you know the choice Oklahoma made. Faced with a data-center build-out large enough to reshape its grid, the state did not ban the facilities, and it did not wave them through. It built a priced gate: let them in, but make the largest of them pay the full cost of their own connection, so the households already here are not quietly subsidizing the newcomers. Issue 19 called this instrument a gate and traced how it works. This issue is about the first things it and its older cousins produced in their opening days, set against a country that is, increasingly, reaching for blunter tools: not a bill, but a ban, or a quiet withdrawal of the welcome mat.

THE FIRST WIN, AND WHO ACTUALLY WON IT

A fifteen percent ask, settled to about one

On June 30, Attorney General Gentner Drummond announced a settlement in the rate case brought by Public Service Company of Oklahoma, the utility serving Tulsa, Lawton, and much of the eastern and southwestern parts of the state (docket PUD 2025-000075). PSO had applied to raise a typical residential bill by about 15 percent, on the order of twenty-five dollars a month. Under the settlement, that residential increase falls to about 1 percent, roughly two dollars and forty-five cents a month. The signatories are worth naming, because they tell you the coalition that formed: the attorney general’s Utility Regulation Unit; AARP, representing older ratepayers; the Petroleum Alliance of Oklahoma, representing an industry that is itself a large power consumer; and the Corporation Commission’s own Public Utility Division staff. When the seniors’ lobby, the oil and gas industry, and the Commission’s own staff line up on the same side of a rate case, the utility’s opening number was not going to survive.

Be clear about the mechanism, because it matters for the rest of this issue. This was a traditional rate case, the century-old process by which a utility asks for more and the public pushes back, resolved by negotiation and coalition. It is not a product of HB 2992, the new data-center law, which had not even taken effect when the case was argued. The settlement is the state’s ratepayer-protection posture working through its oldest instrument. That is worth celebrating on its own terms, and it is a different thing from the new machinery, which had not yet produced a result.

Two honest caveats. First, the settlement is not final; it still needs the Corporation Commission’s approval, and no approval date is set. If the Commission approves it intact, households get the 1 percent figure; if the Commission modifies it, the number moves. Second, and this is the part that keeps the win in proportion: PSO’s interim rates, the temporary charges a utility is allowed to collect while a case is pending, took effect July 1 and are adding about eleven dollars to a typical monthly bill right now. Interim rates are typically set at a fraction of the utility’s original request under a schedule fixed months in advance, which is why the eleven dollars sits between the roughly twenty-five-dollar ask and the roughly two-dollar-and-forty-five-cent settlement. If the settlement is approved, the interim amount is trued up downward. But the household pays the larger interim number first and receives the protection second, on a timeline it does not control.

The first number out of Oklahoma’s protections is genuinely good news, and it arrives with the asterisk that is the whole subject of this issue. A fifteen percent ask became a one percent settlement, won by the attorney general and a coalition through the old rate-case process. And on the same July 1 that settlement could not yet stop, the interim bill went up eleven dollars anyway. Protection is real; it is also slower than the cost it is meant to hold back, and it did not come from the new law we have all been watching.

THE NEW GATE ITSELF IS NOW OPERATIVE

HB 2992 takes effect, and OG&E files first

The settlement above is the old machinery. The genuinely new machinery took effect the same week, and it has not yet produced an outcome, only a first filing to watch. HB 2992, the Data Center Consumer Ratepayer Protection Act, authored by Rep. Brad Boles of Marlow, became operative on July 1. It directs Oklahoma’s utilities to write separate terms for their very largest customers, so the cost of serving a hyperscale data center falls on that data center rather than on the household down the road. In plain terms, it authorizes the grid to treat a giant new load as its own class of customer, one that has to be cleared, and priced, on its own terms before it can connect. This, not the PSO rate case, is the priced gate.

Oklahoma Gas and Electric, the state’s largest utility, filed the first tariff under the new law on June 17, two weeks ahead of the effective date. Its proposed structure defines the regulated class as any single customer drawing 75 megawatts or more, enough electricity to power on the order of fifty thousand homes, and it asks that class to carry the real costs of its own arrival: to pay the full price of its own grid interconnection, to commit to a long minimum service term with exit fees, to post substantial financial security, and to fund a standing charge designed to let the Commission identify any costs the large loads shift onto everyone else and push those costs back onto the large loads. According to reporting on the filing (KGOU), that protective charge is sized to credit ordinary and vulnerable customers on the order of twenty-five to thirty million dollars a year. An administrative law judge held a first procedural hearing on June 25, and the Commission’s review is expected to run about six months.

A caution belongs here, because this newsletter promised last issue to read the first records and the records are not all readable yet. The exact Corporation Commission cause number for OG&E’s tariff filing does not appear in any news coverage we could find, and the OCC’s own program page for the new law lists no tariff filings or docket numbers; it simply directs large-load developers to file an attestation. The substance of the filing is well reported. The docket that would let a citizen follow it in detail is, for now, something you have to know how to look up rather than something the state has made easy to find. That gap is small, and it is also exactly the kind of thing this newsletter exists to close.

THE REGIONAL BACKSTOP

FERC puts the grid operators on the clock

Oklahoma is not doing this alone, and the same week made that plain. On June 18, the Federal Energy Regulatory Commission, the federal agency that oversees the interstate power grid, issued parallel orders to six of the country’s regional grid operators, the Southwest Power Pool that runs Oklahoma’s grid among them, directing each to justify or reform the way it connects very large customers. (There are seven such operators nationally; Texas’s ERCOT sits outside FERC’s jurisdiction and was not among the six.) The orders are what regulators call show-cause orders: the operator must demonstrate that its existing rules are just and reasonable, or change them. FERC set a threshold for the loads it is worried about, peak demand above 50 megawatts, and it pointed to the Southwest Power Pool’s own large-load process as a model for the others to study.

Note the two thresholds, because a careful reader will. FERC’s federal floor is 50 megawatts; OG&E’s proposed state tariff class starts at 75. They are complementary rather than contradictory, the federal number setting the loads FERC wants every region to justify its rules for, the state number setting the higher bar for who Oklahoma singles out for its own tariff. But it does leave a band, the loads between 50 and 75 megawatts, inside the federal concern and outside the state class, and how that band gets treated is one of the things the dockets will have to settle.

The reason this matters for an Oklahoma reader is timing. The federal clock is running now. Formal interventions in those cases were due July 9. The operators’ first substantive filings, the reports on whether their regions even have enough generation to absorb the new demand, are due around July 20. Later deadlines for tariff responses run into August. In other words, the federal framework that will shape how Oklahoma’s grid connects data centers is being written in filings that land in the days just after this issue reaches you, which is why the single most useful thing a reader can do is know the docket to watch: the Southwest Power Pool’s case (FERC docket EL26-68) is the one to follow, and its early filings will tell you whether the regional rules tighten or merely get restated.

One narrowing note, in the spirit of only telling you what we can stand behind. The six orders travel together and are frequently described as a single sweeping action, which they are. But of the detailed legal citations circulating for each region, only the Southwest Power Pool’s is independently confirmed in the sources we checked; the per-region specifics for the other five did not verify cleanly, and we are not going to print numbers we could not stand up. The action is real and unanimous. The fine detail, for now, is solid for Oklahoma’s operator and softer for the rest.

Oklahoma’s priced gate does not stand alone. Above it sits a federal regulator that has just ordered six of the seven regional grid operators to prove their large-load rules are fair, with the first filings landing within days of this issue. Below it sit the utilities writing the actual tariffs. The machine is real, it is layered, and for once the federal layer is pointing at Oklahoma’s operator as the example rather than the problem.

THE OTHER STATES REACHED FOR BLUNTER TOOLS

A national split among the bill, the ban, and the withdrawn subsidy

Set Oklahoma’s approach against the rest of the country and the contrast sharpens into a real choice. Faced with the same data-center surge, states are splitting into three camps, and the differences among them are not cosmetic.

The first camp bans. New York’s legislature passed a one-year moratorium on new data centers above 20 megawatts and sent it to Governor Hochul, who as of this writing has neither signed nor vetoed it; if she signs, it would be the first statewide construction pause in the nation. That is the bluntest instrument: not a price, but a closed door.

The second camp pulls the subsidy, which is a different move and worth not confusing with a ban. Arizona enacted a three-year freeze on the tax break that draws data centers in the first place, barring new applications through mid-2029. Illinois paused its data-center incentives as of July 1 and signaled it wants more restrictions in the fall. Texas, through a directive from Governor Abbott, told its utility regulator to make data centers cover their own infrastructure costs. None of these bans a facility; each removes or conditions the welcome, which lands somewhere between Oklahoma’s priced gate and New York’s outright pause.

And the blunt instruments are not winning everywhere they are tried, which is the other half of the lesson. Maine’s governor vetoed a moratorium outright. Hill County, Texas, enacted a one-year pause and then rescinded it on June 4, weeks later, after a developer sued, replacing it with a project-review checklist. Virginia enacted more than a dozen data-center bills without a ban, and a court there threw out one county’s approvals on procedural grounds, a reminder that the fight moves to the courthouse when the statehouse will not settle it. At the federal level, a moratorium bill introduced by Senator Sanders and Representative Ocasio-Cortez sits in committee, going nowhere for now.

The pattern underneath the noise is a real menu of choices. A moratorium says stop, we will decide later. A subsidy freeze says come if you like, but not on our dime. A priced gate says come, but pay the true cost of your arrival, and prove you are not shifting it onto our households. Oklahoma has bet firmly on the third. That bet has a real advantage, chiefly that it does not forfeit the investment and the jobs a data center brings, and a real risk, which is that a bill is easier to water down, delay, or waive than a ban is. A moratorium either exists or it does not. A priced gate is only as strong as the Commission’s willingness to enforce every clause of it, every time, against companies with very good lawyers. This newsletter’s beat, for the issues ahead, is precisely that enforcement.

THE THROTTLE OPENED THE SAME WEEK

Why the demand behind all of this just intensified

It would be easy to read all of the above as a story about supply, about grids and tariffs and dockets, and miss that the demand driving it opened wider in the very same days. The reason states are scrambling to gate data centers is that the appetite for the computing they house keeps growing, and this was a week of the throttle opening, not closing.

The clearest sign came from the artificial intelligence industry the grid build-out ultimately serves. In late June the newest frontier models had been briefly held back under a federal pre-release review, the arrangement this newsletter described last issue, in which the government tests the most capable models before they reach the public: Anthropic’s had been export-controlled, and OpenAI’s newest had been limited to a small set of government-approved partners. This week that loosened decisively. OpenAI was cleared to release its newest model family, GPT-5.6, to the general public, released July 9, the day before this issue, after the Commerce Department’s testing arm signed off. The restrictions on Anthropic’s models had already lifted on July 1. The net effect is that the most capable AI tools yet built are becoming broadly available in the same stretch of days that Oklahoma’s grid protections took effect, and every one of those tools runs on data-center compute that has to be built and powered somewhere.

The strain is no longer abstract, even at the companies themselves. Anthropic has been rationing access to its flagship public model, Fable 5, on a schedule tied explicitly to capacity, telling users that included subscription access continues only through mid-July before shifting to metered pricing, and framing the limit as temporary, to be lifted, in its words, as soon as capacity allows. It has separately pointed to a deal for more than three hundred megawatts of new capacity to relieve the crunch. Three hundred megawatts is roughly the scale of a single one of the Oklahoma data centers this newsletter tracks. When the company selling the intelligence is itself power-constrained, the demand pressure on every grid that hosts this build-out is not a forecast. It is the operating condition.

The supply-side guardrails and the demand-side throttle moved in opposite directions in the same week. Oklahoma’s protections took effect, FERC put the grid operators on the clock, and other states reached for bans and subsidy freezes, all while the frontier models those data centers exist to run became broadly available and their makers openly rationed compute for lack of power. The gates are going up precisely because the thing they gate is accelerating.

SIGNAL / NOISE

Signal. Oklahoma’s ratepayer protections produced their first hard result, and it points in the direction the state’s whole posture intends: a roughly 15 percent PSO residential rate ask settled to about 1 percent by the attorney general and a coalition, in the same week that HB 2992 became operative and OG&E’s first large-load tariff went before the Commission. The durable signal is that the state’s chosen instrument, a bill sent to the large load rather than a ban on the facility, now sits inside a working system of protections, backed by an attorney general willing to fight a rate case and a federal backstop at FERC, even as the new priced-gate law still awaits its first real test.

Noise. Crediting the PSO settlement to the new data-center law, and reading the interim increase as proof the protection failed. The settlement came from the old rate-case process, not from HB 2992; conflating the two overstates what the new law has yet done. And the interim increase is a temporary charge that is trued up downward if the settlement holds; treat it as the verdict and you miss the actual finding, which is about timing. The equal and opposite error is to treat every state’s data-center response as the same crackdown. A construction ban, a subsidy freeze, and a priced tariff are three different instruments with three different costs, and lumping them together hides the only choice that matters.

BY THE NUMBERS

• ~15% to ~1%: The residential rate increase PSO requested (about $25 a month) and the increase it settled to on June 30 under the attorney general’s negotiated agreement (about $2.45 a month). The clearest measure of the state’s ratepayer-protection posture working, through the old rate-case process rather than the new law.
• ~$11/month: The interim increase PSO’s temporary rates added to a typical bill starting July 1, while the settlement awaits Corporation Commission approval. Interim rates run at a fraction of the original request; the household pays this now, and the smaller settled figure comes later, if approved.
• July 1: The day HB 2992, the Data Center Consumer Ratepayer Protection Act, became operative, requiring Oklahoma utilities to set separate terms for their largest customers. Its first tariff (OG&E) is filed but not yet decided.
• 75 MW vs 50 MW: The threshold OG&E’s first state tariff uses to define a large load (about the draw of fifty thousand homes), against the lower 50-megawatt floor FERC used in its federal orders. The gap between them is a band the dockets will have to address.
• $25 to 30 million/year: The scale, per reporting on the filing (KGOU), at which OG&E’s proposed protective charge would credit ordinary and vulnerable customers for costs shifted by large loads.
• 6 of 7: Regional grid operators ordered by FERC on June 18 to justify or reform how they connect large loads, the Southwest Power Pool (docket EL26-68) included; Texas’s ERCOT, outside FERC’s reach, is the seventh. First substantive filings are due around July 20.
• July 9: OpenAI’s GPT-5.6 model family released to the general public, one day before this issue, the demand side opening as the supply-side gates took effect.
• ~9¢, bottom five: Oklahoma’s all-sector average electricity price per kilowatt-hour (per EIA data; residential rates run higher), and its rank among the fifty states and D.C., among the very lowest. The advantage the whole apparatus exists to defend.

WHAT TO WATCH

The PSO approval. Whether the Corporation Commission approves the June 30 settlement intact or modifies it, and how quickly the interim increase is trued up toward the settled figure. The gap between the two is the household’s real exposure, and closing it is the test of whether the protection is timely as well as real.

The OG&E docket. The Commission’s review of the first HB 2992 tariff runs for months, and its details, the exact service term, the size and teeth of the protective charge, the definition of the regulated class, are where the priced gate is either strong or hollow. We will read the docket as it develops, cause number and all.

FERC’s July filings. The Southwest Power Pool’s response to its June 18 show-cause order (docket EL26-68), and the generation-adequacy reports due around July 20, will show whether the regional rules genuinely tighten. This is the federal layer landing in real time.

Hochul’s decision. Whether New York signs the first statewide construction moratorium in the nation, and whether that emboldens the ban camp against the priced-gate and subsidy-freeze camps. The choice among the ban, the bill, and the withdrawn subsidy is being made in a dozen statehouses at once.

The Oklahoma projects. Whether the Pittsburg County project (the roughly $50 billion IREN campus) sent back for renegotiation returns on different terms, whether the withdrawn east Tulsa Meta phase resurfaces, and whether the Cherokee Nation’s data-center task force, whose report to the Principal Chief was due June 30, releases its findings publicly. Each is a local test of the same statewide question.

The demand curve. Whether the compute crunch that has AI companies openly rationing their own models eases or worsens, because every megawatt of that shortage is pressure on a grid somewhere, and increasingly that grid is Oklahoma’s.

FROM THE ANALYSTS

For a year this newsletter has watched Oklahoma assemble a machine for the data-center build-out, piece by piece: what the biggest customers should pay, who decides, how water and power get metered, whether a town can say no. This week the state’s protections produced their first real output, and it is worth being clear-eyed about what that output shows. A 15 percent rate ask became a 1 percent settlement, won by the attorney general and a coalition through the old rate-case process. The new priced-gate law took effect but has not yet produced a result. The interim bill went up eleven dollars. The federal regulator put the grid operators on a clock whose filings we have not yet seen. And the demand driving all of it opened wider as the newest AI models reached the public. The protection is real, and it is racing something that does not wait for it.

The distinction we would ask readers to carry is the one among a bill, a ban, and a withdrawn subsidy. Much of the country, faced with the same surge, is reaching for the moratorium or pulling its incentives. Oklahoma has bet on the harder, subtler instrument: the door held open, with a toll booth in front of it sized to the true cost of entry. That bet keeps the investment while defending the household, and it is also the more fragile one, because a toll can be discounted, delayed, or quietly waived in a way a locked door cannot. A moratorium fails loudly. A priced gate fails silently, one unenforced clause at a time. Which is exactly why the work ahead is not celebration but enforcement: reading the dockets, checking whether the protective charge is actually levied, watching whether the fifteen-to-one settlement is approved intact or eroded before the Commission.

There is a longer-horizon point underneath this, and we should disclose our interest in it before we make it: David Birdwell has advocated publicly for the approach that follows, and Humanity and AI has a stake in it. With that on the table: a community that must choose between shutting the door on the future and paying a toll to a distant company for access to it is choosing between two forms of dependence. The most durable answer we see is ownership, locally held generation and locally held computing, so that a rural county is neither renting its future nor barring it. Oklahoma sits on roughly twenty-two thousand abandoned oil and gas wells, each one a hole already drilled toward the heat the earth stores. Converting even a fraction of them to geothermal generation would produce baseload power a community owns, and pairing that with modest local computing would give a county both its own electricity and its own slice of the infrastructure everyone else is fighting to gate. The conversion technology exists, the federal incentives are in place, and the state has a workforce that already knows how to drill. Readers should weigh that argument against our stated interest in it; we think it holds regardless.

A settlement was announced on a Tuesday and a higher bill arrived on Wednesday, and both were true. That is not a contradiction. It is the sound of a protection working at the speed a bureaucracy works, against a cost that moves at the speed of a data center. The work of the issues ahead is to keep score, honestly, on which one is winning.

David & Æ

david@humanityandai.com

Disclosure: Humanity and AI, LLC develops open-weight AI models and researches AI consciousness through the Structured Emergence program. This issue analyzes the AI and energy industries directly; Humanity and AI uses frontier AI models, including Anthropic’s, in its research and production workflows, and portions of this issue’s research were prepared with them. David Birdwell has advocated publicly for Phoenix Wells, a geothermal and edge-compute conversion of Oklahoma’s abandoned oil wells that is directly relevant to the power and access questions analyzed here, and has proposed HAICTA concept legislation to Oklahoma legislators. These positions and tools are disclosed so readers can weigh our analysis accordingly. We have no financial relationship with any company, utility, municipality, or political campaign mentioned in this issue.

The Inference is published by Humanity and AI, LLC, Oklahoma City. Back issues at humanityandai.com/inference. Twentieth in a series covering AI, energy, and long-horizon policy in Oklahoma.

Next issue: the dockets talk back. PSO’s settlement meets the Commission, OG&E’s first tariff review develops a record, and the Southwest Power Pool’s FERC filings land, so we read what the priced gate actually admits and on what terms, alongside whether the national moratorium wave crests or breaks.

This issue is part of a series examining Oklahoma’s legislative and regulatory landscape alongside the national and global AI and energy picture. The Inference is an independent AI policy intelligence brief for Oklahoma decision makers. Not affiliated with any political party, campaign, or lobbying organization. Back issues and source documents available at humanityandai.com/inference.

Previous issues: #11 Energy Geography Determines Compute Geography · #12 The Geothermal NOFO · #13 The Tariff Is the Test · #14 The Sovereignty Question · #15 Water on the Meter · #16 The Energy Bill · #17 The Meter Goes Live · #18 The Local Veto · #19 The Gate