Texas has seen this movie before. In 2022, after China banned crypto mining, the state threw open its doors: ten-year tax abatements, cheap power, a light regulatory touch. ERCOT's interim chief executive said 17 gigawatts of miners had inquired about connecting — "two-and-a-half New York Cities" of demand. The pitch was that miners were ideal customers, big flexible loads that would soak up cheap electricity and curtail when the grid was stressed. For a while it worked. Then the state cooled: a special interconnection process for miners in 2022, a law trimming their incentives in 2023.
The sequel is roughly thirty times bigger, and Texas isn't cooling this time. It's auditing. On August 3, the governor ordered every data center in ERCOT's interconnection queue audited before another one advances, with non-compliant projects to be denied connection. On September 14, he put the state's water board on enforcement duty, with civil and criminal liability for data centers that ignore its surveys. On September 21, he halted all state environmental permits for data centers. On September 24, the attorney general opened an investigation. Responses to the state's information requests were due October 12; the water board reports October 14; the environmental agency updates the governor on October 19; ERCOT's full audit lands December 10.
Texas runs America's only major energy-only power grid: generators are paid solely for electricity actually sold, with no capacity market paying them to keep plants available. Investment is supposed to come from volatility itself — prices spiking to $5,000 a megawatt-hour when reserves run thin, collapsing near zero when the wind blows. That design is genuinely good at one thing: letting cheap resources in fast. Merchant batteries went from a rounding error to 16.5 gigawatts, mostly without subsidies. But the design has no mechanism for telling real demand from fiction, because filing an interconnection request cost nothing. When load growth arrived faster and lumpier than any spot market could coordinate, the fiction piled up until the state had to do the coordinating itself.
"Market-purist" needs a qualifier, though. Winter Storm Uri in February 2021 — more than a third of generation offline, millions without power — already cracked the philosophy open. Since then: weatherization mandates, the offer cap cut from $9,000 to $5,000, a new capacity-like credit mechanism for dispatchable generators, and a $7.2 billion state loan fund to build new gas plants. By the time the data centers arrived, Texas was running a hybrid — market in name, increasingly administered at the edges. The boom didn't encounter a pure system. It encountered a patched one, and added a problem none of the patches addressed.
The theoretical case against the design was always the same, and economists gave it a name: missing money. A plant that only runs during the hundred scarcest hours of the year can't earn enough, even at $5,000 a megawatt-hour, to justify existing — so peaker plants, the ones you most need in a crisis, are the ones the market least wants to build. Texas argued for years that a high enough cap solved this. Uri cut the cap nearly in half, and the argument got harder. The data-center boom added a second problem, and a more practical one. The market was already unreliable at delivering the peaker plants it needed. Now it had to sort 474 gigawatts of maybe with nothing but the spot price — and the spot price can't tell who's real.
If that framing is right, it points to the questions worth asking: what breaks when a market can't answer "who's real?", who pays for a grid planned around fiction, and is Texas the preview of what's coming everywhere else?
Texas offered what data centers wanted: cheap land, cheap power, historically fast interconnection, no capacity-market costs baked into bills, abundant natural gas nearby, and no state income tax. Then the state sweetened it. Qualifying data centers were exempt from the 6.25 percent state sales tax on equipment and electricity — a subsidy the comptroller says cost about $1 billion in foregone revenue in 2025 alone, projected at $1.3 billion next year. Texas paid roughly a billion dollars a year to attract the boom it then halted. The governor's stated priority for the 2027 legislative session is eliminating the exemption entirely. The incentives took a decade to build and about a year to reverse.
ERCOT's large-load interconnection queue stood at 63 gigawatts in December 2024. By November 2025 it was 226 gigawatts, nearly three-quarters of it data centers. By August 2026, the state was considering 474 gigawatts of requests — a figure that sweeps in projects beyond the formal queue — against a record peak demand of about 91 gigawatts. Five times the grid, on paper, nearly all of it data centers. ERCOT's chief executive has said the forecast exceeds expected load growth. Everyone agrees most of it will never be built. The disagreement is over what to do about the pileup.
The political pressure arrived on a different frequency. County hearings filled with residents worried about well water, generator noise, and property values; by one industry estimate, more than $170 billion in AI data-center capacity has been blocked, withdrawn, or stalled by community opposition since the start of 2024. The grid problem and the water problem and the neighbor problem turned out to be the same problem: nobody could say which projects were real, so every project looked like a threat.
What Texas did, step by step, was put a price on sincerity. Senate Bill 6, signed in the summer of 2025, requires large loads to demonstrate site control, fund their own grid upgrades, disclose duplicate interconnection requests filed elsewhere in the state, and accept curtailment during emergencies. The implementing rules added the sharpest tooth: $50,000 per megawatt in financial security before an interconnection study even begins — $5 million for a 100-megawatt project, posted upfront, explicitly to discourage speculative filings. In June 2026, regulators approved Batch Zero, a cluster-study process for loads over 75 megawatts: first-ready, first-served, the Texas translation of the federal reforms.
Then came the gates. The August audit demands five disclosures from every data center 25 megawatts and up: what state incentives it received, how much of its power it brings itself versus drawing from the grid, its water sources and uses, its impact on neighboring communities, and who actually owns it. Fail to comply and the project "must be denied connection to the Texas grid." The September water order added teeth the market never had: ignore the survey and face civil and criminal liability. The September permit halt closed the other door — no environmental permits, no water rights, no wastewater approvals until the audit completes, with no megawatt threshold and no clear exemption even for behind-the-meter generation. So a project now needs both to lift before it can proceed: ERCOT’s energization pause and the state’s permit freeze, each on its own timeline.
What each step has in common: none of them is a price signal. They're all tests — of money, of disclosure, of patience. The state looked at a market producing 474 gigawatts of maybe and concluded that no price adjustment would sort it. So it stopped trying to price the problem and started administering it.
The audit treats bringing your own power as the closest thing to a safe harbor — self-sufficiency is what the state says it wants. But the TCEQ halt may cover the air permits for on-site generation too; the directive names no exemptions, and the agency isn't clarifying until October 19. So a developer doing exactly what the state asks for may find that permit frozen as well. Rationing regimes produce these contradictions. When every gate is administered by a different agency on a different calendar, the rules stop adding up.
An energy-only market produces prices for energy. It produces no price for credible commitment — for the difference between a developer who will build and one who filed a form. Filing was free, so fiction was free, and 474 gigawatts of fiction is not a rounding error. It's a planning input. ERCOT had to plan transmission around demand that didn't exist, order equipment for load that would never arrive, and explain to legislators why the forecast showed five grids. That's the point where the market's answer to "who gets power?" — whoever filed first — stopped being acceptable. The state stepped in with a different answer: whoever can prove they're real.
The winners and losers follow directly. The regime favors the deep-pocketed and the shovel-ready: hyperscalers with funded projects survive any readiness screen, and projects that bring their own power look safer than those that don't. It punishes the speculative and the thinly capitalized — $50,000 a megawatt in security plus self-funded upgrades is existential for a small developer, a rounding error for a hyperscaler. The intended casualty is the phantom filer; the collateral damage is anyone who can't post collateral, a distinction the filter doesn't bother making. There's already jurisdictional arbitrage: much of the Panhandle sits in the Southwest Power Pool, not ERCOT, so the governor's directives don't touch it — the freeze has a map, and developers can read maps. And the filter is working, in the narrow sense: lawyers in Houston report clients quietly withdrawing Texas projects, mostly the complex, permit-heavy ones, while projects with real work already done stay put.
The capital that leaves has to go somewhere. BloombergNEF estimates 49.8 gigawatts of ERCOT data-center load — about a fifth of the national pipeline — faces delay, with up to $15 billion of revenue at risk in the worst scenario, and expects barely a gigawatt of new Texas data-center capacity through early 2027. Analysts point to states with structured large-load tariffs and no moratoriums — Oklahoma, Georgia, Ohio, parts of the Midwest — though those flows are still speculation, not tracked fact. What's not speculation is that Texas isn't alone anymore. New York has imposed a statewide moratorium on hyperscale data centers. Pennsylvania's governor ordered what he called the strictest guardrails in the nation: self-funded infrastructure, local approval before state permits. Florida now makes facilities over 50 megawatts bear their full interconnection cost. And at the federal level, FERC's June show-cause orders on large-load interconnection — the same diagnosis, the same prescription of readiness screens and cost responsibility — are expected to resume in mid-November. Texas is the leading edge of a national pattern, not an outlier.
December 10 brings ERCOT's report. January brings the legislature, with the incentive exemption, SB 6's implementation, and the permit regime all on the agenda — and possibly a longer leash for the whole apparatus if the pause stretches past the November midterms, as some analysts expect. What emerges will likely be the country's first comprehensive large-load regulatory code: part market, part audit, part criminal statute. Other states are watching less to copy Texas than to avoid needing to.
Underneath it all is the asymmetry that explains why everyone's so nervous. Jim Matheson, who runs the national association of rural electric cooperatives, put it plainly this week: "When you build a whole power plant to serve [a] 1,000 megawatt data center and the data center goes away, that's a problem." Power plants are built to run for up to 80 years. Data centers sign ten-to-fifteen-year leases. The physical commitment is lopsided — concrete and steel on one side, a maybe on the other — and every piece of the Texas response is an attempt to make the maybe legible before the concrete gets poured. The $50,000-a-megawatt security, the ownership disclosures, the criminal liability for lying about water: all of it is the state trying to manufacture, by administrative force, the one thing its market never produced. Credible commitment.
The job changed. The old task was pricing energy efficiently; the new one is deciding, among 474 gigawatts of maybe, who gets to build. Prices can’t answer that question. What the state needs is credible commitment, a way to tell which developers will actually build. That is now the central problem in American power markets.
Sources consulted for this essay: ERCOT and PUCT publications on the energy-only market and the Batch Zero framework; the August and September 2026 directive letters from Governor Abbott; Baker Botts' analysis of Senate Bill 6; the McGuireWoods 50-state survey of data-center legislation; the Texas Comptroller's revenue estimates; BloombergNEF's August assessment of delayed load (scenario analysis); reporting from Utility Dive, PowerMag, New Project Media, and the Texas House Focus report on grid reliability.