Texas is being asked to plan for roughly seven times the electricity the entire United States data center industry currently draws, all inside one state. That is not a typo or an exaggeration for effect. According to a Reuters analysis published September 1, requests to connect large electricity users, overwhelmingly data centers, to the Texas grid have climbed from about 48 gigawatts in 2023 to more than 474 gigawatts today, a figure more than five times the state's own record peak electricity demand. Texas just became the first major U.S. data center hub to stop taking those requests at face value.

What Abbott Actually Ordered

Texas Governor Greg Abbott issued the directive on August 3, calling for a comprehensive audit of every data center project sitting in the interconnection queue managed by ERCOT, the state's grid operator. The order requires project developers to provide verifiable information proving their plans are real, including details on funding, ownership, water usage, taxpayer-funded incentives, and any on-site power generation. Projects that fail to substantiate their filings can be denied a grid connection outright. Texas Public Utility Commission Chairman Thomas Gleeson captured the underlying problem in blunt terms: "When you don't know what is real, you really don't know how to build the infrastructure for it."

That is the actual crisis here, not a shortage of power exactly, but a fog of uncertainty so thick that regulators can no longer tell which requests represent genuine, funded construction and which represent something closer to speculative real estate hoarding. Abbott reinforced the point in a follow-up letter to ERCOT and state regulators, writing that "the PUCT and ERCOT cannot make decisions to guarantee grid stability and reliability based on substantially incomplete information."

The Mechanics Behind "Ghost Demand"

The industry's own term for the phenomenon driving this freeze is "ghost demand" or "phantom demand," and the mechanism behind it is straightforward once you understand how interconnection queues actually work. Developers routinely file multiple requests across several different utilities for what is, in reality, a single eventual project, hedging against permitting delays, land acquisition falling through, or financing not coming together, often at little to no upfront cost to the developer. ECM Source's analysis of the Reuters reporting frames the scale plainly: the entire existing U.S. data center fleet consumes an estimated 60 to 70 gigawatts at any given moment. Texas alone is currently fielding requests equivalent to roughly seven times that entire national baseline.

Outside Texas, the same pattern is compounding nationally. Ten of the largest utilities across the Midwest, Mid-Atlantic, and South, including AEP Ohio, Southern Co, and PPL, have logged a combined 270 gigawatts of additional large-load requests, according to Reuters' review of recent utility earnings calls. Combine every regional queue nationally and the total tops 700 gigawatts, an order of magnitude beyond what the AI and cloud computing industry currently actually draws from the grid at any point in time.

ERCOT's New Approach to Sorting Real From Speculative

ERCOT has already begun restructuring how it evaluates these requests rather than continuing to process them one at a time. The grid operator introduced what it calls a "Batch Zero" system, examining major electricity users together as a group rather than individually, specifically so planners can determine how much demand the grid can realistically accommodate and what new transmission infrastructure that demand would actually require. ERCOT president and CEO Pablo Vegas described the shift as "a fundamental shift in how ERCOT manages the significant growth of large load interconnection." The grid operator has told regulators it aims to complete the full audit by December 2026, in time to feed accurate results into its next long-term load forecast.

There is already evidence that simply applying more scrutiny makes a meaningful share of the projected demand evaporate. Vision Times' reporting notes that ERCOT's own preliminary forecast in May had projected peak demand could more than quadruple by 2032, driven by data centers and other large-load customers, a forecast now facing direct pressure from the audit's early findings about how much of the underlying demand is genuinely committed versus speculative.

Other States Are Watching, and Some Are Already Copying

Texas is not acting alone, and the timing of similar moves elsewhere suggests other states saw this reckoning coming before Reuters' analysis made it national news. Pennsylvania Governor Josh Shapiro signed his own data center executive order on August 18, imposing stricter permitting requirements for any project consuming 25 megawatts or more, along with greater disclosure requirements covering project plans and end users. The results there are similarly telling: of more than 100 data center projects proposed across Pennsylvania, only about 20 have actually applied for the permits needed to move forward, according to a member of Shapiro's own office speaking to Reuters. Ohio has taken a more direct financial approach, with utility AEP restructuring its tariff to require a $100,000 study fee upfront specifically to price out purely speculative interconnection filings before they ever reach a formal review.

That pattern, a wave of loudly announced projects followed by a much smaller number of actual permit applications, is precisely the gap regulators are now trying to force into the open before committing billions of dollars in grid infrastructure to serve demand that may never materialize.

Why This Matters Beyond Texas Politics

This regulatory reckoning arrives at a moment when the broader AI infrastructure buildout is already straining under cost pressures from an entirely different direction. Nvidia itself recently had to pass double-digit price increases on to its largest customers as memory chip costs surged, a supply constraint expected to persist well into 2027. Layering genuine uncertainty about how much of the industry's projected electricity demand is even real on top of that hardware cost pressure paints a considerably murkier picture of the AI buildout's actual near-term trajectory than the headline capital expenditure figures alone suggest. Total AI-related infrastructure spending has been reported topping $700 billion, a number regulators are now treating with real skepticism given how disconnected Texas's own interconnection queue has become from its actual peak demand.

The Risk to Ordinary Electricity Customers

The consequences of getting this wrong extend well past the data center industry itself. If Texas or any other state builds new power plants, transmission lines, and grid capacity sized to serve inflated, partly fictional demand projections, the cost of that overbuilt infrastructure does not simply disappear. It gets recovered through the rates charged to every other electricity customer on the grid, whether or not the phantom data center projects that justified the buildout ever actually get constructed. That is the practical stake behind Abbott's audit and Shapiro's permitting overhaul: protecting ordinary ratepayers from subsidizing infrastructure built for demand that exists mostly on paper.

The economic geography of where AI-related growth is actually landing has already been shifting in ways that add another layer to this story. Texas has marketed itself aggressively as the preferred home for AI infrastructure investment, competing directly with other states for exactly the kind of large-scale data center projects now under this new audit. A state that spent years courting this specific category of investment is now the first to publicly admit it cannot tell how much of that investment interest is genuine without forcing developers to prove it first.

What Happens If the Real Number Turns Out Much Smaller

None of this means America's AI-driven electricity demand boom is entirely imaginary, and Vision Times' own reporting is careful to note that distinction directly. Real, well-funded, already-under-construction data center campuses exist across Texas and nationally in genuinely enormous numbers. The same underlying capital, chasing custom AI chip supply deals and infrastructure partnerships worth tens of billions of dollars each, reflects demand that is unambiguously real, even as the interconnection queue data suggests a meaningful share of the broader total is not. The genuine question ERCOT's audit is trying to answer, and the one every other state now watching Texas closely will eventually have to answer for itself, is how much smaller the real number turns out to be once developers are actually required to prove their projects exist beyond a filing in a queue. Whatever that final figure ends up being, it will shape how much new power generation capacity gets built across the American grid for the next decade, and who ultimately pays for it.