Texas Freezes the Queue: What a Data-Center Moratorium Reveals About Hidden Contracts and Phantom Load

In early August 2026, Texas Gov. Greg Abbott ordered a halt to all new data-center approvals pending an audit of projects seeking grid connection through ERCOT, the state's independent grid operator.^[1] The moratorium targets a queue holding approximately 474 GW of interconnection requests, of which roughly 90 percent are data centers.^[2] The EIA responded by cutting Texas's projected electricity demand growth forecast in half, from 14 percent to 6 percent.^[1] On its surface, this is a public-interest governor checking runaway AI buildout. Read closer, it is an audit of a cost-allocation scandal hiding in plain sight.

The 474 GW figure is the scandal's measure. To contextualize: it exceeds five times ERCOT's record peak electricity demand.^[2] BloombergNEF estimates that the audit could delay approximately 49.8 GW of data-center load, nearly 20 percent of the entire US development pipeline.^[5] Yet how much of that queue represents signed, collateralized contracts with long-term minimum-take guarantees? The record does not say. Enverus Intelligence Research estimates roughly 12 GW of "high-confidence load" among ERCOT's data-center projects.^[6] That leaves 37+ GW of announced or speculative capacity, developers shopping interconnection requests across multiple utilities, filing duplicate queues, and treating the grid as a free option on future demand. Every MW sitting in that queue either is backed by a utility's load forecast (justifying capex today, socialized into rates) or it is phantom load obscuring the real buildout beneath it. Abbott's audit cannot fix the cost-allocation mechanism; it can only force disclosure of what is contracted versus conjured.

The deeper problem is tariff structure. Data centers in Texas, as in most states, negotiate special contracts with ERCOT and its member utilities, typically confidential documents filed with the Public Utility Commission of Texas under a seal that hides the price, the demand ratchet, the collateral, and the exit fee.^[1] Ratepayers fund the new generation, transmission, and reserved capacity. The customer class that drives the load growth pays a discounted rate. Residential and small-business customers carry the cost if the load does not materialize or underperforms its 10 to 15 year contract term against a 40-year asset life. Harvard Electricity Law Initiative research has documented this structure across multiple states: utilities extract quasi-monopoly rents from the public to serve hyperscalers, justified by confidential contracts and opaque load forecasts that no regulator can interrogate in real time.

The moratorium creates a window. Abbott's directive asks for disclosure of tax breaks, power and water use, community impact, and ownership.^[3] Notably absent: any requirement that new data-center capacity be assigned to a separate customer class with cost isolation, that projects bring their own generation or battery storage, or that grid interconnections be curtailable and flexible rather than firm and full-time. The state has not demanded that special contracts contain a minimum-take ratchet (say, 85 percent of contracted transmission and 60 percent of generation capacity over the contract term) or that exit fees cover stranded investment. Behind-the-meter power is already booming, Meta, Microsoft, Amazon, and others are installing on-site gas turbines and diesel backup without extensive environmental review.^[8] This is the workaround; it leaves the grid-dependent projects (the ones clogging the queue) competing for capacity that ratepayers are funding on forecast alone.

The audit itself is not a solution; it is a disclosure mechanism. Its usefulness depends on whether the PUCT and ERCOT are willing to (1) name which projects are contracted versus announced; (2) net out duplicate requests across utilities to eliminate phantom load; (3) separate real from speculative pipelines; and (4) require that any new load-growth forecasts used to justify rate-base capex include historical realization rates from prior IRPs. If the audit simply clears the queue of weak projects while leaving tariff and cost-allocation structures intact, the cycle repeats: new forecasts, new special contracts (still sealed), new capex, new stranded costs. The window is open now, roughly mid-August to late fall 2026, to file comments demanding that the PUCT establish a large-load tariff modeled on Virginia's GS-5 or Ohio's AEP deal, with binding minimum terms, high demand ratchets, collateral, and cost isolation. Without that, Texas's pause on approvals becomes cover for the same cost-shifting to happen more slowly.