State of the ERCOT large-load queue.

As of ERCOT's most recent public reporting, the Texas grid operator is tracking approximately 410 gigawatts of large loads seeking interconnection through 2030 — roughly five times the grid's all-time peak demand. About 87% of it is data centers. And of everything in that queue, less than 6 GW is actually energized. This page tracks the queue in plain numbers, updated monthly, so the scale of the line is never a matter of opinion.

The number

In testimony to the Texas Senate on April 1, 2026, ERCOT reported approximately 410 GW of large loads seeking interconnection — 410,618 MW in its March 26 status data, projected through 2030. For scale: ERCOT's entire generation interconnection queue, every solar farm, battery, and gas plant hoping to ever sell power in Texas, totals about 454 GW. The loads asking to plug in are nearly the size of all the generation asking to plug in.

How fast it's growing

The queue is not growing linearly — it's compounding. ERCOT's quarterly intake data shows large-load requests running at a few gigawatts per quarter through 2023 and most of 2024, then 63 GW, 76 GW, and 48 GW quarters through 2025 — and then 198 GW requested in the first quarter of 2026 alone, which ERCOT attributes to a "rush of requests for AI data center loads seeking firm service." One quarter brought more requested load than the previous three years combined.

How little gets through

The queue's composition, from ERCOT's March 26, 2026 status data:

Read that bottom line again. After years of applications, studies, and headlines, the share of ERCOT's tracked large-load queue that is actually drawing power is about one and a half percent.

Why it takes years

ERCOT's own process documentation is candid about the mechanics: under the current single-study framework, each new request in a transmission zone can invalidate the studies of every project that hasn't yet energized, pulling them into restudies — "delaying approvals and potentially extending timelines by years," in ERCOT's words. The fix now underway is a batch study framework, clustering applications and studying them together roughly every six months, with protocols targeted effective August 2026. Batching should make the process more predictable. It does not make a 400-gigawatt line short — and a six-month batching cadence means missing one window costs half a year before the study even begins.

What waiting now costs

The queue is no longer free to stand in. Under large-load interconnection standards proposed by the PUCT in March 2026, interconnection agreements would carry financial security of $50,000 per megawatt — five million dollars committed against a 100 MW request, converting to non-refundable at the interconnection-agreement stage — plus site control and disclosure requirements. And the security deposit is the small number: the large one is carry. Land held, capital committed, GPUs depreciating in crates — every quarter between decision and energization is cost with no revenue against it, multiplied by a queue measured in years. The full arithmetic is in the cost math.

The path that skips the lake

None of this is an indictment of ERCOT — the grid is doing careful engineering on an intake it was never designed for. But it is a fact pattern with a conclusion: for AI loads that need firm power on a construction timeline, the fastest path is the one that never enters the queue. Behind-the-meter generation — dedicated power built with the data center, on the customer's side of the meter — replaces studies and restudies with equipment, permitting, and construction. That's the model Corley Energy builds at Power Foundry in the Permian Basin — a faster alternative to the grid. How the queue works mechanically: the ERCOT queue, explained. How to get out of it: current capacity.