Research note · 2026-04-24
11 GW of ERCOT's incoming data-center queue concentrates in the Panhandle — a load-pocket binding constraint Norris's aggregate methodology can't resolve
Tyler Norris's Rethinking Load Growth (Duke, Feb 2025) gives ERCOT a system-wide curtailment-enabled headroom of 10 GW at 0.5% annual curtailment. Cliff's publicly-disclosed ERCOT data-center site dataset suggests the binding constraint is geographic, not aggregate: 11 GW of the 17.6 GW incoming queue concentrates in the Panhandle across two sites — a load pocket whose ability to clear depends on transmission build-out, not ERCOT-wide capacity reserves.
- Cliff dataset compiled
- 2026-04-24
- ERCOT sites analysed
- 30
- Total public MW
- 22.6 GW
- Incoming (con. + planned)
- 18.1 GW
The lead finding
ERCOT's data-center queue concentrates geographically
The 11 GW in the Panhandle across just two sites is the most extreme load-pocket signal in Cliff's public ERCOT dataset — the kind of geographic concentration Tyler H. Norris et al.'s system-wide aggregate methodology cannot resolve. Load-zone assignment here uses nearest-centroid distance to Houston, DFW, San Antonio, West (Permian), and Panhandle reference points, which is operationally closer to transmission-binding than ERCOT's administrative four-zone mapping.
| Zone | Operational MW | Construction MW | Planned MW | At-risk MW | Total MW | Sites |
|---|---|---|---|---|---|---|
| Panhandle | 200 | 0 | 11,000 | 0 | 11,200 | 2 |
| North (DFW) | 1,980 | 2,230 | 1,500 | 0 | 5,710 | 16 |
| West | 300 | 2,200 | 500 | 0 | 3,000 | 4 |
| Houston | 1,352 | 0 | 200 | 0 | 1,552 | 4 |
| South | 700 | 485 | 0 | 0 | 1,185 | 4 |
11 GW planned across two sites in a geographically isolated corner of ERCOT is not a uniformly-distributed load addition — it is a transmission-binding load pocket whose ability to clear depends on SPP-seam and panhandle-expansion transmission projects, not ERCOT-wide capacity reserves. Norris's aggregate headroom framing is correct at the system level; the zone-level picture is what says where curtailment tolerance will actually be priced.
Aggregate upper bound
Even fully-flexible, the queue overshoots Norris's 0.5% headroom
This comparison assumes every MW of Cliff's 17.6 GW incoming-queue (construction + planned) would accept the given curtailment cap — the most-flexible upper bound. Most planned data-center load is currently structured as firm interconnection, so the realistic fit ratio is much lower than these bars show. Even at the upper bound, however, 7.6 GW remains outside Tyler H. Norris et al.'s 0.5% aggregate headroom, which reinforces the load-pocket finding above. Headroom values from Rethinking Load Growth, Duke Nicholas Institute (February 2025), Figure 8.
0.25% annual curtailment
6.5 GW fits · 11.6 GW overshoots
Norris headroom: 6.5 GW · fit rate 36%
0.5% annual curtailment
10.0 GW fits · 8.1 GW overshoots
Norris headroom: 10 GW · fit rate 55%
1.0% annual curtailment
14.7 GW fits · 3.4 GW overshoots
Norris headroom: 14.7 GW · fit rate 81%
Reading this chart at 0.5% curtailment — the budget most existing utility interruptible programs target — only 10 of 17.6 GW fits even under the most-flexible upper-bound assumption. The residual 7.6 GW must either (a) accept tolerance above 0.5%, (b) wait for transmission expansion, or (c) interconnect as a controllable load resource (CLR) under PGRR145 / NPRR1325. The Cliff dataset counts only publicly-disclosed sites; ERCOT's full Large Load Interconnection queue is materially larger, so this comparison is conservative by construction.
What this means for developers
Curtailment tolerance is now a gating site-selection parameter
If you can accept 1.0%
83% of Cliff's incoming queue fits. ~2.9 GW still must wait on transmission or accept tighter load-zone placement.
If you can accept 0.5%
57% fits. This is the curtailment budget most existing utility interruptible tariffs target; above it, you are in PGRR145 Provisional CLR territory.
If you can accept 0.25%
37% fits. Essentially only the operational and earliest- construction MW clears; most planned MW must reshape, relocate, or tolerate more curtailment.
Methodology
What went into the numbers
Norris side (paper)
Per the paper: 9 years of hourly EIA-930 load data (2016-2024), scipy.optimize.root_scalar goal-seek for max constant-24/7 new load addition at each curtailment cap. ERCOT-wide aggregate; no transmission constraints; no plant intertemporal constraints.
Cliff side (dataset)
Curated from official press releases, ERCOT meeting materials, SEC filings, and major-trade-press coverage. Lat/lon are city/county centroids unless a specific facility coordinate was disclosed; precision is sufficient for state-wide visualization, not for parcel work. MW values are public nameplate or disclosed targets, not metered demand. Compiled 2026-04-24. This is a leadgen visualization, not an authoritative datacenter registry. Expect entries to be ±15% on MW, missing recent disclosures, and stale on stage transitions. Refresh quarterly.
Known limitations of this analysis
- The Cliff dataset counts only publicly-disclosed sites; the comparison here is conservative by construction relative to ERCOT's full Large Load Interconnection queue.
- Norris's aggregate headroom is ERCOT-wide. The zone-level binding claim assumes load pockets bind on transmission before system-wide capacity reserves are tested; this is consistent with Panhandle transmission realities but has not been formally modelled here.
- MW values in the Cliff dataset are nameplate, not coincident peak. Applying realistic utilization factors would shift the fit ratios upward.
- Norris assumes constant 24/7 new load. Actual AI training duty cycles are bursty; relaxing this assumption (a future extension of this note) would expand the effective headroom.
Methodology review welcome before we publish more broadly
Every quantitative claim below extends the Norris methodology with Cliff's public ERCOT site dataset. Before publishing to a wider audience, we would value a methodology read from the authors of the paper — particularly on whether the aggregate headroom framing holds at the load-zone level, and whether the Panhandle concentration warrants a separate treatment.
Suggested citation: Cliff Research, “ERCOT's load-pocket binding constraint: 11 GW concentrated in the Panhandle,” 2026-04-24, cliffcenter.com/research/ercot-curtailment-headroom.
Built on Tyler H. Norris et al., Rethinking Load Growth, Duke Nicholas Institute, February 2025 · https://nicholasinstitute.duke.edu/publications/rethinking-load-growth.