Skip to content
By NiLu Updated

Electric grid in the United States — structure and bottlenecks

There's no single national bottleneck in the US: the FERC/NERC/ISO map, the generation queue, and the large-load process are distinct — with the right label.

Article summary

  • There is no "the queue," nor a single national bottleneck in one GW figure, in the US — there is an institutional map (FERC/NERC/ISO vs. vertically integrated utilities), two distinct interconnection families, and cost rules that change by ISO and by docket.
  • FERC Order 2023 reformed the generation queue; large loads (data centers, among others) follow another federal track, still open (RM26-4 / 2026 show-cause orders).
  • In ERCOT, large-load connection requests (hundreds of GW) are not the historical demand peak, nor NERC's growth forecast, nor the generation queue — they are four different numbers.
  • PJM, ERCOT, and MISO have distinct market and queue designs; there is no national average that describes all three at once.

The institutional map: FERC, NERC, and the ISOs

To understand any bottleneck figure in the US, one must first identify who decides what. FERC (Federal Energy Regulatory Commission) exercises federal economic regulation over wholesale electricity and interstate transmission — rates and terms — in addition to reliability oversight, approving and enforcing Reliability Standards under FPA §215. [1] NERC is the Electric Reliability Organization (ERO) certified by FERC for the continental US (excluding Alaska and Hawaii): it develops the Reliability Standards, subject to FERC approval, with Regional Entities operating under it. [1]

The EIA’s page on wholesale electricity markets lists ISO-NE, NYISO, PJM, MISO, SPP, ERCOT, and CAISO as the organizations that operate bulk systems and wholesale markets in the US — RTOs (Regional Transmission Organizations) are independent, membership-based entities that match supply and demand offers. [2] But this institutional list is not the country’s entire electrical map: a material share of load remains outside RTO/ISO footprints, notably in the Southeast and parts of the West, where vertically integrated utilities and balancing authorities, under state regulation, dominate day-to-day operation. [2]

ERCOT is a special case within this list. It appears as an ISO in the EIA’s index, but its wholesale market design and, above all, its large-loads process are governed primarily by Texas state law and the PUCT (Public Utility Commission of Texas) — via, for example, SB 6 — and not by FERC’s federal pro forma in the same way as the other jurisdictional RTOs/ISOs. [3] Not by coincidence, ERCOT is not among the six RTO/ISOs that FERC targeted in its June 2026 show-cause orders, as we’ll see next.

Two queues: generation and large load are not the same thing

The most common mistake in reading news about “the American queue” is treating generation and large load as the same process. They are not.

FERC Order No. 2023 (issued 2023-07-28, with Order 2023-A on 2024-03-21) reforms the pro forma Large/Small Generator Interconnection Procedures and Agreements: first-ready, first-served cluster studies, with financial-readiness and site-control requirements, including specific pathways for storage and hybrid projects. It is a rule for connecting generation — it is not the mature federal rule for connecting large loads (such as data centers). [4]

The instrument that measures this generation queue is Queued Up, published annually by LBNL. The 2026 edition (with data through the end of 2025) aggregates the public queues of more than 50 transmission operators, covering about 98% of US installed capacity: about 8.200 active projects, totaling approximately 1.312 GW of generation and 749 GW of storage — a total of about 2.061 GW of active capacity seeking interconnection. The methodology explicitly excludes load interconnection requests, distribution-connected projects, and behind-the-meter capacity. [5] In other words: that GW figure is the generation/storage queue, not the load queue.

And that number doesn’t equate to capacity that will actually operate, either. Of the total that requested interconnection between 2000 and 2020, only 13% had reached commercial operation date (COD) by the end of 2025; 75% was withdrawn; 10% remained active in the queue. For projects that entered operation in 2025 (in regions with data), the median between the interconnection request and COD exceeded five years. [5] Still within the generation/storage queue, 549 GW already had a draft or executed interconnection agreement (IA) without yet having a COD — a signed agreement is not synonymous with operation. [5]

The federal track for large loads is different, and still under construction. On October 23, 2025, a directive from the Energy Secretary under §403 led to an ANOPR (Advance Notice of Proposed Rulemaking) on the interconnection of large loads to interstate transmission — FERC docket RM26-4-000, framing large loads as generally above 20 MW. On June 18, 2026, FERC issued six show-cause orders under FPA §206, tailored to six jurisdictional RTOs/ISOs — PJM, CAISO, MISO, NYISO, ISO-NE, and SPP —, and not a single final national rule; the ANOPR remains open. The categories cited include the study process and alternative transmission technologies; preventing cost shifting and cost transparency; co-location and behind-the-meter generation; new transmission services for flexible large loads; and studying generating facilities that serve electrically close or co-located loads. The deadlines cited are 60 days for a justify-or-file tariff and 30 days for an informational resource-adequacy report. [10]

ERCOT: number hygiene

ERCOT is, today, the area where the most numbers circulate without labels — and where confusion between the load queue and the generation queue is most common. Mid-2026 materials clearly separate two queues:

The load queue — Large Load Interconnection (LLI). An ERCOT document from June 2026 cites more than 438.000 MW in connection requests from large users (about 90% data centers); a deck for the Texas Senate from July 2026 cites about 474 GW of Large Loads seeking interconnection, with about 205 GW preliminarily eligible for the so-called Batch Zero — preliminary eligibility is not approval. This same document defines a large user, for the process in question, as single-site peak demand above 75 MW, and describes the transition from case-by-case studies to a batch framework (Batch Zero in July 2026; Batch 1 requests in summer 2027), under implementation by the PUCT via SB 6 — not under FERC’s Order 2023. [3] [6]

The generation queue, separately. In the same mid-2026 snapshot, ERCOT reported about 1.949 active generation interconnection requests, totaling approximately 463 GW as of June 30, 2026 (solar ~162 GW, storage ~171 GW, gas ~78 GW, wind ~48 GW). [6]

None of these numbers is ERCOT’s historical demand peak, cited as 85.508 MW on August 10, 2023 — which is already-connected, realized peak demand, not an interconnection request. [3] [6] And none of them is NERC’s forecast for large-load growth in ERCOT through 2030, which we’ll see next. Treating any one of these four numbers as synonymous with the other three is the easiest hygiene mistake to make when reading about Texas.

Data centers in TWh: a different scale of measurement

While ERCOT and the interconnection queues measure MW and GW of requests, the Department of Energy and LBNL measure data-center demand in energy — TWh and percentage share of national electricity. These are different scales, and should not be added to the MW queues.

The DOE’s December 20, 2024 release on the U.S. Data Center Energy Usage Report states that data centers consumed about 4,4% of US electricity in 2023 (176 TWh, against 58 TWh in 2014), with a 2028 scenario range between 6,7% and 12% of national electricity (325 to 580 TWh). [7] LBNL’s 2025 update presents a reference case of 649 TWh in 2030, about 11,8% of US electricity (scenario band between 9,5% and 15,3%; 521 to 843 TWh). The model, a bottom-up type (based on IT shipments, energy per device, and cooling), explicitly does not model interconnection queues nor on-site supply adequacy. [8]

NERC, meanwhile, in its 2025 Long-Term Reliability Assessment (LTRA, with mid-2025 forecasts), measures something else: planning peak growth by assessment area. The report reports aggregate summer-peak growth across assessment areas exceeding 224 GW over ten years (+69% versus the 132 GW growth of the previous cycle), with winter peak growing about 245 GW; the text attributes most of the projected increase to new data centers from the AI/digital economy, but notes that later forecast revisions from ERCOT and PJM — which reduced some short-term large loads — were not included in this edition. [9]

By area (keeping the numbers separate, without a national average): MISO projects summer peak rising from 127 GW (2026) to 143,7 GW (2035), with about 18 GW of data-center loads projected by 2035; PJM projects +56 GW to 210 GW in 2035; and Texas RE (the area covering ERCOT) projects summer peak rising from 94.650 MW (2026) to 154.077 MW (2035), with a large-load interconnection forecast of 45 GW by 2030 (of which 23 GW would be data centers). [9] Note that this 45 GW forecast for ERCOT is a planning forecast — not the roughly 438-to-474 GW of LLI requests seen in the previous section.

PJM and MISO: different designs, no national average

Each ISO/RTO solved this same problem — generation growing, large loads arriving, queues tightening — in a different way. It’s worth looking at two examples alongside ERCOT, without trying to extract an average of the three.

At PJM, the 2026 Long-Term Load Forecast (LTLF) shows continued long-term growth, but a reduction in short-term peaks relative to the 2025 report, after a tighter vetting of large-load and data-center adjustments submitted by the distribution companies (EDCs), of electric vehicles, and of economic assumptions. PJM classifies these adjustments between Firm (with committed contract/capacity) and Non-Firm, with defaults such as a minimum 36-month ramp and 70% utilization unless supported otherwise. Public materials cite a framing of data-center growth of up to about 30 GW between 2025 and 2030 — but this is a planning and adjustment input, not a confirmed commercial-operation (COD) figure. [14] On the generation side, PJM reports, since 2023, processing more than 170.000 MW in new requests, with about 30.000 MW remaining in the transition queue for 2026, about 57 GW already with completed studies and signed or offered GIAs, and about 3.045 MW of new generation actually interconnected in 2025 (mostly solar). [15]

At MISO, the standard generation path goes through cluster studies in the Definitive Planning Phase (DPP), under Order 2023. A mechanism of its own, the Queue Cap, limits the total capacity studied per cycle to 50% of the region’s non-coincident peak — a backstop against a queue approaching or exceeding the regional peak itself. ERAS is an expedited, temporary path, with a serial design (in contrast to the DPP’s clusters), for “shovel-ready” short-term resource-adequacy resources. JTIQ, a joint initiative with SPP, is a portfolio designed to enable about 28,6 GW of generation along the seam between the two regions, with board approvals in December 2024. [16] [17] All of these are generation-side tools — not a national standard for a load queue.

The practical takeaway: PJM combines a capacity market with Firm/Non-Firm vetting of large-load adjustments and a cycle-based generation queue; ERCOT is energy-only, state-designed for wholesale, with LLI batches separate from the generation queue; MISO stands out for its 50%-of-regional-peak cap on the generation queue and for an expedited serial path. Three designs, with no average that represents all of them at once.

Who pays: Order 1920 and the 2026 show-cause orders

The question “who pays for grid reinforcement” has a two-layer answer, and it’s worth not mixing them.

The first is FERC Order No. 1920 (May 2024), with the 1920-A and 1920-B refinements: it requires transmission providers to conduct long-term regional planning (a horizon of at least 20 years) and to file standard, ex ante cost-allocation method(s) for the selected long-term regional facilities — a method that must allocate costs at least “roughly commensurate” with estimated benefits. Methods agreed to by the states must be filed even when the transmission provider prefers an alternative; the final decision rests with FERC. [13] This is the regional wire-reinforcement cost-allocation layer — different from the network-upgrade cost assignment for a specific generator under Order 2023.

The second layer is the suite of show-cause orders from June 18, 2026, already mentioned. Among the categories named in FERC’s materials are explicitly “preventing cost shifting” and cost transparency, alongside co-location, alternative transmission technologies, and new services for flexible large loads. The associated staff presentation cites that nearly two-thirds of the country’s load is under the Commission’s jurisdictional tariffs, and that the ANOPR received more than 3.500 pages of comments. [11] [12] The tariff outcomes of these orders were still pending as of this text’s access date.

It’s worth noting that the language about “ratepayer” protection circulating around these orders is docket and regulatory-process language — not an editorial verdict by NiLu about who should pay. At the Open Meeting of June 18, 2026, at which the suite was issued, FERC’s chairmanship characterized the action with terms like “fair cost allocation,” “transparency for the American ratepayer,” and speed-to-power, as well as respect for states’ rights. That is the institutional position of whoever signs the order on the day of issuance — legitimizing rhetoric for the process, not a closed tariff outcome. [18]

Uncertainties

  • There is no national database, among this dossier’s sources, for the large-load (load) interconnection queue comparable to the generation/storage Queued Up, with multi-ISO coverage and coverage outside RTO footprints as well — without it, any “American large-load queue of X GW” tends to be an extrapolation from the ERCOT case or a false equivalence between generation and load.
  • It is not yet known which tariffs or processes the six jurisdictional RTOs/ISOs will actually file or justify in response to the 2026 §206 show-cause orders, nor whether the RM26-4 ANOPR will become a final rule — the tariff outcomes were pending as of the access date.
  • It is not possible to say, with the data available, how stable the mid-2025 LTRA peaks are (and PJM’s framing of about 30 GW of data centers between 2025 and 2030) after the short-term revisions from ERCOT and PJM themselves and the criticisms about double-counting or flexible load.
  • Large-load rules in areas outside RTOs (the Southeast and parts of the Western US) and in ISOs with little coverage in this cycle (CAISO, SPP, NYISO, ISO-NE) remain under-documented here — the picture presented is robust for PJM, ERCOT, and MISO, not for the whole country.

References

  1. FERC — Reliability Explainer. FERC. Accessed August 09, 2026.
  2. EIA — Electricity Wholesale Markets (ISO/RTO). U.S. Energy Information Administration. Accessed August 09, 2026.
  3. ERCOT — Trending Topic: New Batch Connection Process for Large Electricity Users. ERCOT. Accessed August 09, 2026.
  4. FERC — Explainer on the Interconnection Final Rule (Order No. 2023). FERC. Accessed August 09, 2026.
  5. LBNL — Queued Up 2026 Edition: Characteristics of Power Plants Seeking Transmission Interconnection. Lawrence Berkeley National Laboratory. Accessed August 09, 2026.
  6. ERCOT — Senate Panel Presentation, “Assessing the Grid” (July 29, 2026). ERCOT. Accessed August 09, 2026.
  7. U.S. Department of Energy — DOE Releases New Report Evaluating Increase in Electricity Demand from Data Centers. U.S. Department of Energy. Accessed August 09, 2026.
  8. LBNL — United States Data Center Energy Usage Report, 2025 Update. Lawrence Berkeley National Laboratory. Accessed August 09, 2026.
  9. NERC — 2025 Long-Term Reliability Assessment. North American Electric Reliability Corporation. Accessed August 09, 2026.
  10. FERC — RM26-4 Large Loads Docket. FERC. Accessed August 09, 2026.
  11. FERC — FERC Launches Aggressive, Targeted Action to Speed Large Load Integration (June 18, 2026). FERC. Accessed August 09, 2026.
  12. FERC — E-7 to E-12 Show-Cause Orders, Staff Presentation (June Commission Meeting 2026). FERC. Accessed August 09, 2026.
  13. FERC — Explainer on the Transmission Planning and Cost Allocation Final Rule (Order No. 1920). FERC. Accessed August 09, 2026.
  14. PJM Inside Lines — PJM’s Updated 20-Year Forecast Continues to See Significant Long-Term Load Growth. PJM. Accessed August 09, 2026.
  15. PJM Inside Lines — 2025 Year in Review: Planning Prepares for Burgeoning Electricity Demand. PJM. Accessed August 09, 2026.
  16. MISO — Generator Interconnection (DPP / ERAS / JTIQ). MISO. Accessed August 09, 2026.
  17. MISO Help — Generator Interconnection Queue Cap FAQ. MISO. Accessed August 09, 2026.
  18. FERC — Chairman Swett’s Remarks on Large Load Show-Cause Orders E-7–E-12 (Open Meeting, June 18, 2026). FERC. Accessed August 09, 2026.