# Cheap Supply, Scarce Connection

## A policy brief for commissioners, legislators, and federal regulators

*Condensed from the series "The Great Inversion." Citations refer to the full papers' source lists.*

---

## 1. The problem

The institutions governing electricity were built for yesterday's technology. Physical constraints are real, but the rules make manageable constraints behave like permanent exclusions.

| Metric | Then | Now (2026) |
|---|---|---|
| Utility-scale solar LCOE | $359/MWh (2009) | $40-98/MWh, avg $69 (Lazard 2026) |
| Battery pack price | ≈$1,475/kWh (2010, real 2025$) | $108/kWh; stationary $70 (BNEF 2025) |
| Active interconnection queue | Fraction of the fleet (mid-2000s) | 2,061 GW, about one and a half times the fleet; 549 GW holds an interconnection agreement (LBNL 2026) |
| Typical queue duration | <2 years (mid-2000s) | ≈5 years (LBNL) |
| Queue completion rate | Majority connected | 13% of capacity (LBNL) |
| Gas turbine lead time | ≈2 years (2019) | ≈5 years (S&P Global) |
| Annual capacity cost | $2.2B (2024/25) | $16.4B (2028/29), 4 straight cap-bound auctions (PJM) |

Electrons got cheap. Connection got scarce. The bottleneck is institutional, which is good news, because rules are cheaper to change than infrastructure is to build. A 500 MW load's engineering timeline is 12-18 months for solar-plus-storage; heavy-duty gas turbines and study-first interconnection each add roughly five years (PJM's average application-to-operation timeline now exceeds eight years, per RMI), while ERCOT-style rules cut connection to two or three and a flexibility agreement (per Google's 2025 deals with I&M and TVA, since scaled toward a contracted gigawatt) serves it in months. Procedure, not physics, sets the schedule.

## 2. Why this happened

Three findings, each from a different discipline, jointly explain the pattern.

**Ratemaking pays for capital.** A utility earning a return on invested capital will prefer a $40 million substation it earns on for forty years over a $2 million per year battery contract it earns nothing on (Averch-Johnson, 1962). Case evidence that the cheaper option works when someone is paid to choose it: Con Edison deferred roughly $1 billion of Brooklyn-Queens grid investment with ~$200 million of demand-side resources under a bespoke NY earnings mechanism (net savings contested, direction not); Arizona Public Service deferred a ~20-mile line rebuild with a 2 MW battery, since retired; Sterling, MA reports ~$400k/year on a Sandia-analyzed municipal battery. Each required a carve-out or public power, which is the diagnosis restated as evidence. The bias operates at every voltage: reconductoring can roughly double a line's capacity at roughly half the cost of new build where conductor and structural conditions allow (PNAS, 2024), and utilities rarely propose it, because a cheaper solution is smaller rate base.

**Gatekeepers face no cost for delay.** Interconnection is run as discretionary, serial study by entities with weak incentives for speed. Texas is the most useful real-world comparison, a mechanism worth testing: ERCOT's connect-and-manage framework connects comparable resources several times faster, managing congestion through the market afterward.

**The reliability construct collapsed under load growth.** Delayed auctions, overnight accreditation changes, and a decade of underestimated data-center demand produced a capacity market that has hit administrative caps four auctions running instead of discovering prices. Most recently it fell 6,623 MW short of its reliability requirement, with the costs of speculative load socialized onto all customers. The data centers themselves would rather contract for their own capacity; Google's 2025 curtailment agreements with Indiana Michigan Power and TVA show the demand for a product the institutions never offered.

Utilities face real pressures (wildfire, storms, cyber, aging assets), and the diagnosis does not require bad faith anywhere. It requires only that people follow the incentives they are given. That is what makes it fixable.

## 3. The five principles, and the preferred mechanism under each

Each principle carries a menu of mechanisms evaluated in Paper III, with a stance and an evidence tier for each, including the alternatives we set aside and the ones we reject. What follows is the principle and the preferred mechanism under it; a commission that reaches the same function by another mechanism has used this program effectively.

**Principle 1. A utility should earn on outcomes delivered, not on capital deployed.** Preferred mechanism: performance-based regulation for wires. Totex revenue caps (erasing the capex/opex accounting line), output incentives tied to interconnection speed and hosting capacity, and yardstick benchmarking against peer utilities. Precedents: Ofgem's RIIO since 2013; Hawaii PUC Docket 2018-0088. British networks kept investment-grade credit and earned roughly 9 percent real on regulated equity under RIIO-1, against a 5-6 percent UK-company average, which the National Audit Office criticized as too generous. That is the message to carry into every utility conversation: the documented failure mode of performance regulation is utilities earning too much.

**Principle 2. The party that owns the wires should not decide who may connect to them.** Preferred mechanism: an independent or ringfenced distribution system operator. Interconnection processing, hosting-capacity maps, envelope calculation, flexibility procurement, and data access move to an independent or ringfenced Distribution System Operator. Precedent: Britain's RIIO-ED2 licence conditions requiring functional separation from April 2023, plus outright separation of the national system operator (NESO, 2024). Principle: the owner of a bottleneck cannot referee access for its own competitors, the same logic as FERC Order 888.

**Principle 3. Where and when a resource operates should affect what it earns.** Preferred mechanisms: the maturity ladder. Default dynamic retail rates with hedged flat options; posted, technology-neutral locational offers wherever deferral value exists on a feeder; dynamic operating envelopes replacing static interconnection caps (flexible exports are now the default connection standard for new South Australian solar). Full distribution nodal pricing waits until this substrate exists.

**Principle 4. Access should follow published rules, not case-by-case permission.** Preferred mechanism: connect by published rules, both voltage levels. A standardized, software-screened pathway for small resources meeting published operating limits; connect-and-manage as the required pathway where technically feasible for bulk generation and large loads; Order 1920 enforcement plus reconductor-first showings before greenfield transmission; fast-track interconnection for hyperscale loads in exchange for verified curtailability during the top 50-100 system hours (Duke's Nicholas Institute estimates on the order of 100 GW of headroom from such flexibility).

**Principle 5. Incremental system costs should sit with those who create them.** Preferred mechanisms: the large-load tariff first, incremental reliability obligations phased behind it. Each load-serving entity, hyperscale loads included, demonstrates forward physical or contracted coverage of its coincident-peak contribution, against a transparent penalty price; the centralized auction shrinks to a residual and sunsets. French precedent. For generators this means long-tenor contracts in place of an auction that paid $28.92/MW-day one cycle and $269.92 the next, an administrative cap it has since hit four times running; contracts are what project finance is built on.

**What this is not.** No nationalization, no elimination of vertical integration where states prefer it, no forced retail choice, no DER mandates, no technology subsidies, no relaxed reliability standards. The objective: restore competitive markets wherever natural monopoly no longer exists, and regulate the remaining monopoly (the wires) on performance.

## 4. What it is worth

**The system bottom line.** A scenario-based estimate suggests roughly $100-220 billion of avoidable generation capex if 100 GW of large-load growth is served through flexibility on the existing system (Duke estimate priced from EIA's $921/kW reference to observed market costs near $2,000/kW); doubled corridor throughput from reconductoring at about half the cost of new build (PNAS), now being financed at scale through DOE's $3.26 billion loan to AEP Texas for 2,800-plus miles; and each ten percentage points of improved queue completion connecting roughly as much capacity as the country added over the past decade (LBNL). Capacity-cost incidence shifts by design from households to the loads that drive the costs.

**The consumer bottom line.** Electric bills have three cost drivers and the reforms discipline each: the supply line (cheap generation finally connects and competes), the reliability line (obligations move the $2.2B-to-$16.4B capacity escalation off households and onto the loads driving it), and the delivery line, now the fastest-growing part of many bills (totex reduces the financial preference for capital-intensive solutions behind 2025's record $31 billion in rate requests; reconductor-first buys transmission at roughly half greenfield cost; deferral savings are shared back). Protections ship with it: hedged flat options under default dynamic rates, transition bill protection, low-income rates, non-retroactive tariffs. The commitment is a published metric: real bill growth in participating states below matched peers.

## 5. Who must act

| Actor | First moves |
|---|---|
| State PUCs | Totex performance-based regulation orders; default dynamic rates; automated interconnection screening; machine-readable planning data |
| State legislatures | Performance-based regulation enabling statutes; separation of access administration; statutory right to connect; standard value-stack tariffs with scheduled, non-retroactive updates |
| FERC | Rulemakings: connect-and-manage default; Order 2222 deadlines with penalties; Order 1920 enforcement; approve obligation transition tariffs |
| RTOs/ISOs | Obligation accounting and settlement; large-load fast tracks; interregional planning |
| Utilities | File performance-based regulation plans; stand up ringfenced DSO functions; reconductor-first showings |
| Large loads and LSEs | Demonstrate forward hedges; standardize curtailability contracts |

No core state reform requires new federal legislation; some federal reforms require FERC action under existing authority, and federal legislation (backstop siting, seam clarification) could accelerate or clarify the program without gating it.

## 6. Measures of success

By 2030: small-resource interconnection in days; connection offers for bulk resources and large loads inside 18 months; five states under totex performance-based regulation; the first delivery year in which new data-center load demonstrably carries its own capacity. By 2035: queue conversion above half for mature cohorts; interregional transfer capability growing; real retail bill growth below matched peer states. The failure metrics of Section 1 provide the annual control chart.

## 7. Objections answered

*"The queue is speculative":* largely true, and the speculation is the institution's product; slow studies with unpredictable cost allocation make multiple positions a rational option. Even at the historical 13 percent completion rate, today's queue implies ~270 GW of eventually built capacity, several years of additions at recent build rates, waiting on procedure. And 549 GW holds a draft or executed interconnection agreement, though over 40 percent withdraws even after that stage. Order 2023's deposits are already trimming the froth (the queue shrank 10 percent in 2025) without easing the scarcity. *"Connect-and-manage threatens reliability":* ERCOT has run it through nearly two decades of the nation's fastest resource growth, Britain adopted it in 2010, and ERCOT's real reliability failure (Uri, 2021) traced to weatherization and fuel in the FERC-NERC inquiry, which no interconnection study prevents. *"Utilities need capital":* agreed; totex funds hardening, wildfire, cyber, and replacement fully, and changes only the distorted choice at the margin (build vs. buy). *"DERs aren't dispatchable":* accreditation already answers this; PJM's demand-response accreditation moved from 69% to 92% between consecutive auctions as availability rules tightened, batteries respond sub-second, and Principle 5's obligation accounting prices every resource at its measured contribution. *Dynamic rates and vulnerable customers:* defaults ship with hedged options and bill protection; the flat status quo already transfers from inflexible to flexible customers. *Utility opposition:* Principle 1 comes first, changing the utility's payoff before anything moves its control, and the British record shows the framework preserving investment-grade financeability, with its documented early weakness being excessive outperformance, answered by tighter calibration rather than by financial distress.

---

*Full argument, sources, coalition analysis, and international lessons: Papers I-III of "The Great Inversion" series. Dates and milestones: the path in the executive summary.*