The Power Bill Briefing — June 2026

In this Briefing

1. PJM clears at the cap again — scarcity is now priced, not discovered
2. Data centers are the whole load-growth story — and EIA's demand curve confirms it
3. The gas-turbine famine — 100 GW of backlog and 10–20% higher prices ration dispatchable capacity
4. Nuclear becomes the AI-power balance sheet — IPPs re-rate, an SMR option enters PJM
5. A record 86 GW of 2026 additions — and 93% of it is solar, storage and wind
6. The battery fleets cross 14 GW each — but the merchant arbitrage trade is compressing

PJM clears at the administrative cap a second time — the auction now prices scarcity instead of discovering it

What's new. On 17 December 2025 PJM released its 2027/2028 Base Residual Auction, which cleared at $333.44/MW-day (UCAP) across the entire RTO footprint — the FERC-approved collar cap, and a second consecutive year at the ceiling after the 2026/2027 auction cleared at $329.17/MW-day. For the first time in PJM's history the market mechanism is no longer setting price; an administrative cap is. The auction secured 134,479 MW of capacity and demand response for the delivery year beginning 1 June 2027 [1][2].

Evidence. The 2026/2027 auction (results 22 July 2025) cleared at $329.17/MW-day, up 22% year-over-year for most of the footprint, with a total procurement cost of $16.1 billion, +9.5% from $14.7 billion the prior year [3]. The 2027/2028 result of $333.44/MW-day is +1.3% on that [1]. Critically, PJM's own simulation shows that without the collar, RTO prices would have reached $529.80/MW-day, with the Dominion Zone (northern Virginia's "data center alley") clearing at $542.83/MW-day — roughly 60% above the capped print [2]. PJM still fell short of its reliability target despite the cap.

The Take: The two-year cap streak masks an embedded forward step-up that the market has not yet priced. The uncapped clear of ~$530/MW-day is the real shadow price of capacity, and the collar's upper bound ratchets each year — meaning the capped print is a ceiling that is itself rising, not a stable plateau. Taking PJM's own uncapped simulation as the marginal signal, the implied additional capacity transfer being suppressed by the collar is on the order of (529.80 − 333.44) × 134,479 MW × 365 ≈ $9.6 billion per delivery year (The Power Bill estimate) — value that accrues to whoever owns unforced capacity the moment the cap relaxes or is litigated away. The cap is not protecting consumers from scarcity; it is deferring the bill and transferring duration risk onto load.

Market read. VST (NYSE) — Add · conviction Med · 12–24 mo — Vistra's large dispatchable + nuclear fleet inside PJM is levered to the uncapped scarcity value the collar is currently suppressing. CEG (NASDAQ) — Hold · conviction Med · 12–24 mo — already richly valued on nuclear optionality; capacity upside is real but largely in the price. NRG (NYSE) — Add · conviction Low · 12–24 mo — capacity-heavy retail+gen book benefits from elevated PJM clears, but less nuclear narrative premium.

Data centers are not a driver of load growth — they are essentially the driver

What's new. PJM's 2026 Long-Term Load Forecast confirms that large loads — principally data centers — account for more than 100% of projected peak-demand growth over the next five years, because underlying base demand is contracting beneath them. PJM's large-load adjustments grow by 35.1 GW between 2026 and 2031, against total demand growth of 34.6 GW [4]. Nationally, EIA's January 2026 outlook called the strongest four-year run of US electricity-demand growth since 2000 [5].

Evidence. PJM's forecast has summer peak rising from 160 GW in 2025 to 253 GW by 2046, a 58% increase led by data centers [4]. EIA's Short-Term Energy Outlook puts US electricity consumption at 4,193 TWh in 2025 and 4,283 TWh in 2026, both fresh all-time highs above the prior 4,097 TWh (2024) record, with commercial (data-center-heavy) demand growing ~2.6%/yr and industrial ~2.1%/yr [6]. PJM's Independent Market Monitor estimated that existing and forecast data-center load added a combined $23.1 billion to capacity-market revenues across the 2025/2026, 2026/2027 and 2027/2028 auctions [4].

The Take: The "more than 100% of growth" framing is the most important and least-appreciated number in US power. It means data-center load is not riding a rising tide — it is cross-subsidizing a shrinking residential/industrial base, and every megawatt of large load is therefore competing directly against existing ratepayers for the same scarce capacity. That structure is politically unstable: when 100% of the cost increase is attributable to a load class that is <20% of energy, the cost-allocation fight (bring-your-own-generation mandates, separate data-center rate classes, co-location tariffs) becomes the dominant regulatory risk of the next 24 months — a bigger swing factor for IPP cash flows than the commodity curve. Watch FERC and state-commission dockets, not just the auction.

Market read. EQIX (NASDAQ) — Hold · conviction Med · 12–24 mo — data-center REIT exposed to rising power cost and interconnection delay as a structural margin headwind, not just a demand tailwind. DLR (NYSE) — Hold · conviction Low · 12–24 mo — same dynamic; power procurement and on-site generation capex become differentiators.

The gas-turbine famine — capacity to build dispatchable power is now the binding constraint, and it is being priced like one

What's new. GE Vernova's gas-power equipment backlog plus slot-reservation agreements reached 100 GW in Q1 2026, up from 83 GW the prior quarter, and management expects to reach at least 110 GW by year-end 2026 and to be effectively sold out of turbine slots through 2030 [7][8]. The bottleneck has moved from the megawatt to the machine: you can have a power-plant permit and a PPA and still wait years for iron.

Evidence. GE Vernova's production runs at roughly 10 GW/year of slots through 2030, so a 100 GW book represents about a decade of output [8]. New-order pricing in H1 2026 is running 10–20 percentage points higher per kW than Q4 2025, outpacing inflation [7]. The company is targeting 20 GW of annualized turbine output by mid-2026 and says it could stretch its two existing facilities to ~24 GW by mid-2028 [8]. Meanwhile EIA's June 2026 STEO has Henry Hub averaging only ~$3.60/MMBtu in 2026 [9] — fuel is cheap; it is the turbine and the queue, not the gas, that are scarce.

The Take: Cheap fuel plus scarce iron inverts the usual gas-economics intuition. With Henry Hub near $3.60/MMBtu the variable cost of a new combined-cycle plant is unremarkable; the scarcity has migrated entirely into the capital term, and a turbine slot is now closer to a traded option than a procurement line. A 10–20% per-kW price step on a ~$1,200–1,400/kW machine implies roughly $120–280/kW of additional overnight cost (The Power Bill estimate) before any plant is built — which, levelized over a 20-year life at an 8% discount rate and a 60% capacity factor, adds on the order of $1.5–3.5/MWh to new gas LCOE purely from equipment inflation (The Power Bill estimate). That is small versus the capacity scarcity it competes against — which is exactly why buyers are paying it. The turbine OEMs, not the fuel, now capture the scarcity rent.

Market read. GEV (NYSE) — Add · conviction High · 12–36 mo — the clearest pure-play on dispatchable-capacity scarcity; pricing power and a decade-deep backlog with limited slot competition. ETN (NYSE) — Add · conviction Med · 12–24 mo — electrical equipment / grid-connection content rides the same build cycle with broader end-market diversification. PWR (NYSE) — Add · conviction Med · 12–24 mo — Quanta's EPC and interconnection labor is the other binding constraint once the turbine arrives.

Nuclear becomes the balance sheet of the AI-power trade — and a 300 MW SMR option enters PJM

What's new. In January 2026 Vistra and Meta announced agreements to support existing nuclear plants inside PJM, including an option for Meta to take power from a potential new 300 MW small modular reactor at a Vistra site [10]. It extends a now-established pattern: as of mid-2026 every major hyperscaler has signed at least one nuclear offtake, with ~13 announced projects committing over 9.8 GW of nuclear capacity to data centers [10].

Evidence. Anchor deals include Microsoft's 20-year, 837 MW PPA with Constellation to restart Three Mile Island Unit 1 (Crane Clean Energy Center), and Amazon's expansion of its Talen offtake at Susquehanna to 1,920 MW through 2042 [10]. The equity re-rating has been violent: Vistra (VST) is up ~45% over the trailing 12 months to March 2026 and >500% over five years; Constellation (CEG) trades at 17–21× EV/EBITDA versus Vistra's ~13×, a ~50% premium pricing in nuclear optionality [11]. The three merchant IPPs — VST, CEG and Talen (TLN) — have become the default vehicles for the "AI power" thesis [11].

The Take: The market is conflating two very different cash flows under one "nuclear-for-AI" label. The uprate/restart/existing-fleet deals (TMI, Susquehanna) are near-term, contracted, and low-execution-risk — they deserve a re-rating. The new-SMR option deals (the Vistra 300 MW SMR, and the broader TerraPower/Oklo pipeline) are real options with first-of-a-kind cost risk, a decade to first power, and no operating SMR fleet in the US to anchor LCOE. Pricing both at the same multiple is the error. A signed offtake on an existing reactor is worth a utility-like multiple; an SMR option is venture risk wearing a PPA. The CEG-over-VST premium is mostly existing-fleet quality — defensible — but any IPP multiple expansion justified by SMR optionality is the part of the trade most exposed to a single cost-overrun headline.

Market read. CEG (NASDAQ) — Hold · conviction Med · 12–24 mo — best-in-class existing nuclear fleet, but a ~50% multiple premium leaves limited room for upside surprise. TLN (NASDAQ) — Hold · conviction Med · 12–24 mo — contracted Susquehanna offtake is high-quality; valuation already reflects the data-center thesis. OKLO (NYSE) — Avoid · conviction Low · 12–24 mo — pre-revenue SMR developer carrying the first-of-a-kind cost and schedule risk the existing-fleet names do not.

A record 86 GW of new capacity is coming in 2026 — and 93% of it is solar, storage and wind

What's new. EIA expects US developers to add a record 86 GW of utility-scale capacity in 2026, of which solar, battery storage and wind together make up roughly 93% — the cheapest-to-build, fastest-to-deploy resources continuing to dominate net additions even as the system screams for firm capacity [12]. The mismatch between what gets built (energy) and what is scarce (capacity) is the defining tension of the current market.

Evidence. Per EIA's December 2025 generator inventory: 43.4 GW solar (51%), 24 GW battery storage (28%), 11.8 GW wind (14%) and 6.3 GW natural gas (7%) [12]. Solar additions would be ~60% above 2025 if realized. Renewables were ~33.4% of US utility-scale capacity as of March 2026, projected to reach 36.6% by February 2027 [13]. EIA's STEO has the combined solar+wind generation share rising from ~18% (2025) toward ~21% (2027), while gas+coal+nuclear slip from 75% to ~72% [6].

The Take: The 86 GW headline flatters the firm-capacity picture. Apply representative summer peak-coincidence (capacity-credit) factors — roughly 25% for solar, ~10–15% for wind, and treat 4-hour storage as energy-limited firm — and the effective firm capacity added by the 2026 class is far smaller than the nameplate: on the order of ~30–35 GW of peak-coincident capacity from 86 GW nameplate (The Power Bill estimate, using ~25% solar / ~12% wind / storage counted at duration-limited nameplate). That is why PJM can clear at a price cap and set a build record in the same year: the grid is adding energy and arbitrage depth far faster than it is adding the dispatchable, peak-coincident capacity the auctions actually pay for. The build record and the capacity crisis are not contradictory — they are the same fact seen from two sides.

Market read. FSLR (NASDAQ) — Add · conviction Med · 12–24 mo — US-manufactured modules favored by domestic-content economics as solar leads the 86 GW class. NEE (NYSE) — Add · conviction Med · 12–36 mo — the largest renewables developer monetizes the build record across solar, wind and storage. GEV (NYSE) — Add · conviction High · 12–36 mo — counterintuitively the better renewables-cycle hedge, since the firm-capacity shortfall the build record leaves behind is what its turbines fill.

ERCOT and CAISO battery fleets each cross 14 GW — but the merchant arbitrage trade is already compressing

What's new. US grid-scale storage scaled faster than any resource in history through 2025, with ERCOT surpassing CAISO as the largest fleet. ERCOT entered 2026 with roughly 13.9 GW / 22.9 GWh of operational grid-scale storage and CAISO close behind — but the very saturation that built the fleets is now eroding the energy-arbitrage spreads that financed them [14][15].

Evidence. US battery capacity reached 38.1 GW by end-Q2 2025, +63% year-over-year, with Q2 2025 the largest quarterly add on record [14]. ERCOT led at 14.2 GW (37% of US), CAISO at 12.4 GW (33%) as of mid-2025 [15]. The economics tell the cautionary half: ERCOT merchant battery revenues fell from ~$192/kW (2023) to an average ~$43/kW across 2024–2025 — a ~78% collapse — as real-time prices averaged $31/MWh lower in 2024 than 2023 and the fleet cannibalized its own spreads [16].

The Take: Storage is transitioning from a price-taker on volatility to a price-maker that suppresses it — and that is a structural, not cyclical, revenue headwind. A ~78% per-kW revenue decline against a still-growing fleet means each incremental ERCOT battery earns dramatically less than the one before; on current trajectory, pure energy-arbitrage merchant economics in ERCOT are approaching the point where new-build is uneconomic without a capacity or ancillary-services backstop (The Power Bill estimate). The investable thesis is therefore migrating from deploying more 4-hour batteries to owning the value stack that survives saturation — ancillary services, capacity accreditation, and longer-duration assets that arbitrage across days rather than hours. The winners of the next storage cycle will look more like capacity providers than traders.

Market read. TSLA (NASDAQ) — Hold · conviction Low · 12–24 mo — Megapack/Energy is a structural grower, but US merchant-arbitrage compression caps the standalone-storage TAM narrative. STEM (NYSE) — Avoid · conviction Med · 12–24 mo — storage software/optimization is squeezed precisely as merchant spreads that justify its value collapse. NEE (NYSE) — Hold · conviction Med · 12–36 mo — contracted/utility-owned storage insulated from merchant compression, unlike pure-play arbitrageurs.

Market Calls

References

1. PJM Interconnection, "2027/2028 Base Residual Auction Report," 17 December 2025. https://www.pjm.com/-/media/DotCom/markets-ops/rpm/rpm-auction-info/2027-2028/2027-2028-bra-report.pdf
2. PJM Inside Lines / PJM News Release, "PJM Auction Procures 134,479 MW of Generation Resources," 17 December 2025. https://insidelines.pjm.com/pjm-auction-procures-134479-mw-of-generation-resources/
3. Utility Dive, "PJM capacity prices hit record high as grid operator falls short of reliability target," July 2025. https://www.utilitydive.com/news/pjm-interconnection-capacity-auction-data-center/808264/
4. PJM Inside Lines, "PJM's Updated 20-Year Forecast Continues To See Significant Long-Term Load Growth," 2026. https://insidelines.pjm.com/pjms-updated-20-year-forecast-continues-to-see-significant-long-term-load-growth/
5. U.S. Energy Information Administration, Press Release, "EIA forecasts strongest four-year growth in U.S. electricity demand since 2000, fueled by data centers," 13 January 2026. https://www.eia.gov/pressroom/releases/press582.php
6. U.S. Energy Information Administration, Short-Term Energy Outlook, June 2026. https://www.eia.gov/outlooks/steo/
7. Utility Dive, "GE Vernova gas turbine backlog hits 100 GW as prices rise," April 2026. https://www.utilitydive.com/news/ge-vernova-gas-turbine-backlog-hits-100-gw-as-prices-rise/818332/
8. GE Vernova Inc., Form 8-K / Q1 2026 results, 22 April 2026. https://www.sec.gov/Archives/edgar/data/0001996810/000199681026000063/gevpressrelease1q26.htm
9. U.S. Energy Information Administration, "We expect Henry Hub natural gas spot prices to fall slightly in 2026 before rising in 2027," 2026. https://www.eia.gov/todayinenergy/detail.php?id=67004
10. SMR Intel, "Every Nuclear-Powered Data Center Deal: Google, Amazon, Meta & Microsoft (2026)," 2026. https://smrintel.com/nuclear-data-center-deals/
11. Lambda Finance, "Vistra vs Constellation vs Talen: Best AI Power Stocks (2026)," 2026. https://www.lambdafin.com/articles/vistra-vs-constellation-vs-talen
12. U.S. Energy Information Administration, "New U.S. electric generating capacity expected to reach a record high in 2026," April 2026. https://www.eia.gov/todayinenergy/detail.php?id=67205
13. North American Clean Energy / EIA data, "Renewables Were 26% of U.S. Electrical Generation and 36% of Installed Capacity in 2025…," 2026. https://www.nacleanenergy.com/solar/renewables-were-26-of-u-s-electrical-generation-and-36-of-installed-capacity-in-2025-as-eia-projects-even-faster-solar-wind-battery-growth-in-2026
14. S&P Global Commodity Insights, "US BATTERY STORAGE: ERCOT surpasses CAISO in Q2 for most operating battery storage capacity in US," 17 September 2025. https://www.spglobal.com/energy/en/news-research/latest-news/electric-power/091725-us-battery-storage-ercot-surpasses-caiso-in-q2-for-most-operating-battery-storage-capacity-in-us
15. Modo Energy, "ERCOT Annual Buildout Report: Battery capacity reaches 14 GW entering 2026," 2026. https://modoenergy.com/research/en/ercot-battery-buildout-2025-annual-report
16. energy-storage.news, "Where is alpha? Market trends in ERCOT and CAISO impacting battery storage revenues," 2025. https://www.energy-storage.news/where-is-alpha-market-trends-in-ercot-and-caiso-impacting-battery-storage-revenues/

Disclosures & Disclaimer

This report is general commentary published for information purposes only. It is not investment advice, a recommendation, or a solicitation to buy or sell any security, and it does not account for the objectives or circumstances of any individual. The Power Bill is a research publication, not a registered investment adviser or broker-dealer. Views are the publication's own analytical opinions, are subject to change, and may prove wrong. Markets involve risk of loss; past performance does not indicate future results. Readers should do their own research and consult a licensed financial professional before acting. The publication and/or its principals may hold positions in securities mentioned. Company facts and figures are drawn from public sources believed reliable but are not guaranteed. © The Power Bill.

About The Power Bill

The Power Bill publishes independent technical and techno-economic research in its field. Reports are prepared for subscribers and are provided for information only. © The Power Bill. All rights reserved.