Cycle on Oct 19, 2026

I'll start by reviewing the open threads, prior notes, and pulling market data for the core entities in parallel.

I have the lay of the land: 15 open threads (several one-shot watches with October/November fire dates), 22 prior notes, and the macro backdrop. Key checks for this cycle: (1) the FERC IRAS ruling watch (fire date was Oct 15 — did FERC rule?), (2) the October EIA STEO (~2nd Tuesday = Oct 13), (3) any new PPAs/load commitments or earnings guidance in the past week. Let me search for the latest news and pull ticker views for the core IPP/utilities.

Major development confirmed: FERC ruled Oct 9 on PJM's IRAS proposal — accepting the Large Load Registry but rejecting the curtailment mechanism and BYONC. That resolves the IRAS watch (fire date Oct 15) and materially changes the BYONC watch. Let me get the full story from Utility Dive and check the October EIA STEO, plus pull more ticker views.

FERC's Oct 9 order is confirmed with rich detail, and the October STEO (released Oct 6) has a striking regional divergence: PJM wholesale power prices +41% in 2026 while the Pacific Northwest's Mid-Columbia hub falls 23%. Let me pull the STEO electricity section for the residential price forecast, get views on the regulated utilities, and scan the past week for new PPAs/load commitments.

Rich week of developments. Key threads: FERC's IRAS rejection (Oct 9), AEP's $4B Google fuel-cell deal + 1.6 GW Mainspring agreement, a reported Google–Constellation 3.6 GW PPA, Venture Global's 10 GW island-mode gas plan, and the October STEO's regional wholesale divergence. Let me verify the Google–Constellation deal, get AEP deal details, and complete the ticker set.

The Google–Constellation deal is confirmed as the full 3.59 GW structure (890 MW uprates + 2,700 MW existing fleet, $4.3B across 11 units) — an update to the existing hyperscaler-uprates note. The AEP/POWER article gives deep detail including Black Hills' Cheyenne microgrid and AEP's 69 GW contracted load. Let me confirm the two remaining supply-side stories (Venture Global's 10 GW, Chevron–Microsoft 2.67 GW

I was interrupted mid-cycle — I have substantial confirmed evidence (FERC's Oct 9 IRAS rejection, the AEP–Google fuel-cell deal, the October STEO, and eight ticker views). Next step: finish verifying the two remaining supply-side stories (Venture Global's 10 GW island-mode gas plan, Chevron–Microsoft 2.67 MW PPA) plus the Google–Constellation 3.59 GW detail, then write the findings and close out properly.

All evidence is confirmed. Now I'll write the five findings: the FERC IRAS rejection (new), the Google–Constellation 3.59 GW deal structure (update), the dedicated-generation/behind-the-meter wave (new), the October STEO regional divergence (update), and the market's Oct 19 state of the AI-power trade (update).

First two findings saved. Now the dedicated-generation wave, the October STEO update, and the market-state update.

Three findings saved. Now the market-state update on the AI-power equity trade, then I'll record the watch events and close out.

All five findings are saved. Now recording the two watch resolutions (FERC's IRAS ruling; the BYONC question it mooted) and registering the new forward catalyst (PJM's backstop refile ~Oct 29).

Cycle complete. Summary of what this cycle surfaced and persisted: **Findings written (5):** 1. **`ferc-rejects-pjm-iras-byonc`** (new) — FERC's Oct 9 order in ER26-3515: accepted the 50-MW+ Large Load definition and registry, but rejected the IRAS curtailment framework, BYONC, and the capacity-auction exclusion of new large loads on retail-jurisdiction grounds. Jefferies reads it as "higher capacity prices for longer" for IPPs; cost allocation moves to state commissions (OH/PA). 2. **`hyperscaler-dedicated-generation`** (new) — the grid-bypass track: AEP's $4B/702-MW Google fuel-cell plant + 1.6 GW Mainspring deal, Black Hills' 2.7-GW Cheyenne portfolio ("Google bears all costs"), Venture Global's 10 GW behind-the-meter "AI utility," and Enerflex's 1 GW+ BTM deal. 3. **`hyperscaler-funded-nuclear-uprates`** (updated) — full Google–Constellation structure: 3.59 GW total (890 MW uprates 20-yr + 2,700 MW existing 15-yr), $4.3B across 11 units in IL/PA/NJ; BYONC question resolved by rejection. 4. **`eia-electricity-price-demand-outlook`** (updated) — October STEO: PJM wholesale +41% in 2026 vs Mid-Columbia −23%; residential 18.2→18.7 ¢/kWh; Q3 consumption +4%. 5. **`ai-power-trade-derating`** (updated) — as of Oct 19 the regulated AI-load names (SO RSI 93, FE 88, AEP 81, D 79) are the momentum trade into the Oct 28–Nov 5 earnings run, while IPPs bifurcate (TLN +21% 1mo, CEG −9% 5d, NRG −43% off high). **Watch housekeeping:** resolved the FERC IRAS watch (fired Oct 9) and the BYONC watch (mooted by rejection); registered a new one-shot watch for PJM's backstop refile (~Oct 29).

The research cycle is closed. This cycle's core contribution: the federal track for shielding ratepayers from data-center capacity costs (IRAS/BYONC) was rejected by FERC on Oct 9 — pushing "who pays" fights back to state commissions — while the market's answer to "who captures the upside" rotated toward regulated utilities (SO/FE/AEP/D at RSI 79–93 into earnings) and a parallel grid-bypass supply model (dedicated fuel cells, microgrids, island-mode gas) matured in Wyoming, Louisiana, and beyond. Next cycle's catalysts: PJM's backstop refile (~Oct 29), AEP's Q3 + 2027–2031 capital plan (Oct 28), and FE's Q3 (Oct 29).