Monday, October 12, 2026

What Needs Your Attention Today

Green means watching — we are tracking it and it looks normal. Amber means warning — it needs a closer look. Red means danger — time to act. The color never carries the meaning alone: every card says it in words.

Module 1 — The 60-Second Regime Matrix

Regime: the referee split the whistle — and handed half of it to the states. Friday's FERC order (docket ER26-3515) is really two rulings in one. First, transparency won: PJM's definition of a Large Load — any end-use customer with a cumulative peak of 50 MW or more at a single electrical site, with affiliated facilities inside a one-mile radius counted together — plus the registry that forces utilities to report those loads, takes effect today, October 12. Second, coercion lost: the Interim Resource Adequacy Service (IRAS), which would have let PJM ask new large loads to reduce power use during emergencies unless they brought their own new capacity, was rejected — FERC ruled that dictating retail curtailment "intrudes on" state authority. The bring-your-own-capacity rules went down with it, as "inextricably linked."

The mechanism, in plain terms: federal regulators can count the elephants in the room, but only states can tell them where to stand. (The legal split: FERC governs wholesale power — the bulk market between generators and utilities — while states govern retail — what you and I actually pay and whether our power can be cut. The rejected plan crossed that line.) Commissioner Rosner's concurrence is the tell — he wants state regulators to write explicit rules on which customers get cut first "when circumstances require it." The doctrine from last week — DOE's October 7 position that new load must fund its own power ("build, bring, or buy") — is now colliding with fifty different state rulebooks. What would prove this wrong: PJM refiling a restructured emergency-cutback plan by the October 29 deadline that FERC accepts as wholesale-only — that would revive the federal path and deflate the state-by-state story. Until then, the fight is fifty-fifty federal-state, and the federal half just got thinner.

Rosner's concurrence is worth reading in his own words, because it is now the de facto doctrine for the state fights: he wrote that data centers not serving critical functions should be shut off before ordinary customers in an emergency — "homes, hospitals, and schools should never be curtailed before data centers" — but only after voluntary demand response is exhausted. (Demand response, for the non-utility reader: customers who already get paid to voluntarily use less power during shortages — the grid's pre-booked volunteers.) The sharp edge: PJM's filing, he wrote, "flips this paradigm" by forcing the new large loads off the system before the customers already paid to volunteer for cutbacks. The commission itself said it does not take such decisions lightly — and it gave PJM guidance on how to refile. Meanwhile the states are already moving: Pennsylvania's governor has directed regulators to seek rules that curtail data centers first unless they secure their own capacity. Add the live context: PJM issued its first-ever Capacity Advisory on October 9 for October 12–13, forecasting ~102,000–104,000 MW peaks against ~70,000 MW of scheduled outages in warm fall weather — the shortage mechanics are real, today, not theoretical. Falsifier: watch the Oct 22 PJM member meeting — if members converge on a restructured federal plan, the state-by-state thesis weakens.

Fresh reporting sharpens what Friday's order does and doesn't settle. Utility Dive (Oct 12) notes the registry FERC approved is genuinely new load intelligence — each facility's location, ramp schedule, capacity supply and other details, all reportable — while the curtailment framework fell on two grounds: it intrudes on state retail-rate jurisdiction and "appears to discriminate against large loads." Rosner's most load-bearing point goes to market design: excluding the curtailed loads from the capacity auction's demand curve, he wrote, produces "the result could be a capacity market signaling that PJM has more than enough capacity while the region sheds thousands of megawatts of demand on an involuntary basis. That's not only economically inefficient, but also likely to prevent supply and demand from coming back into balance." (Utility Dive, quoting his concurring statement.) In plain terms: a rigged demand curve would let the auction price signal abundance while the grid sheds gigawatts by force — the opposite of a functioning market. One market-structure footnote: demand-response aggregator Voltus had backed the rejected plan in September comments (The Duck Curve), so the commercial demand-response industry was on the losing side of Friday's order.

One more piece of live machinery — with a correction to this morning's read. PJM's one-time Reliability Backstop Procurement — the auction meant to close the 6,831 MW capacity shortfall left by July's 2028/29 auction — did not open its bid window as planned on September 30. After FERC's September 29 partial approval found the cost-allocation, transmission-owner exit, and collateral rules "may be unjust and unreasonable," PJM delayed the auction; spokesman Jeffrey Shields told Utility Dive a new timeline has yet to be determined. (This morning's edition carried a conflicting trade-press claim that the auction is "open through October 21" — that claim is now corrected.) What still stands: the pay-as-bid design capped at $555/MW-day — nearly 70% above the $325 cap on last summer's base auction — with results originally targeted for December 2, now unverified pending the refile. The arithmetic, labeled as arithmetic: at the cap, the full backstop would cost roughly $3.8 million per day — and the Natural Resources Defense Council estimates the auction could pay out up to $20 billion in total for new power plants. Scale context, for the scale of the problem: PJM serves ~67 million people across 13 states and DC, failed to clear enough capacity in its last two base auctions, and estimates data-center and large-load demand could grow by up to 70 GW by 2038. Friday's orders (FERC's state-authority ruling, DOE's "build, bring, or buy" position) are an argument that this bill lands on the large loads that created the shortfall, not on households — and the auction's eventual price print is now the most important unresolved data point in American power markets. Falsifier: if PJM's refile lands cleanly by Oct 29 and the auction opens on a firm new schedule, the delay is procedural, not structural.

The DOE layer got sharper this week too. Its October 7 statement of position at FERC asks PJM to file a revised backstop proposal by October 29 — answering FERC's September 29 ruling that the auction's cost-allocation framework needed fixing — and explicitly agrees with FERC that backstop costs should be allocated on updated load forecasts, so the zones adding the load absorb the cost. PJM told Utility Dive it intends to file by that date rather than wait out the five-month hearing process FERC had left open. That's the mechanism of "build, bring, or buy" becoming real: the federal government is using a filing deadline, not a subsidy, to force the money question. Then the macro fuel behind the whole fight, from Bloomberg's October 6 reporting: global investment-grade and high-yield AI-infrastructure bond issuance is already ≥$360 billion year-to-date — 5.8% of all global bond sales, up from 1.9% last year — and SpaceX is in early talks to raise ~$40 billion ($10B in bank loans, $30B in investment-grade debt) to buy Nvidia chips, with Apollo leading and an expected 2027 close (talks could still end with no deal). Nvidia's chip-buying coalition pledged, per Bloomberg, to "create dedicated pools of capital at significant scale at attractive rates for Nvidia customers." For context on what those chips become: xAI's Colossus 2 cluster in Memphis already runs 110,000 Nvidia GB200 and 440,000 GB300 chips — and every rack of them is a large load PJM must now count.

Module 2 — The Macro Vector: Gas First, Reactors Later

The most concrete new power announcement of the weekend came from Texas — and its timeline is the story. Blue Energy Global asked Texas's grid operator, ERCOT, for permission to plug in up to five 300-MW small nuclear reactors at the Port of Victoria — the first-ever grid application for the smaller, factory-built reactor type, reported by Bloomberg. But the project's staging is what matters: ~1 GW of gas-fired generation is slated for 2030–2031, while the first nuclear unit is targeted for about 2033 — a decade-long gas bridge before the first atom splits for the Crusoe AI data-center campus at the site. Total cost: above $6 billion. Blue Energy already filed the first portion of a federal construction permit application in September.

This is the new industry's honest architecture: sell gas today to afford reactors tomorrow. The analogy, with its break stated: it resembles the 1990s wireless buildout, when carriers sold today's voice minutes to fund tomorrow's data networks — except here the bridge fuel emits carbon for a decade, and the "tomorrow" asset (BWRX-300) has yet to generate a single commercial kilowatt-hour anywhere. Note the parallel context: Bloomberg also reported that Texas has ordered a temporary pause on new data-center projects amid public backlash — the queue is lengthening even as the first nuclear applications arrive. (That pause is Bloomberg's reporting, not a confirmed order text in hand — treat it as reported, developing.) What would prove this wrong: a financed, permitted BWRX-300 project with a signed construction contract before 2030 — one real order would collapse the gas bridge and validate the developer story. Until then, the megawatts that matter are the gas ones.

The new detail since filing week: Blue Energy is a Maryland-based startup founded in 2023; it holds an option on ~70 acres at the Port of Victoria, with local economic-development reporting putting the opening phase near $2.4B of the $6B+ total. On the federal track, the NRC Part 1 filing landed September 15 — a limited work authorization request plus environmental report, with the full construction permit due no later than about March 2028. ERCOT's own data confirms the gas turbines (two GE Vernova 7HA.02s) are slated for 2030–31, and Blue Energy has separately applied to interconnect those too. The anchor customer: Crusoe, whose partnership with Blue Energy dates to 2025 for a 1.5 GW AI data-center campus on a 1,600-acre site in Calhoun County. And for the skeptics' column: no BWRX-300 — or any small modular reactor — has ever generated a commercial kilowatt-hour in the United States; a handful run in Russia and China, and Canada's first is still years out. The closest domestic peer is X-Energy's project with Dow at Seadrift, Texas (four 80-MW units, environmental review complete in May). So "first-ever SMR grid request" is real — and it is also just the first piece of paper.

Federal paper arrived too. On September 16 Blue Energy announced it had filed the first part of its NRC construction-permit application — a limited work authorization request (permission to begin certain early construction activities at the Victoria site) plus the environmental report — filed September 15, with the full permit due no later than about March 2028; the NRC confirmed receipt in an October 8 Federal Register notice. The remaining portion lands after one calendar year of site-specific data collection (per GE Vernova, which signed an engineering, licensing, and safety-analysis agreement with Blue Energy on August 13), and the whole project remains subject to a final investment decision in 2027. Translation for the skeptic: an interconnection request plus a Part 1 filing is still two full approvals, one FID, and roughly a decade from atoms. Falsifier: a 2027 FID that green-lights construction would collapse the gas-bridge thesis — until then, the 2030–31 gas turbines are the only contracted megawatts in the story.

Module 3 — Dispersion & Anomalies

The past week showed what happens when weather, outages, and grids interact. In Southeast Europe over the weekend, Serbia's day-ahead benchmark (wholesale power sold for the next day) fell nearly 25% to €181.02/MWh for Saturday delivery as weekend demand eased — but Romania (at €233.88/MWh), Bulgaria (€233.40) and Montenegro (€232.04) barely budged. The reason is mechanical, not market-y: Romania is running without its two Cernavodă reactors (~1.36 GW of always-on generation), and Bulgaria's Kozloduy-6 (~1 GW) went into scheduled maintenance on October 8 while unit 5 is constrained by low Danube water levels. Weather gives the grid weekends off; missing reactors don't.

The thread: the grid's binding constraint is rarely generation technology anymore — it is the queue, the permit, the transmission wire, or the statehouse. The megawatts exist; the agreements to move them don't.

Europe's gas mechanics, deeper: IEEFA's lead energy analyst, Ana Jaller-Makarewicz, puts it plainly — record-low storage leaves Europe "with less of a buffer" against global disruptions and "vulnerable to price spikes." One reason storage is so low: high prices all summer rewarded traders for selling gas rather than injecting it, so the cushion never got built. On October 3 the EU sat at 72.4% full vs 83% a year earlier; the Strait of Hormuz closure has trapped roughly a fifth of the world's LNG supply (most of Qatar's still offline); and the January 1 Russian-LNG ban removes ~19% of the EU's LNG imports (H1 2026 share) mid-winter — replacing those volumes is estimated to cost Europeans an extra €3B, 12% more than last year. ENTSOG's winter scenarios: a cold winter with optimal LNG supply still leaves storage at 11% (the strategic reserve) and requires demand cuts of ~7% (22 bcm); tight LNG pushes that to ~12% (40 bcm). Priced in: price spikes on any cold snap; falsifier: an early, warm autumn that lets refill catch up — unlikely at the current ~2.3 TWh/day pace.

The official stress test, in ENTSOG's own numbers: the grid operators' Winter Supply Outlook 2026/27, published this week, puts EU storage at 72% full on October 1 — about 811 TWh (~74 bcm), versus 83% (~940 TWh, ~85 bcm) a year earlier. Europe's cushion against a cold winter is roughly 130 TWh thinner than last year's. The counterweight: Europe now holds ~1,600 TWh of LNG regasification capacity per winter season (~145 bcm), which ENTSOG says can "partially compensate" — but if LNG availability is limited, storage could fall "well below 30%" by winter's end, wrecking both the late-winter cushion and next summer's refill. The European Commission (Oct 8) calls the system "sufficiently flexible" while conceding that "the decision whether to import LNG or withdraw gas from storage ultimately lies with the market" — i.e., with prices. And prices are doing their job: energy analyst John Kemp notes the highest prices in nearly four years have already forced consumption cuts and pulled LNG cargoes toward Europe, with storage possibly still gaining ~35 TWh before drawdown begins (peaking near 847 TWh, ~74%). The honest mechanism, from IEEFA: summer gas prices sat above winter forward prices, so traders had no commercial incentive to inject — Europe's thin cushion is a rational market outcome, not an accident. Falsifier: a cold December with tight LNG would force the 7% demand cuts IEEFA modeled, and they'd land on industry first. If you heat a home in Europe: this is the year fixed-rate energy plans earn their keep — the cushion that used to smooth winter prices simply isn't there.

The UK's wind-collapse story deserves its own footnote — the system worked. The preparedbritain.co.uk chronology of the October 6 margin notice shows the signal doing its job: NESO's projected gap shrank from 1,880 MW at 6:45am to 679 MW at noon to 321 MW at 2:15pm, and the notice was cancelled at 3:45pm, before the 4pm window even opened. This was the sixth such notice since June — most are answered by the market and stood down, which is the design, not a symptom. NESO's early view of winter 2026/27 (published Oct 9) gives a de-rated margin of 5.5 GW, an 8.8% buffer over expected peak demand. The read for the investor: margin notices are a feature of a wind-heavy grid, and the £400/MWh print is the price of the market buying its own insurance. Falsifier: if notices start standing (uncancelled) into the 3–11pm window, that's when "routine" becomes "structural." And if none of this is your job: none of it means blackouts are coming — it means a wind-heavy grid now buys its insurance in public, and that insurance eventually shows up in your bill.

Module 4 — What to Watch

• FERC accepts PJM large-load definition and registry, rejects interim curtailment service — Headwaters Wire, Oct 9. Registry and 50 MW definition effective Oct 12; IRAS and bring-your-own-capacity rules rejected as intruding on retail rate regulation.
• FERC rejects PJM's large-load curtailment plan, keeps the load registry — The Duck Curve, Oct 10. Independent take: the plan to cut new loads off early in a shortage did not survive; the order itself (docket ER26-3515) is the primary source.
• FERC rejects most of PJM's large load interconnection plan — Utility Dive, Oct 12. The curtailment framework "intrudes on state jurisdiction" and "appears to discriminate against large loads"; Rosner's concurrence warns a rigged demand curve would signal "more than enough capacity" while the grid sheds gigawatts by force.
• DOE presses PJM on ratepayer protections from large load costs — Utility Dive. DOE's Oct 7 statement asks PJM to file a revised backstop plan by Oct 29; PJM intends to file rather than wait out the five-month hearing process.
• PJM delays backstop procurement auction after FERC only partly approves plan — Utility Dive. The auction planned for Sept 30–Oct 21 was put on hold; PJM says a new timeline has yet to be determined.
• ENTSOG publishes its Winter Supply Outlook 2026/27 — ENTSOG. EU storage 72% full Oct 1 (~811 TWh) vs 83% a year ago; ~1,600 TWh/season of LNG regas capacity can partially compensate; tight LNG could push storage "well below 30%" by winter's end.
• Request to begin work at Texas gas-to-nuclear project puts Blue Energy in rare company — Blue Energy, Sept 16. First part of the NRC construction-permit application filed for the Port of Victoria gas-to-nuclear project; final investment decision targeted 2027.
• First small nuclear reactors seeking to connect to Texas grid — Bloomberg wire (via Energy Connects), Oct 6. Blue Energy Global's ERCOT application: five 300-MW BWRX-300 units, first-ever SMR interconnection request; Texas has paused new data-center projects amid backlash.
• EU faces potential winter gas shortfall as storage hits record low — IndexBox, Oct 10, summarizing IEEFA's Oct 9 report: ~73% full, lowest October since 2011 records began; EU may need to cut winter gas demand 7% (14 bcm).
• India wastes 4% of solar power amid power shortages in July, August — Reuters (via The Hindu BusinessLine), Oct 12. CREA data: ~1.8B units of solar curtailed (4.2%) despite record demand; India has ~304 GW clean capacity with 500 GW target by 2030.
• Rooftop solar reaches new highs across Australia — pv magazine, Oct 12. 17,488 MW at noon Sunday, a new NEM record; distributed + utility solar covered a record 72.29% of demand.
• UK electricity margin notice: 1.88 GW shortfall as wind power drops — IndexBox, Oct 10, with Bloomberg/Elexon data: wind ~40% below forecast, ~25% of the year's average; 7–8pm intraday delivery topped £400/MWh.