Grid Alpha

PODBISHOP1UNITS settled at -$763.50/MWh in the 11:00 UTC real-time interval on Wednesday, against $69.19/MWh at BCNPPS52_SOLAR, for a nodal spread of $832.69/MWh across the CAISO footprint. The timestamp is the first thing to notice. 11:00 UTC is 04:00 Pacific, before sunrise, so whatever pushed Bishop that deep in this interval, it was not solar oversupply.

The decomposition makes the point cleanly. The system energy component was $31.72/MWh at both ends of the spread; Bishop's print is a -$794.43 congestion component with a -$7.39 loss term stacked on it, while the BCNPPS solar nodes carried a +$30.98 congestion component and a -$0.10 loss. Four nodes in the Bishop cluster shared the identical -$794.43 congestion figure: PODBISHOP1UNITS, PODCONTRL1POOLE and PODCONTRL1QF all at -$763.50, and PODBISHOP1ALAMO at -$762.92 on a marginally smaller -$6.80 loss component. Identical congestion across a cluster is the signature of a single binding element behind those nodes, and system conditions offer no cover for it: load was 23,163 MW, net load 19,502 MW, gas was the marginal fuel, and THNP15GEN cleared at $36.90/MWh, close to the energy component. The constraint itself is unnamed. The snapshot's congestion table lists PCE1 GOSHEN-IPCO KPRT345, GRACEONEIDA1385 and SHERIDANYELLOWTA_230 with no shadow prices, and those names sit outside the CAISO balancing area, which reads like a mislabeled feed rather than a diagnosis.

The supporting data is thin enough to say so plainly. AVA_DRYCREEK printed $0.00 on every component and still made the bottom-five list, which looks like a null rather than a price. The seven-day curtailment feed logged 1,668 events, but the displayed sample is dated 2026-09-09 with near-zero MW, so it says nothing about Tuesday night or Wednesday morning. A 48-hour public RSS sweep returned nothing on Bishop, CAISO congestion, or any outage. Forced outage, planned derate, local must-run condition, or data artifact all remain live explanations, and the material rules out none of them.

The angle's midday question is the right one, but it needs a baseline first. If PODBISHOP1_UNITS still shows a congestion component at or below -$794.43 at 12:00 Pacific on 2026-09-16, the constraint is structural for the day and the solar ramp deepens it rather than causing it. If the same node prints a congestion component at or above $0.00 in the 11:00 UTC interval on 2026-09-17, treat Wednesday's print as a single-interval event and discount the persistence thesis. Separately, watch BCNPPS: five identical $69.19 prints with +$30.98 congestion in a pre-dawn interval means the solar cluster sits on the constrained side of an element while producing nothing, and if that premium goes negative at 12:00 Pacific once the units are on, the spread compresses from both ends rather than one.

> A -$794.43 congestion component at 04:00 is a wire problem, not a solar problem, and until CAISO names the wire the spread is a fact without a cause.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

A $351.54/MWh gap between the top and bottom of the PJM real-time stack opened in the 10:45 UTC interval, 06:45 ET on a Tuesday morning. EMUNI 69 kV printed $334.67/MWh; the Sanderson 13.8 kV bus printed -$16.87/MWh. The RTO hub at 06:55 ET sat at $29.52/MWh, which says this is a wires story, not a fuel story.

The decomposition makes that explicit. Both ends of the spread carry the same $28.23/MWh system energy component, so nearly the whole $351.54 is congestion plus a small loss term: $305.63/MWh of congestion at EMUNI against -$44.15/MWh at Sanderson, with losses of $0.81 and -$0.95 respectively. The high side is not a single stray bus. EMUNI 69 kV, EMUNI 12 kV G1, EASTON and the DPLEASTON residual aggregate all printed the identical $334.67, which points to one electrical pocket behind one binding element rather than a data glitch. The low side is equally uniform: Sanderson SAN1, SAN2 and SAN3 at 13.8 kV plus IMPATIE and WEBSTER at 138 kV all cleared at -$16.87. What the feed does not give is the element. The three constraints listed at 07:02 ET, COOPERPE 230 kV, BERLNLK 138 kV and WMOULTON 69 kV, all show shadow price as unavailable, so nothing in the pack ties either pocket to a named flowgate.

Context around the print is unremarkable. System load was 85,855.9 MW at 07:00 ET with net load of 79,465 MW and gas as the main source. The outage feed logged at least 147 anomaly events over the trailing seven days, but the sampled rows all carry the same 2026-09-09 04:00 stamp with forced outages ranging 2,740 to 4,392 MW and maintenance 3,689 to 6,556 MW, so the series looks like duplicates or revisions and cannot be read as a trend. The one piece of adjacent news, PJM's September 14 proposal on large-load ride-through after nearly 4 GW of data centers dropped when a line tripped in July, concerns disturbance response, and no source connects it to this morning's spread. Whether EMUNI and Sanderson sit on the same constraint, or whether the spread outlasted a single five-minute interval, is not in the data.

That is why the basis-risk framing in the angle holds and a fundamentals call does not. Two checkable markers. If the EMUNI 69 kV congestion component still exceeds $305.63/MWh at the 09:00 ET interval, the pocket has survived the morning ramp and any FTR or physical position across the EMUNI/Easton nodes is carrying a persistent charge, not a transient. If any of the three listed constraints posts an actual shadow price above $0 by 08:00 ET, the driver becomes nameable and the setup moves from watch to analysis; if all three remain n/a while the hub holds near $29.52, the gap lives entirely in one unidentified element and the DART settlement at these nodes is the only place the exposure resolves.

> A $351.54 spread with $28.23 of energy on both ends is congestion wearing a price tag with no name on it yet.

Not investment advice. For informational purposes only.

Yesterday's tape: NYISO N.Y.C. lmppeak >= 91.78 — verified (observed 356.19); NYISO N.Y.C. lmppeak <= 91.78 — not triggered (observed 356.19).

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

N.Y.C. real-time LMP printed $433.48/MWh at 23:30 UTC on Sunday, 19:30 local, against a $47.75/MWh mean over the 23.3-hour window that closed at 10:50 UTC Monday. The window low was $32.40/MWh at 13:10 UTC Sunday morning. The peak-to-mean multiple in the dek is arithmetic on those two prints, not a figure any feed reports, and no source in the pack names a driver for the spike.

The open is a range, not a point. The hub series ends at $91.78/MWh at 10:50 UTC; the live snapshot stamps N.Y.C. at $48.24/MWh for 11:00 UTC, ten minutes later. The two feeds disagree and neither print is flagged as a settled interval, so "collapses to $92" holds on one feed and "already back at the mean" holds on the other. System load stood at 17,417.3 MW at 10:45 UTC with net load at 16,658 MW, and Dual Fuel was reported as the main generation source at that snapshot; that is a fleet-mix reading, not a statement about which unit set price. The constraint list carries CHESTROR-SHOEMAKR 138 kV, SCRIBA-VOLNEY 345 kV and the SCH-PJ-NY interface, all with shadow prices unavailable. Across 280 five-minute bars the 23:30 UTC bar stands alone, and the tape shows only that it happened and unwound, not why.

The background is not quiet. A drone strike shut Saudi Arabia's East-West pipeline in the week of 12 September, taking roughly 5 million b/d offline, about 5% of global supply, and Brent moved above $104/bbl with US diesel at record highs, per Electrek. Whether any of NYISO's dual-fuel units switched to oil sits in the open questions, and nothing in the pack connects the crude move to Sunday's evening bar. What it does establish is that the oil-side cost reference for that fleet has moved, which matters if a repeat ramp forces fuel-switching decisions into the offer stack.

The angle frames Monday as a repeat-or-revert test, and the checks are concrete. If the N.Y.C. real-time bar at 23:30 UTC Monday, 19:30 local, prints above $91.78/MWh, the higher of the two disputed open prints, the evening ramp has repeated at a level above anything the morning tape offered; if every bar through 23:30 UTC stays below $91.78/MWh and the day's mean sits near $47.75/MWh, reversion has won and Sunday's bar is a one-off. Separately, if the 11:00 UTC $48.24/MWh print is confirmed at settlement while the 10:50 UTC $91.78/MWh print is revised, the "collapse to $92" framing was a feed artifact and the session opened at the mean. On mechanism, if CHESTROR-SHOEMAKR 138 kV or SCRIBA-VOLNEY 345 kV shows a nonzero shadow price in the 23:30 UTC interval, Sunday's spike has a candidate cause worth pricing into the evening; if all three constraints still read n/a at that hour, the spike stays unexplained and the cleaner comparison is the 23:30 UTC bar against the $47.75/MWh window mean rather than against a constraint story nobody has evidenced.

Until one feed is confirmed as the settled open and one constraint shows a price, the trade is the 23:30 UTC bar itself, not a narrative around it.

> A $433 bar with no named driver and a $92-or-$48 open is a data problem before it is a scarcity problem; Monday's 23:30 UTC print decides which.

Not investment advice. For informational purposes only.

Yesterday's tape: PJM WESTERN HUB lmppeak >= 86.43 — verified (observed 816.58); PJM WESTERN HUB lmppeak <= 37.18 — not triggered (observed 816.58); PJM WESTERN HUB lmp_peak <= 28.92 — not triggered (observed 816.58).

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

Western Hub printed $591.38/MWh in the 23:45 UTC five-minute interval on Saturday, 19:45 local, and opened Sunday at $37.18/MWh. The 23.8-hour window mean sits at $86.43/MWh, which tells you one evening print did most of the work on an otherwise soft tape. The window low was $28.92/MWh at 14:00 UTC Saturday, a mid-morning trough, so the hub travelled more than an order of magnitude inside a single afternoon and evening.

What the feed does not say is the story. No driver is named. The three constraints listed as top congestion at the 11:00 UTC Sunday snapshot, BERWICK 69 kV, GLENBROO 69 kV and the Chicago-Praxair3 138 kV contingency on Wilton Center-Dumont 765 kV, all carry shadow prices reported as n/a, so nothing confirms any of them bound during the spike. The outage anomaly feed logged 147 events over seven days, but the five sampled rows are all stamped 2026-09-07 04:00, with forced outages ranging 2,685 to 3,303 MW and maintenance outages 3,253 to 5,936 MW. Five identical timestamps reads as a data-quality question, not a current outage picture, and it says nothing about what was offline at 19:45 Saturday. System load at 11:00 UTC Sunday was 84,391.82 MW against net load of 82,784 MW, with gas the main generation source. A Saturday evening in mid-September is shoulder season: solar rolling off, load modest, and the supply stack that clears the ramp only as deep as the units committed for the day. That is the mechanical setting in which a single interval can decouple from load, but the material offers no confirmation that this is what happened. The morning basis is also flat: Western Hub at $37.18/MWh at 10:50 UTC against PJM-RTO at $34.13/MWh at 10:35 UTC leaves no residual congestion premium at the hub, which argues against a constraint that persisted overnight.

One adjacent item worth a line: on 2026-09-11 the D.C. Circuit vacated the DOE order that had kept the J.H. Campbell coal plant in Michigan from retiring, finding states hold primary authority over generation and resource adequacy. Campbell is a MISO unit. Whether any PJM units sit under comparable orders, and whether the ruling touches their status, is not answered in the source material and belongs on the week's question list rather than in today's price.

The angle already frames the 23Z hours as the priced risk, so the job is to make that checkable. If the Western Hub 23:45 UTC interval on Sunday prints above the $86.43/MWh window mean, the evening-ramp thinness is repeating and the Saturday print was not a one-off. If instead the 23:45 UTC interval clears below $37.18/MWh, Sunday's open, the spike was idiosyncratic and the weekday setup should be priced off the RTO tape. Separately, if any of the three listed constraints shows a reported shadow price at the 11:00 UTC snapshot on Monday rather than n/a, that is the first hard evidence of a congestion driver, and the hub-to-RTO basis at 10:35 UTC Monday should widen from the near-flat $37.18 versus $34.13 read if it does. A Monday 14:00 UTC print back near the $28.92/MWh Saturday low would confirm the midday trough is structural and the risk remains concentrated in the evening ramp.

> One five-minute print at $591 is a question, not a regime; until a constraint or an outage shows up with a number attached, the evening ramp is priced risk without a confirmed cause.

Not investment advice. For informational purposes only.

Yesterday's tape: SPP SPPNORTHHUB lmppeak >= 47.61 — verified (observed 223.21); SPP SPPNORTHHUB negativelmp <= 0 — verified (observed -9.49).

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

SPP North Hub real-time LMP cleared $1,135.99/MWh at 00:20 UTC on September 12, against a $47.61/MWh mean for the 23.5-hour window ending 10:50 UTC. Two hours and five minutes later, at 02:25 UTC, the same hub printed -$20.23/MWh. By 10:50 UTC the series feed showed $15.66/MWh, and the live snapshot showed $16.55/MWh at 11:00 UTC; the two feeds sit ten minutes apart and the gap is unexplained.

The series behind those numbers is 282 five-minute bars running from 11:05 UTC September 11 to 10:50 UTC September 12. In local time, the peak landed at 19:20 CDT Friday and the trough at 21:25 CDT Friday. That is all the feed says about the shape of the event: it reports the peak bar, not how many intervals cleared above any given level, so duration is unknown. No source gives a cause. The top three SPP constraints in the snapshot report shadow prices as n/a, so a binding-constraint read cannot be confirmed or ruled out, and whether the print was hub-wide or node-driven is equally open. Whether the day-ahead market priced those hours anywhere near the real-time outcome is also not in the data.

What the snapshot does show is the system state this morning, not at the spike. At 10:50 UTC, SPP load was 32,553 MW and net load 15,580 MW, with wind the main generation source. The anomaly feed logs at least 1,173 generation-capacity-on-outage events and at least 1,943 variable-energy-resource curtailment events over the last seven days, but both counts are lower bounds from truncated samples and carry low confidence. The sampled outage rows are dated 12:00 on September 5 and show coal on outage between 1,392 MW and 1,996 MW plus 249 MW of hydro; the sampled curtailment rows from 11:05 to 11:25 the same day show zero solar curtailment. None of that is dated to Friday evening, and no public RSS item in the 48-hour lookback mentions SPP or the print. Third-party confirmation does not exist yet.

The angle frames this as repeat risk rather than a one-off, and the honest position is that the data cannot yet distinguish the two. So the test is whether the same clock window reproduces the same shape. If North Hub real-time LMP clears above the $47.61/MWh window mean in any five-minute interval between 00:20 and 02:25 UTC on Sunday September 13, the event has a second data point and the repeat thesis strengthens. If net load at 00:20 UTC Sunday prints below the 15,580 MW seen at 10:50 UTC today with wind still the main source, note the low-net-load condition, but do not connect it to the spike without a cause on record. If any of the three listed constraints posts an actual shadow price instead of n/a at 00:20 UTC Sunday, the congestion channel moves from unknown to testable. And if the hub goes negative again by 02:25 UTC, the full round trip repeated, which matters more for DART exposure than the peak alone.

Until one of those checks fires, the $1,136 print is a single bar with a timestamp and no attribution. Trade the uncertainty, not the story.

> A $1,136 peak and a -$20 trough two hours apart is a shape worth watching for tonight; it is not yet a mechanism anyone can name.

Not investment advice. For informational purposes only.

Yesterday's tape: MISO NIPS.CC.WHITN negative_lmp <= 0 — verified (observed -113.01).

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

$627.68/MWh. That was the gap between NSP.CMMPA.JANE at $529.26/MWh and NIPS.CC.WHITN at -$98.42/MWh in MISO's 10:55 UTC real-time interval on Friday, 05:55 local time, while ILLINOIS.HUB cleared a sleepy $43.31 at 11:00 UTC. Whatever is binding, it is not the energy stack.

The system energy component was $34.82/MWh at every node on both ends of the ladder, so the entire spread lives in congestion and losses. Jane's price is $497.60 of congestion on top of $34.82 of energy, with a loss term of -$3.16. The Whiting complex is the mirror image: NIPS.CC.WHITN, WHITNGT1, WHITNGT2 and WHITNST all cleared at an identical -$98.42 with -$134.15 congestion, which is what four units behind one constrained element look like. Between the extremes sits a second high cluster, NSP.OWATONSIG and NSP.OWATONA7 at $240.09 and ALTW.MMPA.OGWF at $239.26, all sharing a $208.22 congestion component, with NSP.KASOTA at -$86.93 on the other side. The NIPS prefix even straddles the ladder: NIPS.BAILY.ARR was fifth-highest at $223.90 with $187.78 of congestion while the Whiting nodes were the floor.

What the feed does not give is the constraint. The snapshot lists SIBLEY-ORRICK, PRARCK-CALAMUS and CHICAGO-PRAXAIR3 as the top flowgates but reports no shadow price or direction for any of them, so none can be tied to Jane or Whiting. The anomaly log shows at least 5,396 real-time binding-constraint events and at least 84 generation-outage events over the past seven days, but the rows actually displayed date from September 4 and September 5 and name constraints elsewhere in the footprint. The angle's north-south read fits NSP nodes at the top and NIPSCO units at the bottom, but the source supplies no geography or paired constraint to confirm it, and Baily's position on the high side argues against a clean regional split. System conditions were unremarkable: 76,612 MW of load, 62,121 MW of net load, coal the lead fuel.

Single-interval prints this size either decay or explain themselves within hours, so the test is persistence. If Jane's congestion component still exceeds $208.22/MWh, the level the Owatonna cluster printed, at 08:00 CT (13:00 UTC), the constraint is structural rather than a dispatch transient, and the day-ahead result for the Jane node becomes the tell for whether the DART is paying for it. If the four Whiting nodes still clear below zero at 12:00 CT (17:00 UTC), the element is holding through the midday load build and the shadow price on whichever flowgate MISO eventually posts is the number that sets the size of the spread, not load. Until a shadow price attaches to a named element, the Jane and Whiting pair is a pattern, not a driver.

> When energy is $34.82 everywhere and one node prints $529.26, the story is the constraint nobody has named yet, not the stack.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

PJM's real-time market printed a $234.14/MWh intranodal spread at 10:25 UTC Thursday (06:25 ET), with MACNEW_T 115 KV at the top at $169.49/MWh and YORKANA 13 KV LOAD at the bottom at -$64.65/MWh. The system price tells you this is not a system story: PJM-RTO hub sat at $66.34/MWh in the same interval, essentially identical to the $66.01/MWh energy component common to both extremes.

The decomposition is clean. MACNEW's congestion component printed $102.54/MWh; YORKANA's printed -$130.62/MWh. Losses are rounding error on both ends ($0.94 and -$0.04). So the entire $234 spread is congestion, a pure transmission story in central PJM. And YORKANA is not alone at the bottom: REDFRONT, PROSPEME, WHELLAM and YOE all printed between -$60.09 and -$64.32/MWh, each carrying congestion components near -$126 to -$130. That is a coherent negative-price pocket across the 115 KV and 13 KV network around York, not a single-node artifact.

Context matters for persistence. System load was 100,348 MW at 07:00 ET with net load at 98,490 MW and gas on the margin, a solid but not stressed morning. The outage backdrop is heavy: PJM logged at least 144 daily outage events over the prior week, with forced outages sampled as high as 13,317 MW on September 4. Whether any specific derate is feeding the central-PJM pocket is unconfirmed, and the listed constraints (ENDCAVRN, HARWGATE, GLENDON lines) returned no shadow prices in the snapshot, so the binding element is not yet identifiable from the data. What is identifiable: negative LMPs at load nodes this deep mean generation behind the constraint is being paid to back down, or more precisely, being charged to run, while MACNEW-side supply clears at a $103 premium to the energy component.

The timing is the trade. This printed at 06:25 ET, before the morning ramp has fully built and hours before the afternoon peak. If the binding constraint holds as load climbs toward the evening ramp, the MACNEW congestion component has room to widen past $102.54/MWh and the YORKANA pocket can go deeper negative; if the constraint clears with the morning topology change, the spread collapses back toward hub levels fast.

What I'd watch: the MACNEW_T congestion component at the 13:00 ET interval print. If it still exceeds $50/MWh at that reading, the constraint is structural for the day and afternoon DART positions against the YORKANA pocket stay live. Second check: YORKANA's LMP at 16:00 ET. If it is still negative into the ramp, the pocket is load-insensitive and likely outage-driven, which argues for persistence into Friday. If both revert toward the $66 hub print by 13:00 ET, this was a morning-topology event and the trade is done. Watch PJM's constraint postings for any of the three listed 115/230 KV lines picking up a shadow price; that names the driver.

> A $234 spread with a $66 hub is not a market signal, it is a map of one broken path in central PJM; the afternoon ramp decides whether it is a wall or a door.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

ERCOT's WEST weather zone posted +23.0% weather-matched year-over-year load growth over the last 90 days, with an interquartile range of +18.3% to +27.7% across 1,982 matched hour pairs. Awarded CRR MW on paths sourcing or sinking at a West hub or the West load zone went the other direction: 3.6% of total MW in the October 2024 through September 2025 delivery window, 2.9% in the year that followed.

The load number survives the obvious objection. The matching method paired each hour against five year-ago hours within 2°C, producing 15,937 pairs across eight weather zones, and the correction bites elsewhere: NORTHCENTRAL's naive +13.6% collapses to +6.8% once temperature is stripped out. WEST is the outlier by a wide margin, with SOUTHERN next at +8.6%, then SOUTHCENTRAL +6.1%, EAST +6.0%, COAST +2.4%. FAR_WEST was flat at +0.3%, consistent with loads already energized and price-responsive curtailment in both windows. The WEST zone holds the Abilene Stargate campus at roughly 1.2 GW announced scale plus wind-corridor large loads. One caveat travels with the study: NORTH printed -10.9% and the authors flag it as an unexplained anomaly, possibly zone boundary or metering reallocation, which means some of WEST's gain could be accounting rather than new electrons.

The auction side shows dispersion, not withdrawal. Active bidding families were 210 then 212, with 20 entrants and 18 exits. HHI sat near 209 in both windows and the largest family, XDCEN, holds 7.0% of awarded MW on a 3% sticky-path share; eight of the top ten families sit at 0-9% sticky. The fastest riser, XDARB, went from 0.2% to 2.4% of awarded MW. An earlier GridAlpha study of the 2023-01 through 2024-03 block put core hedgers near 62% of MW, though the sticky definitions differ enough that the drift toward opportunistic bidders is directional only.

What I would watch is whether the share decline is absolute or arithmetic: if total auction volume grew, 2.9% can still be more West MW than 3.6% was, and that distinction decides whether the under-owned framing holds. Track the next monthly CRR auction results posting for West-path awarded MW in absolute terms and West-path clearing prices against the prior window; a falling share at rising clears is a very different market than a falling share at flat clears. Near term, the DA system lambda forecast for 2026-09-09 frozen pre-DAM runs $19 to $60/MWh with a $31 mean, and the elevated-risk hours flagged on the seven-day strip are HE20 and HE21 on both 2026-09-10 and 2026-09-11: if realized WEST zonal congestion in those four hours clears above the forecast evening peak of $60/MWh, the positioning gap is being priced in real time rather than in the auction. The model's trailing 30-day MAE is $8.1/MWh against a $10.6/MWh naive baseline over 684 scored hours, so a $60-plus print in those hours is outside routine forecast error. Gas is not the constraint here: the Texas Eastern Lebanon compressor outage is complete and the Nexus operational flow order lifted 9/8.

> A zone growing 23% a year while its congestion hedges shrink to 2.9% of the auction is either the cheapest exposure in ERCOT or a measurement error, and the next auction print tells you which.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

CAISO's real-time market printed an $899.33/MWh nodal spread at 10:05 UTC Wednesday, 03:05 local. COLWATER1N011 cleared at $734.03/MWh while AEC3001N001 printed negative $165.29, a simultaneous scarcity-side and curtailment-side signature inside the same five-minute interval.

The decomposition is the story. System energy sat at $73.31/MWh at both ends of the spread. Colwater's print was $639.87 of pure congestion plus $6.87 of losses; AEC300's was negative $248.73 of congestion against a $3.86 loss credit. This was not a system-wide scarcity event. It was transmission. Four additional nodes, SEGS1G7B1, GALE1N001, COLWATER1N002 and COLWATER1N001, clustered at $733.74 to $733.76 with identical $639.87 congestion components, marking a single binding interface isolating the Colwater pocket. On the other side, four GWF Tracy-area nodes printed between negative $124.68 and negative $124.96 on a shared negative $207.13 congestion component: a second, separate negative-price pocket. One hour later the NP15 hub cleared $46.36 with load at 27,243 MW and net load at 24,938 MW, gas on the margin. The system was calm; the topology was not. Whether the negative prints reflect active renewable curtailment is unconfirmed: GridAlpha logged 1,668 CAISO curtailment events over the prior seven days, but sampled volumes were near zero, and no scarcity or reserve-shortage flag appears in the source data. Treat the "scarcity pricing" half of the thesis as inference, not fact.

The open question is duration. A single-interval $899 spread is noise for most books; a persistent one reprices the local basis. If the $639.87 congestion component at COLWATER1N011 still exceeds $300/MWh in the 12:00 UTC (05:00 PDT) interval, the constraint is structural into the morning ramp rather than an overnight artifact, and Colwater-area CRRs and congestion revenue rights against AEC300 and the Tracy cluster carry real value into the day-ahead run. If AEC300's congestion component is back above negative $50 by 13:00 UTC, the negative pocket cleared with the overnight trough and the trade is done. Watch whether the day-ahead market prices the Colwater cluster above $200 for tomorrow's overnight block; that is the market confirming or dismissing the constraint.

> When both ends of an $899 spread share the same $73 energy component, the grid is fine and the wires are the market.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

Michigan Hub printed $404.58/MWh at 18:30 CT Monday, the final interval of the 24-hour window, then opened Tuesday's tape at $40.11/MWh. That peak ran nearly 11x the $37.17/MWh daily mean, and the same day produced a $0.00/MWh print at 07:55 CT. Zero to four hundred dollars inside eleven hours is not a normal Monday in the MISO North region.

The dispersion is the story. A $0 floor in the morning followed by a $404 evening peak is the signature of a system long on midday energy and short on evening ramp capability. The backdrop supports the tightness read: MISO logged 84 generation outage events over the prior week, with sampled snapshots showing roughly 8,754 to 9,115 MW derated and 6,293 to 6,446 MW on forced outage. Those samples are dated September 2, so they describe the fleet heading into the week, not necessarily the stack at 18:30 Monday. On the transmission side, MISO recorded 7,246 real-time binding-constraint shadow-price events over the same seven days, with repeat appearances from the TMP293RUGBY230RUGBY115kV flowgate and the AnitaTap 161 kV element. A hub that binds that often at the margin will price scarcity violently when the evening ramp meets a thin stack.

What the data does not say matters as much as what it does. No source confirms the specific driver of the $404.58 print: no named outage, no scarcity adder, no reserve shortage declaration. There is no load, weather, or renewable output data to explain either the midday zero or the evening spike. And there is no day-ahead or same-day pricing for Tuesday to confirm whether a repeat ramp is forming. Treat the peak as an observed event with a plausible mechanism, not a diagnosed one.

The trading implication is in the setup, not the explanation. If the forced-outage stack is still running near the 6,300 to 6,400 MW range sampled last week and the Rugby and Anita_Tap flowgates keep binding, the evening ramp has the same ingredients today. Watch the Michigan Hub real-time print into the 17:00 to 19:00 CT window Tuesday: if LMPs clear $100/MWh by 18:00 CT, the scarcity repricing is repeating and the $37 mean understates where the evening settles. If the hub holds under $60/MWh through 19:00 CT, Monday's peak reads as a one-interval event and the fade to $40 was the correction, not the pause. Also watch the midday print: another $0.00/MWh hour before 09:00 CT would confirm the intraday dispersion regime is intact, which is the precondition for the evening spike, not a separate signal.

For position framing, the spread that matters is intraday, not day-over-day. A hub that can print zero and $404 in the same session is pricing ramp hours as a distinct commodity. Hourly shape products and real-time exposure around the evening block carry the risk; the daily average is noise.

> When a hub prints zero at breakfast and $405 at dinner, the daily mean is a fiction; trade the ramp hours or get run over by them.

Not investment advice. For informational purposes only.

Yesterday's tape: ERCOT HBHOUSTON lmppeak >= 60 — verified (observed 107.82); ERCOT HBHOUSTON lmppeak <= 50 — not triggered (observed 107.82).

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record