Grid Alpha

$742.45/MWh is the maximum real-time nodal spread CAISO printed in the 10:05 UTC snapshot, with ELCAJNGT7N011 at $104.85/MWh on top and PODBISHOP1_UNITS-APND at $-637.60/MWh on the floor. That is 3:05 a.m. Pacific on a Wednesday, with system load at 25,462 MW, net load at 24,082 MW, and NP15 at $50.93/MWh. The number is real; the pairing is less so.

Decompose the two ends. El Cajon's $104.85 is $38.22 of energy, $1.52 of losses and $53.25 of congestion, and five ELCAJNGT7 nodes (N011, N001, N010, N009, N004) sit between $104.83 and $104.85 carrying the identical $53.25 congestion mark, which reads as one constraint pricing one pocket rather than noise. Bishop is the mirror: energy $37.25, losses $-8.82, congestion $-678.58. PODCONTRL1POOLE-APND and PODCONTRL1QF-APND clear at exactly the same $-637.60 with the same $-678.58, and PODBISHOP1ALAMO-APND sits at $-635.85 with $-677.90 of congestion. The entire Bishop and Control cluster is pinned behind one limit, and the dispatch is paying $-678.58 of congestion to make output there go away. The catch is the clock. The El Cajon highs are stamped 10:05 UTC; the Bishop lows are stamped 11:00 UTC, a 55-minute gap. The $742.45 headline therefore spans two intervals, and nothing in the feed shows both extremes coincident in a single five-minute run. CSADIAB7N001 at $-245.33 with $-284.40 of congestion is a third, separate negative pocket, and it does sit in the 10:05 interval.

The constraint list names YU XFMR2 A 69KV, SD1 NGILA-IVALLY 500 and YU XFMR4 A 69KV, with no shadow prices attached. SD1 NGILA-IVALLY 500 is San Diego and Imperial Valley territory and is the natural candidate behind El Cajon's $53.25, but the source does not say so, and nothing on the list points at the Eastern Sierra. The "oversupplied pocket" reading in the angle is a hypothesis; the material contains no generation, outage, or export-limit data for Bishop. The curtailment feed logs 1,668 events over seven days, but all five sampled rows show 0 MW and 0 MWh, so it confirms nothing either way. Gas is the main source at 4 a.m. Pacific, which is not a solar-flood setup. Whatever is being pushed back at Bishop is being pushed back by a wire, not by system-wide oversupply.

Three checks settle this. If PODBISHOP1UNITS-APND's congestion component is still more negative than CSADIAB's $-284.40/MWh at 13:00 UTC, the Eastern Sierra limit is structural and not an interval artifact. If ELCAJNGT7_N011's congestion component still exceeds $50/MWh in that same 13:00 UTC interval, the two pockets are live simultaneously and the spread finally earns its headline. If CAISO's binding-constraint report, out by 17:00 UTC, shows SD1 NGILA-IVALLY 500 with a shadow price at or above $53.25/MWh for the 10:05 UTC interval, the El Cajon side has its named driver; if the Eastern Sierra still has no constraint on that list, the Bishop side remains unexplained and the negative print should be treated as unrepeatable until it repeats.

> A $742 spread stitched from two intervals is a flag, not a setup; the question that matters is whether Bishop and El Cajon ever print $-637 and $104 in the same five minutes.

Not investment advice. For informational purposes only.

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

A $1,192.28/MWh gap separated the top and bottom of PJM's nodal stack in the 10:50 UTC interval on Tuesday, 06:50 EDT. ASPENSLR34.5 KV ASPNRDSP printed $733.06/MWh while DILLSBUR115 KV LD1 and LD2 cleared at -$459.22/MWh, and the PJM-RTO hub sat at $108.68/MWh in the same interval. Nothing in the system price moved; the entire spread lives in the congestion column.

The decomposition is unusually clean. Every listed node carried the same $108.23/MWh energy component, so Aspen's premium is $615.81 of congestion plus $9.02 of losses, and Dillsburg's discount is a -$568.22 congestion component against $0.77 of losses. The +615.81 figure is not unique to Aspen: ROXBURY 23 KV (two nodes), SHIPNBRG115 KV LD1 and CARL PN 23 KV LD1 all carried the identical component, printing $732.90 and $732.10/MWh. On the other side, three ASYLUM 23 KV nodes cleared at -$377.43/MWh on -$484.07 of congestion. Identical components across clusters of nodes is the signature of one binding element with one set of shift factors, not a scatter of local limits. Which element is the open question. The 11:01 UTC snapshot lists TODDSTVN115KV198C1_LN, Lake George-Tower Road 138 kV on loss of Lake George-Babcock 345 kV, and LENOX-MACNEW 115 kV as top constraints, but all three show shadow price unavailable and none is tied to the Aspen or Dillsburg nodes.

System conditions do not explain it either. PJM load was 84,643.62 MW with net load of 81,819 MW at 10:55 UTC, gas the main source, and a $108.68 hub says the RTO was not short. The outage feed logged at least 147 events over the past seven days, a lower bound given the sample marker, but reports only system-wide totals: the 2026-09-30 04:00 rows show forced outages of 3,226 to 4,982 MW and planned outages of 15,035 to 16,084 MW, with five rows sharing one timestamp and reporting different values. No element-level outage is visible, so a forced transmission outage near the path remains an inference, not a fact. The node names read as a solar interconnection on one end and a load bus on the other, which fits the angle, but the feed carries names only, no asset type or location.

The first test is persistence. If the congestion component at ASPENSLR34.5 KV ASPNRDSP still exceeds the $108.23/MWh energy component at 09:00 ET, this is a sustained constraint rather than a one-interval dispatch artifact, and the ROXBURY and SHIPNBRG nodes should still share its sign. If DILLSBUR115 KV LD1 has returned above zero by 09:00 ET, the load pocket decongested first and the 06:50 print was transient. Second, compare the day-ahead LMP for hour ending 07:00 ET at both nodes against the $733.06 and -$459.22 real-time prints; a DA/RT divergence of that size across a cluster of nodes is what FTR holders on the path will settle against. It lands the same week MISO, SPP and PJM floated a replacement for the 2004 freeze date used to portion out firm rights on transmission lines, though whether that touches this path is unknown.

> A $1,192 spread on a flat $108.23 energy component is not a price signal; it is one line, one shift factor, and a shadow price the feed still reads as n/a.

Not investment advice. For informational purposes only.

Yesterday's tape: ERCOT JUNORTHRN lmppeak >= 208.99 — not triggered (observed 48.47); ERCOT BATCAVERN negativelmp <= -21.18 — not triggered (observed 42.15).

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

$689.31/MWh separated the top and bottom of the ERCOT nodal stack in the 10:50 UTC interval, 05:50 CDT Monday. JUNORTHRN and JUNCTIONRN both cleared at $528.24/MWh; BATCAVERN and FORTMARN both printed $-161.07/MWh. HB_NORTH at 11:00 UTC was $47.57/MWh, so the hub says almost nothing about what either tail is doing.

The split is wide but thin in membership. Below the two $528.24 nodes, CFLATSUNIT cleared $334.42, KNAPPRN $312.96 and BULLCRK12 $208.99/MWh. On the low side the drop-off is steeper: after the two nodes at $-161.07, SEVNFESRN sat at $-21.18 and DCL and LARDVFTNG4 at $-15.86/MWh. Two nodes sharing an identical price at each extreme is the signature of points sitting on the same side of a single binding element, but the record does not name that element. The 11:01 UTC snapshot lists 6437_F, FORTMAYELWJC11 and COLLEJUPIT1 as the top constraints with shadow prices of n/a for all three. The FORTMA name shared between the constraint and the $-161.07 node is a string match, not a confirmed mapping. Recent SCED records describe 6437F as a 138 kV element with a listed limit of 229.3 and GREENLWEAVER1_1 as a 69 kV element with limits of 36.1 and 35.9; both are lower-voltage facilities, and nothing in the record ties either to the two tails.

System conditions give no cover story. Load was 50,445 MW and net load 43,518 MW at 10:55 UTC, with natural gas the main source. The trailing seven-day outage log shows at least 1,914 unplanned events, a count that may be truncated. The sampled entries are small and dated 2026-10-01: SANMIGL lignite forced down 66 MW to 325 MW of 391 MW, SAMSON_1 solar cut 46 MW and 52 MW from a 125 MW maximum, SALVTION wind cut 27 MW and 17 MW from 125 MW. None of those names match the five high or five low nodes. The Texas Eastern Line 15 force majeure at Tompkinsville and Danville is a Kentucky-area event with no established link to ERCOT gas supply or to this interval. The angle's West/South Texas framing is not confirmed by any source here; the node locations are simply not stated.

That leaves the tradeable question as persistence, not cause. Day-ahead prices for these nodes are not in the record, so the real-time premium over DAM is unknown and the CRR exposure the angle flags cannot be sized yet. What I'd watch: if JUNORTHRN still clears above the $208.99 that BULLCRK12 printed at 10:50 UTC when the 12:00 UTC (07:00 CDT) interval posts, the event has outlasted the overnight trough and basis for the afternoon should not be marked to the $47.57 hub. If BATCAVERN or FORTMARN is still below the $-21.18 SEVNFESRN level at 12:00 UTC, the low tail is structural for the session rather than a one-interval print. If ERCOT publishes a non-zero shadow price for FORTMAYELWJC11 or 6437_F in any SCED interval before 14:00 UTC (09:00 CDT), that is the first evidence linking the constraint list to the price tails; until then the driver stays open.

> A $689 spread with a $47.57 hub is a nodal story with no named constraint yet, and the only thing worth pricing before 12:00 UTC is whether it prints twice.

Not investment advice. For informational purposes only.

Yesterday's tape: NYISO N.Y.C. lmppeak >= 500 — not triggered (observed 85.52); NYISO N.Y.C. lmppeak <= 57.13 — not triggered (observed 85.52).

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

N.Y.C. real-time LMP hit $4,032.14/MWh at 02:55Z on 2026-10-04, which is 22:55 EDT on a Saturday night in Zone J. The 23.8-hour window mean sits at $57.13/MWh and the low was $27.10/MWh at 09:50 EDT Saturday morning. By the 06:50 EDT Sunday series end the hub had returned to $41.20/MWh, and the live snapshot stamped ten minutes later at 07:00 EDT shows $36.94/MWh; the two readings are from different feeds and ten minutes apart, so this reads as a continuing slide rather than a data conflict.

What the record omits matters as much as what it holds. The hub series is 286 five-minute bars reported as peak, mean and low only, so the duration of the excursion is unknown; a $57.13 mean sitting between a $27.10 low and a $41.20 close is consistent with a very short burst rather than a sustained scarcity hour, but the bar count above any threshold is not in the data. No source in the pack names a trigger: no generator trip, no transmission outage, no 10-minute or 30-minute reserve shortfall, no N.Y.C. locality reserve requirement binding, no scarcity-pricing declaration. The three constraints carried in the live snapshot, SPR/DUN-SOUTH, FALCONER 115 MOONROAD 115 2 and SCH - PJ - NY, all show shadow prices as n/a, so congestion can be neither confirmed nor excluded. Sunday morning the system looks ordinary: load of 13,093.8 MW at 06:45 EDT, net load of 11,739 MW, natural gas as the main source. Prints above $4,000/MWh in NYISO real time belong to the territory of reserve demand-curve pricing or a single resource setting price on a thin stack, but the cause here is unestablished and should be reported as such.

The angle frames today as the test of one-off versus repeatable, and the test is cheap to run. First, pull the NYISO real-time 5-minute history for 22:00 to 23:30 EDT Saturday by 12:00 EDT Sunday: if more than a single interval sits above $1,000/MWh, the event had duration and a reserve-side explanation becomes more likely; if it is one bar, the mean already told you that. Second, if any N.Y.C. five-minute interval between 22:00 and 23:59 EDT Sunday prints above $500/MWh, the overnight setup repeats on a second low-load weekend night and the DA/RT spread for those hours needs repricing; if the 23:00 EDT Sunday interval settles below the $57.13 window mean, the Saturday print stays an isolated bar. Third, check NYISO operations announcements by 12:00 EDT Sunday for any reserve shortage or Zone J outage notice covering 22:00 to 23:00 EDT Saturday; absence of a notice does not clear the event, but presence of one settles the question. Transco Zone 6 N.Y. gas and the Sunday evening load forecast are both missing from the pack and are the two inputs that would turn this from a tail observation into a view on tonight.

> A $4,032 bar that leaves behind a $57 mean and a $41 open is the city zone showing its tail on a quiet Saturday night; until a cause is named, it is a reminder of the distribution, not a change in the level.

Not investment advice. For informational purposes only.

Yesterday's tape: CAISO CSADIAB7N001 negative_lmp <= 0 — verified (observed -1187.63).

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

$1,015.17/MWh separated the top and bottom of CAISO's real-time nodal stack in the 10:05 UTC interval on Saturday. ELCAJNGT7N011 cleared at $232.02/MWh while CSADIAB7N001 printed $-783.14/MWh. The clock matters: 10:05 UTC is 03:05 PDT, deep in the overnight trough and hours before any solar reaches the system.

The system energy component was $35.55/MWh at both ends, so nothing about this is a marginal-fuel story. ELCAJNGT7N011 carried $181.26/MWh of congestion and $0.04 of losses; CSADIAB7N001 carried $-821.37/MWh of congestion and $-12.50 of losses. Each side is a clean cluster. All five top nodes were ELCAJNGT7 points (N001, N004, N009, N010, N011) at an identical $232.02/MWh, and CONTROLX1N008, N009, N010 plus CSADIAB7_N003 all sat at $-783.01/MWh on the same $-821.37 congestion term with $-12.36 of losses. Identical congestion components across a cluster are the signature of a single binding element with the two groups on opposite sides of it. Which element is not established. The 11:02 UTC snapshot lists YU XFMR2 A 69KV, SD1 NGILA-IVALLY 500 and YU XFMR4 A 69KV as top constraints with shadow prices unreported; SD1 is a San Diego-area 500 kV path, but the feed does not tie any of the three to these nodes, and no direction or shadow price is attached.

The rest of the system looked ordinary. THNP15GEN-APND traded at $48.22/MWh at 11:00 UTC. Load was 24,240 MW at 10:55 UTC against net load of 23,576 MW, with natural gas the main generation source. A $-783 print at 03:05 PDT on a Saturday, with no solar on the system, does not fit the usual oversupply template. The curtailment feed logged at least 1,380 events over the trailing seven days, with the count possibly truncated by pagination, but the sampled entries are stamped 2026-09-26 at 12:00 UTC and the largest is 2.701 MW of wind. Those samples say nothing about dispatch in the San Diego area during this interval, and real-time curtailment MW for 10:05 UTC is not in hand.

The angle hinges on persistence into the solar ramp, and this print predates that ramp by hours; the question is whether the congestion term survives the transition rather than whether solar created it. If CSADIAB7N001 still shows a negative LMP at 18:00 UTC (11:00 PDT), the overnight constraint has carried into peak solar and the curtailment-pressure read holds. If the ELCAJNGT7N011 congestion component is back at $0/MWh by 20:00 UTC (13:00 PDT), the overnight event was transient and the spread is a non-recurring print. Separately, watch the shadow price field on SD1 NGILA-IVALLY 500 in the next snapshot: any nonzero value would be the first evidence linking a named constraint to this divergence, and the day-ahead LMPs for both node groups, once pulled, will show whether the market saw it coming.

> A four-digit spread at three in the morning on a flat $35.55 energy component is a wires story, not a solar story, until the midday prints say otherwise.

Not investment advice. For informational purposes only.

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

A $322.37/MWh gap separated MISO's highest and lowest real-time nodes in the 10:50 UTC interval Friday, and both ends sat inside the NIPSCO footprint. NIPS.MICHCP12 printed $223.27/MWh while NIPS.MUNSTR.LN cleared at $-99.10/MWh. ILLINOIS.HUB, by contrast, settled at $40.50/MWh at 11:00 UTC, within a few dollars of the system energy component.

The energy component was $35.96/MWh at every node in both the top-five and bottom-five lists, so the whole spread is congestion plus pennies of losses. MICHCP12 carried $187.23/MWh of congestion; MUNSTR.LN carried $-134.73/MWh. Four of the five highest nodes were NIPSCO: MCHCP.ARR at $217.44, BAILY.ARR at $216.81, MAIZEMSV at $191.26, then DUNNBR1SF at $113.90. The second-lowest node was also NIPSCO, STJOHN.LN at $-48.58, before the depression bleeds into ComEd at CRETE.NU ($-30.87) and fades to near-flat at ZIONEC.MVP and WEC's PLSNT1.MVP ($4.81 and $5.12). Read as a map, that is a positive pocket around the Michigan City and Bailly nodes and a negative pocket around Munster and St. John on the Illinois border, with the sign flipping inside one utility zone. The driver is not confirmed. The snapshot's three listed top constraints all show n/a shadow prices, and none of their names points to northern Indiana. MISO logged 6,579 real-time binding-constraint events over the trailing seven days, and a LAKEGEORGETOWERROAD138 flowgate appears in the September 25 sample, but nothing in the source ties it to today's split. The outage feed is footprint-level and its sample rows conflict at the same timestamp, so it cannot confirm or rule out a NIPSCO unit or line outage either. Footprint load was 74,425 MW with net load at 64,507 MW and gas as the main source; this is not a system-stress print. It is local.

The angle treats the negative node as the tradeable edge, and the structure supports that reading: a $-99.10 LMP with $-134.73 of congestion is a deeper deviation from hub than anything on the high side except MICHCP12 itself. The caveat is that a single 5-minute interval without a named constraint is as consistent with a transient as with a persistent pocket. Three checks settle it. If the MUNSTR.LN congestion component is still below $-50/MWh at 13:00 UTC (08:00 CDT), the pocket has survived the morning load pickup and is not a dawn artifact. If MISO's real-time binding-constraint feed populates a shadow price for any NIPSCO-area flowgate in the 12:00 UTC interval, the spread acquires a name and a limit and becomes modelable rather than anecdotal. If the MICHCP12 minus MUNSTR.LN spread compresses below $20/MWh by 14:00 UTC with no shadow price ever posted, treat Friday's print as noise and move on. The day-ahead echo matters too: a NIPSCO basis of comparable sign in the October 3 DA run would mean the constraint is being modeled forward, not just hit once.

> A $322 spread with no named constraint is a coordinate, not a thesis; the edge at MUNSTR.LN is real only if it is still negative when the shadow price finally prints.

Not investment advice. For informational purposes only.

Yesterday's tape: ERCOT KNAPPRN lmppeak >= 200 — verified (observed 252.73); ERCOT KNAPPRN lmppeak <= 75 — not triggered (observed 252.73); ERCOT ENDPARKESS1 negativelmp <= 0 — verified (observed -39.24).

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

KNAPPRN cleared at $905.13/MWh in ERCOT's 10:50 UTC real-time interval on Thursday while ENDPARKESS1 and FLUVANNA12 both settled at -$103.77/MWh, a maximum nodal spread of $1008.90/MWh. That is 05:50 CDT, before sunrise, on a system carrying 58,420 MW of load with natural gas as the main source. HB_NORTH printed $49.19/MWh at 11:00 UTC. The hub is quiet; the nodes are not.

The high side is not a single bad print. BULLCRK12 cleared at $550.98/MWh, LAMESASLRG and ALPBESSRN tied at $275.23/MWh, and CROSSTRLRN came in at $203.74/MWh, five nodes stacked well above the energy component. The low side runs deeper than the two -$103.77 prints: DERMOTTALL at -$39.25/MWh, JADESLRALL at -$37.28/MWh and ANDMDSLRALL at -$22.03/MWh fill out the bottom five. Node names carrying SLR, BESS and ESS suffixes point toward solar and storage resources, but the fact pack does not confirm their location, their MW size, or whether the positive and negative clusters sit on the same constrained path. The 11:01 UTC snapshot lists 6940_D, ARGENTAAMATH1 and OAKCPERSHI11 as the top congestion constraints, with no shadow prices attached, so the binding element behind the KNAPPRN and ENDPARK_ESS1 split remains unidentified. Net load of 49,598 MW against 58,420 MW gross tells you renewables were contributing in the dark hours, which is consistent with export-limited wind or storage discharge being priced out behind a constraint, but that is inference, not a confirmed driver.

The outage tape offers context without a smoking gun. ERCOT logged at least 1,649 unplanned resource outage events over the trailing seven days, a count that may be truncated by pagination and should be read as a floor. The sampled events cluster on 2026-09-27: gas unit WCPP on forced outage, down 33 MW from 245 MW to 212 MW available, and solar resources PAR and TNG fully offline for Maintenance Level 1, removing 243 MW and 125 MW. None of those is dated Thursday morning, and no public coverage in the 48-hour window addresses this event. Day-ahead prices at the affected settlement points are not in the pack either, so the DART exposure at KNAPPRN and ENDPARKESS1 cannot be sized from here. Whatever a day-ahead position at KNAPP_RN was priced at, a $905.13 real-time clear against it is the scale of the problem.

The first checkable question is persistence. If KNAPPRN still clears above $200/MWh in the 13:00 UTC interval while HBNORTH stays under $60/MWh, the split survived the overnight-to-morning transition and is structural rather than a one-interval SCED artifact; a KNAPPRN print under $75/MWh at 13:00 UTC would mark it as a spike. If ENDPARKESS1 is still below $0/MWh at 14:00 UTC, the low-side curtailment is holding into daylight and the negative cluster deserves its own constraint attribution. Watch the posted shadow prices for 6940_D, ARGENTAAMATH1 and OAKCPERSHI11 in the 12:00 UTC SCED file: any of the three above $100/MWh names the driver this fact pack could not, and a widening between KNAPPRN and BULLCRK12 beyond their current gap at 13:00 UTC would argue the two are on different elements rather than one.

> A $1008.90 spread with a $49.19 hub is pure basis, and until a shadow price names the constraint, every nodal position near KNAPPRN or ENDPARKESS1 is carrying risk nobody can yet attribute.

Not investment advice. For informational purposes only.

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

$1,844.91/MWh separated the top and bottom of PJM's real-time nodal stack in the 10:50 UTC interval (06:50 ET) Wednesday. MACNEW_T115 KV cleared at $1,146.11/MWh; LENOX 115 KV cleared at -$698.80/MWh. The PJM-RTO aggregate printed $95.44/MWh in the same interval, so this is not a system event. It is one path.

The decomposition makes that plain. Energy was $95.00/MWh at every one of the ten extreme nodes; the entire spread is congestion plus rounding-error losses. MACNEW_T carried $1,049.56/MWh of congestion against $1.55 of losses, and Lenox carried -$793.96/MWh of congestion against $0.16 of losses. The rest of the high side lines up in a neat gradient: WMPOTTER115 at $1,076.10 with $978.97 congestion, WILLIAMM115 (MIDSTREM) at $998.88 with $901.29, then THOMPSON115 and TIFFANY 115 at $881.65 and $881.11 sharing an identical $783.10 congestion component. On the low side, four MEHOOPAN buses (13.8 kV, 13 kV, 115 kV) cleared between -$214.19 and -$213.67/MWh, every one of them with the same -$306.35 congestion. Identical congestion components across distinct buses mean identical shift factors to a single binding element; the ladder from $1,049.56 down to $783.10 on one side and from -$793.96 to -$306.35 on the other is the electrical distance from that element. Two of the negative MEHOOPAN buses are the PGCOG1CT and PGCOG2CT combustion-turbine nodes, which is where a unit either sits on the wrong side of a constraint or gets backed down to relieve it; the feed does not say which.

What the feed also does not say is the constraint itself. The three top congestion entries reported at 11:01 UTC are all 138 kV elements (ALBRIGHT ALB-GAR, BUTL_APS BUT-KAR, Lake George to Tower Road), none on the MACNEW/Lenox 115 kV path, and none with a shadow price attached. System context is unremarkable: load 86,645.7 MW, net load 82,657 MW, gas the main source at 11:00 UTC. PJM logged 144 daily outage events over the trailing seven days, with sampled planned-outage rows between 12,277 and 26,668 MW, but no zonal breakdown ties any of that to this 115 kV cluster. Treat the cause as unidentified, not as a transformer outage or a line trip until an outage ticket or a named constraint shows up.

The angle's basis and FTR framing depends entirely on persistence, and persistence is checkable. If MACNEWT's congestion component is still above the $783.10/MWh THOMPSON/TIFFANY level at 09:00 ET, the element has survived the morning ramp and the next question is whether day-ahead LMPs at MACNEWT, WMPOTTER and Lenox for October 1 carry congestion above the $95.00 energy component or price it away. If that component is back under $95.00 by 12:00 ET, the 10:50 print was an interval spike and the nodal-basis case collapses with it. Watch Lenox against the MEHOOPAN cluster as the second check: if Lenox congestion remains below -$306.35/MWh at 09:00 ET while MEHOOPAN holds near that level, the sink is a single-bus problem and the wider source-to-sink path is narrower than the headline spread suggests.

> Energy is $95 at every node on the list; the whole $1,845 is one unnamed constraint's shadow, and the trade lasts exactly as long as that element does.

Not investment advice. For informational purposes only.

Yesterday's tape: MISO NIPS.MUNSTR.LN negative_lmp >= 0 — verified (observed 22.93).

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

$782.61/MWh separated the top and bottom of the MISO nodal stack in the 10:50 UTC interval on Tuesday, 05:50 CDT. UPPC.AUTRAIN in the Upper Peninsula printed $693.97/MWh while NIPS.MUNSTR.LN in the NIPSCO footprint cleared at -$88.64/MWh. ILLINOIS.HUB sat at $56.70/MWh ten minutes later, so the hub told you almost nothing about either tail.

The system energy component was $41.82/MWh at every node in both tails; this is congestion, not scarcity. AUTRAIN carried $646.00 of congestion against $6.15 of loss, and the four nodes behind it are also Upper Peninsula: MIUP.MARQ at $672.59 on $624.50 congestion, and MIUP.KUESTER12, 13 and 14 at an identical $672.46 on $625.05. The entire top five sits on congestion between $624.50 and $646.00, which is a single regional price island, whatever the binding element turns out to be. At the other end, MUNSTR's -$131.10 congestion component is a clear outlier below the four NSP nodes in MISO North, which clustered between -$63.54 and -$63.07 on a shared -$98.61 congestion figure. The live feed lists three flowgates at the top of the congestion table, led by Lake George to Tower Road 138 kV, followed by Loon Lake to Faribault and the Barton Lake transformer, but every shadow price reads n/a and no named constraint is tied to AUTRAIN or MUNSTR in the source. Lake George to Tower Road did bind in real time a week ago at 11:05 UTC on 2026-09-22 with no override, alongside Wardwa to Bismark2 230 kV and Abbott to Traer 161 kV. MISO logged at least 5,937 real-time binding-constraint events over the trailing seven days, and the feed showed only a five-record sample, so read that count as a floor.

Background conditions are ordinary. System load was 72,292 MW at 10:55 UTC, net load 59,795 MW, with coal the largest source. The estimated outage stack sampled at 05:00 UTC on 2026-09-23 showed 9,666 MW planned, 7,724 to 7,813 MW forced and 6,885 to 7,128 MW derated; at least 84 outage events were logged over the same seven days. None of that data carries transmission-outage detail for the Upper Peninsula or northern Indiana, so the physical trigger for either tail is unconfirmed.

The angle frames this as one unresolved constraint risk; the data leaves open whether it is one constraint or two, and that is the trade. If AUTRAIN's congestion component still exceeds its $41.82 energy component at 12:00 UTC (07:00 CDT), the Upper Peninsula island has outlived the interval that produced it and the MARQ and KUESTER nodes should be checked for the same congestion figure. If MUNSTR prints back above $0/MWh at 12:00 UTC while AUTRAIN congestion stays above $41.82, the two tails are decoupled and the Indiana leg was independent. Watch the 5-minute binding-constraint feed for a non-null shadow price on Lake George to Tower Road by 12:00 UTC; a repeat of the 2026-09-22 binding with a populated shadow price would be the first hard link between a named element and the UP prices.

> A $782.61 spread on a $41.82 energy price is a transmission story with two unexplained ends, and until a shadow price attaches to either one, the risk is that each tail resolves on its own clock.

Not investment advice. For informational purposes only.

Yesterday's tape: SPP SPPNORTHHUB lmppeak >= 56.89 — verified (observed 433.08); SPP SPPNORTHHUB lmppeak <= 36.19 — not triggered (observed 433.08).

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

SPP North Hub real-time LMP printed $601.71/MWh at 19:55 UTC on Sunday, 14:55 local, against a $56.89/MWh mean across the trailing 287 five-minute bars. By 07:10 UTC Monday, 02:10 local, the same hub cleared $22.85/MWh, and the series closed at $25.50/MWh at 10:55 UTC. The entire round trip, spike and unwind, fits inside a single 23.9-hour window.

What the source material does not say matters as much as what it does. No binding-constraint magnitude accompanies the spike: the three flowgates named in the live snapshot, TMP72732664, TMP72833443 and WTP100_92100, all report shadow prices of n/a. No wind, load-forecast or forced-outage record is attached to the 19:55 UTC interval, so the driver of the print is unrecorded here. What the feed does show is a system carrying 29,852 MW of load at 10:55 UTC Monday, net load of 22,524 MW, and coal as the main generation source. Around that sits a week of outage noise: the anomaly log counts at least 1,173 generation-capacity-on-outage events since 21 September, and sampled rows from 12:00 UTC on 21 September show between 3,563 MW and 3,876 MW of coal offline. Those counts carry a truncation flag and the outage rows are a week old. They are background, not cause.

The current-price picture is itself contested. The hub series ends at $25.50/MWh, while the live snapshot shows $36.19/MWh for the 11:00 UTC interval. Both readings are flagged medium confidence and both are marked as conflicting, so the "open" for Monday depends on which feed you trust. Neither is the $26 figure in the angle. The variable-energy-resource curtailment log adds a further wrinkle: at least 1,943 events over seven days, yet every sampled solar row reads zero. Whatever was curtailed, the sample does not show it.

The angle frames this as a day-ahead versus real-time exposure at North Hub. The fact pack carries no day-ahead LMP for the hub, so that spread cannot be measured from this material, and nothing here forecasts Monday evening load or wind. What can be checked is narrower. If North Hub real-time LMP at 19:55 UTC on 28 September clears above the $56.89/MWh trailing mean, the afternoon tail has repeated in the same clock interval two days running. If it stays below $36.19/MWh at that interval, Sunday reads as a one-off. Separately, if the SPP outage report for 12:00 UTC on 28 September shows coal on outage above 3,876 MW, the top of last week's sampled range, the supply-side backdrop has tightened rather than eased. And if any of the three named flowgates posts a non-zero shadow price in the 11:00 UTC snapshot on 29 September, the congestion channel that is blank today has started reporting.

The honest read is that a single five-minute bar at $601.71/MWh moved the 23.9-hour mean to $56.89/MWh while the hub spent Monday's early hours between $22.85/MWh and $36.19/MWh. That is a tail, not a level, and the tail has no documented cause in this pack.

> A $601.71 print with no recorded driver is not a thesis; it is a reminder that North Hub's afternoon risk lives in five-minute bars, not in the daily mean.

Not investment advice. For informational purposes only.

Yesterday's tape: CAISO THSP15GEN-APND lmppeak >= 54.04 — verified (observed 114.22); CAISO THSP15GEN-APND lmppeak <= 44.54 — not triggered (observed 114.22).

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