Grid Alpha

The last five-minute bar of MISO's Michigan Hub real-time tape Wednesday cleared $1,032.08/MWh at 18:40 CDT. The 24-hour mean across the same window was $43.30/MWh; the low was $23.70/MWh at 11:50 CDT. Roughly 24 times the mean, printed on the terminal interval of the series.

That last detail governs everything else. The snapshot runs 303 five-minute bars from 06:05 to 18:40 CDT, and the peak lands on bar 303, which means nothing in this dataset distinguishes a single interval of congestion rent from the first rung of a longer evening ramp. Michigan Hub is back at $42.48/MWh as of Thursday 06:04 CDT, inside the prior window's mean, so the event resolved overnight one way or the other.

Attribution is thinner than the print deserves. MISO logged at least 84 generation-outage estimate events over the trailing seven days, but every sampled row carries the same 2026-08-21 00:00 CDT stamp, the largest showing 11,857 MW derated against 10,055 MW forced and 7,098 MW planned, with no regional split and no way to tie any of it to Michigan five days later. The real-time binding-constraint feed carries at least 4,203 shadow-price events over the same seven days; the two sampled constraints are a Sheffield 161/69 kV transformer and the Chicago_Praxair flowgate into Wilton Center, both stamped 2026-08-20, and neither sits in Michigan. Both counts are lower bounds truncated by pagination. There is no transmission outage, no unit trip, and no MISO-PJM or MISO-IESO interchange data covering the 18:40 interval at all.

The premise that no driver has been confirmed is therefore a statement about the feed, not about the grid. What I would watch first is the reconstructed tape past 18:40 CDT: if the print extends across multiple intervals, it reprices evening-ramp scarcity at the hub and day-ahead-to-real-time basis should widen to follow it; if it stands alone at bar 303, it is a congestion artifact and the reversion to $42.48 is the entire story. Second, watch whether a Michigan-area constraint enters the binding set during the evening ramp on subsequent days. The sampled flowgates point west and south into Illinois and Iowa, and a latent-scarcity thesis at Michigan Hub needs either a Michigan-area constraint or a Michigan-area forced outage underneath it to carry weight. Third, note that no day-ahead Michigan Hub price for today appears anywhere in this material, so the gap between what the forward session priced and what real time delivered is inferred rather than measured. Traders sizing a DART position on this event are working from one number and a hole where the counterfactual should be.

The wider point is about the shape of the distribution, not the level. A hub that spends a full session between $23.70 and roughly $43 and then touches four figures on one bar is telling you the supply stack has a cliff in it somewhere above the working range. Where that cliff sits, and how much of it is forced outage versus transmission, is the question worth answering before the next evening ramp.

> Scarcity you cannot attribute is scarcity you cannot hedge: until a Michigan-area constraint shows up in the binding set, $1,032 on the final bar is a data point, not yet a signal.

Not investment advice. For informational purposes only.

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

SPP North Hub real-time LMP printed $1,210.49/MWh at 23:00Z Tuesday, 18:00 local. Five hours earlier the same hub cleared $17.70/MWh, and the window mean sat at $44.28. As of the 06:05 CDT snapshot Wednesday, the hub is marked $26.30.

The shape is a conventional evening net-load ramp taken to an unconventional level, but no source in the pack ties a driver to that bar: not a load forecast miss, not a wind ramp-down, not a binding constraint. The outage stack is suggestive rather than diagnostic. SPP logged at least 1,173 generation-on-outage events over the trailing seven days, with coal dominant in the sampled rows and a spread of 767.7 MW to 2,286 MW on outage inside a single 2026-08-19 timestamp, which reads more like inconsistent reporting than a coherent capacity picture. The 1,943 VER curtailment events are similarly unhelpful: every sampled solar row shows zero across energy, manual, and redispatch categories. Two data caveats matter more than either count. The series terminates on the peak bar itself, so there is no observed decay path from $1,210 down to $26.30; the mean-reversion is inferred from a single current tag, not watched. And the series carries 309 five-minute bars against a stated window of 11:05Z to 23:00Z, roughly 12 hours of clock time for about 25.75 hours of data. If the window is the shorter one, $44.28 is a half-day mean anchored on the daylight trough, and the true 24-hour baseline the spike traded against is unknown.

That ambiguity is the whole trade. If the $44.28 mean is a genuine full-day average, the $1,210 bar is a fat tail on an otherwise normal Tuesday and $26.30 is the honest anchor into today's ramp. If it is a half-day mean built from midday hours near $17.70, the baseline is understated and last night's evening block was running hotter than the mean suggests before it broke. What I would watch is the 17:00 to 19:00 local window against the same coal outage set: if the forced capacity that was out Tuesday evening is still out Wednesday evening and load repeats, the five-minute market re-prices toward scarcity regardless of where the overnight tape settled. The absent inputs are the ones that would size this properly. No SPP wind or solar forecast, no load forecast, no reserve margin, and no day-ahead North Hub prices for today, so there is no way to judge whether the market has already priced a repeat into the DA-to-RT spread or is leaving it open.

FERC's approval Wednesday of SPP's economic topology reconfiguration tariff revision cuts the other way over time: reconfiguring topology to relieve congestion and displace higher-cost generation is structurally bearish for the tail. The filing's effective date and its practical reach into North Hub congestion are not specified, so it is a 2027 consideration, not a tonight consideration.

> A $1,210 print with no observed decay path is not a price signal about yesterday; it is an unanswered question about tonight's ramp.

Not investment advice. For informational purposes only.

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

The widest real-time nodal spread in PJM hit $684.17/MWh in the 10:40 UTC interval, 06:40 EDT Tuesday morning. MACNEW_T115 KV cleared at $420.75/MWh while LENOX 115 KV settled at -$263.42/MWh. Both nodes sit on the same system, priced off the same energy stack, roughly 685 dollars apart.

The mechanism is visible in the components. Every one of the ten extreme nodes carries an identical energy component of $30.26/MWh, which strips the narrative down to a single variable: MACNEW_T115 shows $389.97/MWh of congestion against $0.52 of loss, and LENOX shows -$293.78/MWh of congestion against $0.10 of loss. This is not a single-node artifact. WMPOTTER115 KV printed $394.73/MWh on $363.78 of congestion and WILLIAMM115 KV printed $366.05/MWh on $334.96, while TIFFANY 115 KV and PTGENRGY115 KV cleared at exactly $322.74/MWh each with congestion of $291.11 apiece. The negative tail is equally clustered: four of the five lowest nodes sit at the Mehoopany wind complex, all at -$112.87/MWh of congestion and LMPs between -$83.42 and -$83.69. Identical congestion values across paired nodes point to a common binding element and near-identical shift factors, though the feed names no constraint and reports no shadow price.

Two things worth separating before extending this into the afternoon session. The snapshot is one five-minute interval at breakfast-hour load, not a shape, and the data contains no evidence the spread survives the morning ramp; persistence is a hypothesis, not an observation. What would test it is whether MACNEW_T115 and WMPOTTER115 hold their rank across successive intervals and whether the TIFFANY/PTGENRGY pair keeps clearing at the same congestion number. If those pairings decay independently, the driver is a transient dispatch outcome around low overnight load and high wind. If they hold their lockstep into the load pickup, the binding element is structural and the congestion component stays the dominant term in any 115 kV basis position on that path. The outage backdrop offers weak support at best: at least 144 PJM outage anomaly events over the trailing seven days, with a sampled 2026-08-19 window showing forced outages between 3,508 and 4,618 MW and maintenance between 2,681 and 4,421 MW, but that data is a low-confidence sample with an unclear unit of observation.

Two gaps limit how far this goes. No day-ahead LMPs are supplied, so the DA-to-RT congestion basis cannot be evaluated and neither the FTR nor the virtual entry point is priceable from what is here. No generation or curtailment data exists for Mehoopany, so the intuitive read that wind output is pinning the negative side remains an inference. Separately, FERC approved a two-year RMR extension for Brandon Shores and H.A. Wagner through 2031, but those units sit outside Baltimore in BGE with no stated electrical link to this pocket; treating it as related would be a stretch.

> A $684 spread with a constant $30.26 energy component is not a scarcity event: it is a transmission limit repricing two ends of the same grid, and until the binding constraint is named, everything else is inference.

Not investment advice. For informational purposes only.

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

HB_HOUSTON real-time LMP hit $570.55/MWh on the 02:00Z bar, roughly eleven times the $52.83 twenty-four-hour mean and about twenty-four times the $23.53 the hub was quoting as of the 06:02 CDT Monday snapshot. One five-minute bar carried the entire distribution.

Start with the clock, because the framing depends on it. 02:00Z on 24 August is 21:00 CDT on 23 August: Sunday evening, after solar rolloff, not an overnight hour. The day's low, $16.61/MWh, printed at 14:00Z, which is 09:00 CDT, well before the midday solar trough most people would assume. On supply, the outage feed shows SCES, a 933 MW subbituminous coal unit, at zero available on a forced extension; SPARTA solar fully forced out at 104 MW; and wind derates of 27 and 14 MW at SAGEDRAW plus 29 MW at SHANNONW. Every one of those rows carries a 2026-08-20 05:01 timestamp, four days ahead of the price event, and nothing in the feed confirms the status as of today. The five sampled SCED shadow prices run $0.053 to $0.762/MWh, which is functionally no congestion, but those rows are untimestamped and cannot be tied to the peak interval.

Two data caveats worth carrying into the session. The series is 288 five-minute bars, which implies a full 24 hours against a stated window that is shorter, so it is unclear whether the $52.83 mean covers a complete diurnal cycle. And the peak sits on the final bar of the stated window while the current price is $23.53, meaning the reset quote comes from a later, unlabeled timestamp. Neither breaks the story: a print eleven times the mean is a print. Both argue against treating $52.83 as a reliable anchor for anything.

So the reading is scarcity pricing into the evening net-load peak, with local congestion unproven. What I would watch is the 21:00 CDT hour specifically, not a generic overnight block, and the day-ahead clear for that hour against the midday trough. If SCES's 933 MW is still out and Monday's ramp arrives on a working-day load shape rather than a Sunday one, that single hour reprices first and the DART on it carries the day. If shadow prices stay sub-$1 through the evening ramp, the move is system-wide reserve scarcity and hubs converge; if constraints start binding meaningfully above that level, Houston separates from the rest of ERCOT and the hub spread is the cleaner expression of the same view. The feed carries at least 1,790 unplanned outage events and at least 7,715 binding shadow-price records over the trailing seven days, both sampled and both lower bounds, so the outage stack is doing work even when the tape is quiet at $23.53.

The honest limit: without day-ahead HB_HOUSTON quotes, a DART history for the evening hours, or a Ship Channel gas print to set the heat-rate floor, the entry level on a peak-hour repricing view is a judgment call rather than a calculation. The asymmetry is visible; the price of taking it is not.

> Scarcity didn't vanish by morning: it compressed into one 21:00 bar, and the hour is the trade, not the spot.

Not investment advice. For informational purposes only.

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

Louisiana Hub real-time LMP printed $1,378.58/MWh at 16:25 CDT Saturday against a 24-hour mean of $60.01. That is a 23x peak-to-mean spread, and it was gone almost as fast as it arrived: the current print sits at $21.83/MWh, barely above the day's low of $18.83 set at 08:45 that morning. Southern MISO flashed scarcity pricing for a few intervals, then unwound into Sunday morning.

The anatomy matters more than the headline number. The peak landed late Saturday afternoon local time, a shoulder-hour slot where air-conditioning load still lingers but committed capacity thins out. Across 285 five-minute bars in the snapshot window, the hub spent most of the day pinned near the $19 to $60 range, which means the $1,379 print was a narrow spike, not a sustained scarcity regime. The backdrop supports a tightness story: MISO's sampled outage records showed roughly 7,206 to 7,472 MW of derated capacity, with 4,841 MW forced and 4,301 MW planned, and at least 84 generation-outage anomaly events logged over the prior seven days. Real-time binding-constraint activity was heavy too, with more than 5,060 shadow-price events over the same week, though none of the sampled constraints sit in the Louisiana or southern MISO region. What is missing is the specific driver. No named flowgate, no ORDC adder confirmation, no reserve-level print, and no day-ahead comparison in the data. The outage figures are also dated 2026-08-17, five days before the spike, so treat them as context rather than cause.

For today's session the setup is the unwind, not the spike. A hub that round-trips from $1,379 to $22 inside a day is telling you the scarcity was interval-specific, likely a local reserve or ramp event rather than structural capacity shortage. If day-ahead cleared nowhere near the real-time peak, the DART spread on Louisiana Hub becomes the number to audit: persistent negative DART after a spike day often means the DA model is under-pricing southern congestion risk. Watch whether Sunday's evening ramp re-tightens; if forced outage levels remain near the 4.8 GW sample and late-afternoon prices re-elevate toward triple digits, the scarcity premium is not finished. If instead the hub holds the $20 to $60 band through the ramp, Saturday was a one-interval event and the volatility premium should compress. Either way, the 23x spread is a reminder that southern MISO's tail risk is fat even on a Saturday in late August.

The one discipline item: the fact set cannot confirm whether a scarcity-pricing adder was active at the peak, so do not assume ORDC mechanics until reserve data confirms it.

> A $1,379 print that dies at $22 is not a scarcity regime; it is an invoice for being short ramp in southern MISO for exactly one afternoon.

Not investment advice. For informational purposes only.

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

A $183.44/MWh intraday spread ripped through SPP's Oklahoma footprint in the 04:00 UTC interval on Aug. 22, with the maximum print landing at 11 p.m. local time Friday night. The high side: CSWSDUNCANLD1_69 at $171.78/MWh. The low side: CSWSDUNCTP4LDOMPA at negative $11.66, two AEP/CSWS Duncan-area nodes priced a world apart on the same system.

The decomposition tells the whole story. Both nodes carried an identical $25.95 energy component; the gap is pure congestion. Duncan LD1 booked a $145.52 congestion adder while the Duncan tap node sat at negative $37.90, a one-sided load pocket with nowhere for local supply to go. The rest of the Oklahoma complex clustered near negative $8: CSWSDUNCN_ESLD138 at negative $8.56, WFECRUSSETTLD4 at negative $8.43, OKGERUSETLD1 at negative $8.03. This was not a regional scarcity event. It was a localized choke, and one open question matters here: which constraint bound, and whether an outage drove it. SPP logged at least 1,173 generation-capacity-on-outage records in the trailing week, with sampled coal outages running 2,278 to 3,428 MW. Sampled solar curtailment intervals showed zero MW curtailed, so renewables were not the binding story. One data caveat: several adjacent high-priced nodes in the same snapshot are stamped a day later, so treat the 08-22 header timestamp as the operative read.

The timing is the angle. FERC approved SPP's topology optimization plan on Aug. 20, two days before this print, giving the ISO authority to reconfigure switching to relieve exactly this class of constraint cost. The precedent is MISO, whose market has saved nearly $100 million year-to-date under a similar reconfiguration approach. If SPP's implementation delivers even a fraction of that, congestion rents at chronically constrained pockets like Duncan compress over time.

For traders, the near-term read is persistence. If the Duncan split extends beyond a single overnight interval, CSWS-area TCR and congestion exposure gets repriced fast; if it fades with returning coal capacity, this was an outage artifact, not a structural pocket. Watch the constraint-level shadow prices around the Duncan nodes versus the broader Oklahoma cluster near negative $8, and watch SPP's implementation timeline: the spread between today's congestion reality and tomorrow's optimization savings is the trade.

> When two nodes sharing a fence line price $183 apart, congestion is the market; topology optimization is now the regulator's bet on killing it.

Not investment advice. For informational purposes only.

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

The widest print of the week came Thursday night local time in CAISO: a $306.14/MWh maximum real-time nodal spread at the 04:55 UTC snapshot, 21:55 PDT. The tail of the distribution tells the story. THSP15GEN_ONPEAK-APND printed the top LMP at $306.14/MWh with a congestion component of $110.59 against an energy component near $188.74. At the other end, five APND scheduling nodes across AVA, AZPS, and BPAT paths, including MALIN, printed $0.00/MWh at 10:50 UTC Friday, 03:50 local. Zero, not deeply negative. The spread tape was wide and bifurcated, with congestion doing the work at both poles.

The congestion signs matter. THPACWGEN_ONPEAK-APND carried a congestion component of -$94.42/MWh against an LMP of $88.31/MWh in the same 04:55 UTC interval, meaning PACW was trapped on the wrong side of a binding path while SP15 cleared at the system's top print. Behind the snapshot, the trailing seven days logged at least 1,668 CAISO curtailment events, with sampled solar rows at 0 MW and wind up to 1.627 MW. The count is large even if the sampled megawatts are small; the signal is frequency of dispatch intervention, not volume. Whether the 04:55 spread was outage-driven, ramp-driven, or a single binding constraint is not identified in the feed, and I won't pretend otherwise.

East of the Rockies, the machinery was strain, not spreads. PJM recorded 147 daily outage records over seven days, with sampled forced outages up to 15,565 MW and maintenance up to 12,639 MW. MISO's estimated outage feed showed 84 records, with one sample at 15,015 MW derated and 12,515 MW forced, and the real-time binding-constraint feed returned at least 5,233 shadow-price events, a number that may be API-truncated. ISONE logged 326 real-time binding intervals, with the SYSTEM_10MINSYNC constraint showing a marginal value of 23.62 in samples. That is the backdrop for the evening hub strength the angle flags: heavy forced-outage stacks plus persistent binding constraints tighten the evening ramp. Exact peak levels are not in this week's feed, so the hub numbers stay out of this column until the data supports them.

Now the housekeeping. One published call on this book failed to trigger: the CAISO HPLNDJT negative-LMP call missed at -28.99. The week's verified snapshot prints bottomed at $0.00/MWh on APND scheduling nodes, and no sustained deep-negative prints appear in the data here. A negative-LMP thesis needs renewable saturation plus congestion trapping, and the tape showed curtailment frequency but not the price outcome. We print the miss because a signal book that hides its losers is marketing, not research. The honest read: negative-price setups at scheduling nodes remain a watch item, not a triggered trade.

What I'd watch from here. If CAISO curtailment events keep compounding at this pace while APND scheduling nodes pin at $0.00, the path to deep negatives runs through the spring-style solar hours, not 04:55 UTC snapshots; watch the midday intervals rather than the overnight extremes. If PACW congestion stays near -$94 while SP15 holds triple-digit congestion, the SP15-PACW basis is the live expression of the same constraint that built the $306 spread. In MISO, 5,233-plus shadow-price events with 15 GW-scale derates says congestion rent is accruing somewhere; the open question is which constraints repeat.

> The spread was real, the strain was real, and the miss was real; the book prints all three because that is what a tape is for.

Not investment advice. For informational purposes only.

Yesterday's tape: CAISO HPLNDJT6N001 negative_lmp >= 37 — not triggered (observed -28.99).

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

CAISO's real-time market printed a $1,497.57/MWh nodal spread overnight Thursday, with SCHLTE1PL1X3-APND at $1,472.38/MWh against HPLNDJT6N001 at negative $25.19. That is a 40x gap between the energy component and the delivered price at the top node. The system energy price never moved; the wires did.

The decomposition is the story. SCHLTE's LMP breaks down to $1,423.54/MWh of congestion against an energy component of $36.76/MWh and a loss component of negative $1.23. At the bottom of the stack, HPLNDJT carried negative $72.07/MWh of congestion against energy of $37.11. Both tails of the curve are pricing the same system energy within 35 cents of each other; the entire $1,498 spread is transmission. This is not a scarcity event, a gas event, or a ramp event. It is a binding constraint isolating load pockets from supply pockets in the 3:20 to 3:50 AM PDT window, when the system should be at its most relaxed. Four PGST_2 nodes (RDRR01, PDRP42, PDRP40, PDRP39) printed a uniform $686.01/MWh with $636.48 of congestion each, which reads as a second, separate constrained cluster rather than smeared system-wide stress.

The negative tail has company. GRANITE6N001A printed negative $14.86, two CLERLKE nodes printed negative $11.20, and HARTLEY6N001 printed negative $9.14, all at 10:20 UTC. A cluster of sub-zero prints in the overnight hours points to must-run or non-dispatchable supply trapped behind the same constraints. CAISO logged at least 1,668 curtailment events over the trailing seven days, though the sampled records are wind-only and dated August 13, so the curtailment linkage to Thursday's negative prints is unconfirmed.

One caveat on the headline number: the two extreme prints are 30 minutes apart. SCHLTE's $1,472.38 is a 10:50 UTC reading; HPLNDJT's negative $25.19 is 10:20. The same-interval spread may be tighter, and no constraint name or outage record has surfaced to explain the $1,423 congestion component. Treat the $1,498 as an upper bound on the location pair, not a cleared hourly basis.

The trade implication turns on persistence and the DA/RT split. No day-ahead prints are in the data, so the basis cannot be sized yet. If SCHLTE's congestion component survives into Thursday's evening ramp, when solar rolls off and the system tightens, the spread widens rather than closes, and virtual convergence at the constrained cluster gets interesting. If the constraint clears with the morning pickup, the negative tail at HPLNDJT and CLERLKE collapses back toward the $37 energy component first, since those nodes are congestion-short, not energy-short. Watch the next intervals for whether PGST_2's $686 cluster moves in sympathy with SCHLTE; correlated congestion means a shared interface, divergent prints mean two separate trades.

> When the energy component sits at $37 on both ends of a $1,498 spread, the grid is the market, and whoever maps the binding constraint first owns the basis.

Not investment advice. For informational purposes only.

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

SPP's real-time market printed a $246.48/MWh top-to-bottom nodal spread in the 04:00 UTC interval, 23:00 CDT Tuesday night. The high side was CSWS46STNOLD1 at $214.45/MWh in the AEP Oklahoma load zone; the low side was EDEAUR1241LDAURORA_124 at -$32.03/MWh in Empire District's Aurora pocket. Both prints came out of the same dispatch interval, and neither was an energy story.

The split is pure congestion. The system energy component sat at $21.47/MWh at every listed node, top and bottom alike. CSWS46 carried a $193.22 congestion adder on top of that flat energy price; Aurora carried -$53.89. The high-price side is a cluster, not an outlier: CSWSYALEARCLD1 and CSWSPNEPEORLD1 both printed near $208.6, and the CSWS Mohawk nodes followed close behind. The negative side is also broader than one node: WFECRUSSETTLD4 at -$22.00 and OKGERUSETLD1/2 at -$21.49 show sub-zero pricing extending across Oklahoma into the OG&E footprint. This is a map problem, not a tick problem. No named flowgate is identified in the available data, so the specific element loading up the $193 congestion into CSWS remains an open question. Trailing-seven-day logs show at least 1,173 generation capacity-on-outage records and 1,943 VER curtailment records in SPP, with sampled coal outages ranging roughly 2,406 to 4,071 MW on August 12, though the outage data is low-confidence sampling and no single event cleanly explains tonight's geometry.

The setup is the classic overnight SPP pattern: cheap energy system-wide, transmission that cannot move it where load or constraint shadow prices want it. With energy at $21.47, anyone positioned between these two pockets is trading the constraint, not the commodity. If the spread holds into morning ramp, congestion revenue rights and any virtual spread positions spanning CSWS to the Aurora/WFEC/OKGE cluster carry real value; if the binding element relieves with the dawn pickup in dispatchable headroom, the spread collapses toward energy and the trade is over. One caveat on the "live" framing: the snapshot is a single 04:00 UTC interval as of the 06:02 CDT pipeline run. Nothing in the data confirms whether later intervals widened, held, or mean-reverted, so sizing should assume the print may already be stale. Watch whether the negative pricing persists at Aurora and the OKGE Russett nodes into the next several intervals; persistence would suggest structural oversupply in that pocket, likely wind-driven given the region, rather than a transient redispatch artifact.

> When energy is flat at $21 and the spread is $246, you are trading the wire, not the watt.

Not investment advice. For informational purposes only.

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

CAISO's real-time market posted a maximum nodal LMP spread of $810.14/MWh in the 10:20 UTC interval, one of the widest single-interval separations on the board. CSADIAB7N001 cleared at $-741.85/MWh while SPRPPGNODE2 held $68.29/MWh at the top of the stack.

The decomposition tells the whole story. The system energy component was a uniform $35.92/MWh across every listed node, so none of this spread is energy; it is congestion and losses, full stop. CSADIAB carried a congestion component of -767.65 against losses of -13.82, and it was not alone: CONTROLX1N008, N009, N010 and CSADIAB7N003 all printed $-741.75 with the identical -767.65 congestion figure, a clustered deep-negative pocket rather than a single-node flicker. On the other side, SPRPPGNODE1 and SPR3PP2SOLARGNODE matched GNODE2 at $68.29 with congestion of 29.20, so the premium side is a zone, not an artifact of one bus. The backdrop fits: CAISO logged at least 1,668 curtailment events over the trailing seven days, consistent with active renewable oversupply pressing against a binding constraint somewhere between these pockets.

One caution on the angle: 10:20 UTC is 03:20 PDT, dead overnight in California. This is not a midday solar bell print, and the sampled curtailment data shows wind megawatts being cut, not solar. That cuts both ways. A negative-742 node clearing at 3 AM local implies either a genuinely binding local constraint with nowhere for the generation to go, or a telemetry artifact, and no public coverage yet confirms which. If the constraint holds into the morning ramp, the spread should compress as load absorbs the pocket and the SPR premium fades toward system energy; if the negative print persists across intervals, it reads as structural, and congestion revenue rights along that path reprice. What I'd watch: whether CSADIAB repeats in the next interval set, whether the -767.65 congestion component releases, and whether these nodes are actually settleable before sizing anything against the print.

> An $810 spread with a $35.92 energy component is pure congestion geometry: trade the constraint's duration, not the price level.

Not investment advice. For informational purposes only.

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