West Hub printed a $287/MWh swing on 2026-04-26, a move big enough to reprice the rest of the week if it holds. The read-through is not subtle: basis widened hard enough to signal a local constraint, not a clean system-wide move.

There is not enough verified detail in the supplied facts to pin the exact day-ahead and real-time LMPs, the congestion component, or the interval that drove the move. [FACT NEEDED] on the West Hub price print, [FACT NEEDED] on the basis pair versus Zone A or adjacent hubs, and [FACT NEEDED] on whether the swing came from one interval or a multi-day run. That matters because a single spike can fade; a durable basis break usually needs a transmission limit, outage, or load-pocket imbalance that lasts past one dispatch cycle.

For now, the mechanics point to western New York supply tightness relative to load. If NYISO keeps posting west-side congestion notices, then West Hub should stay bid versus the rest of the state. If the move was tied to a short-lived outage or a temporary interface limit, then the spread can snap back quickly once the constraint clears. Watch West Hub versus Zone A, and watch whether the spread persists into the next operating day rather than just the next interval.

> When West Hub gaps this far, the market is usually telling you the wire, not the load, is setting the price.

Not investment advice. For informational purposes only.

HB_HOUSTON touched $1153/MWh last night, roughly 12x the 24-hour mean embedded in the thesis. That is not a rounding error in a hub that normally trades like a liquid reference point; it is a scarcity print, and it says the evening stack got thin enough to clear at punitive levels.

The backdrop is not subtle. ERCOT’s unplanned resource outage feed showed at least 19,306 events over the last 7 days, though that result is sample-limited, and the sample includes forced reductions across gas, wind, and water units: DEC was down 22 MW, SANTACRU 25 MW, BUCHAN 17 MW, AQUILLA 49 MW, and FOARDCTY another 10 MW on a forced-extension basis. That is not one big failure, just a steady accumulation of missing megawatts. ERCOT’s SCED shadow-price feed also printed at least 7,330 events over the same window, with constraint samples at 6.827 on WESTEX, 46.446 on 630_B, and 62.388 on BOWFMR1. In parallel, the LMP/DART feed logged at least 433,228 events, a reminder that price discovery is active and the system is already moving through a lot of congestion and redispatch.

The market question is whether last night was a one-off evening squeeze or the first signal of a repeatable pattern. If the same outage stack remains in place and the same generation stack is called again into the evening ramp, HB_HOUSTON can reprice quickly before the broader ERCOT screen catches up. Watch Houston versus the wider ERCOT hubs and the behavior of binding constraints into the load pocket: if congestion tightens while operating reserves thin, the spread can widen faster than outright hub averages suggest.

The clean read is that scarcity is no longer theoretical when Houston can print four digits on a routine evening.

> In ERCOT, the price is telling you the evening stack, not the average day, still owns the tape.

Not investment advice. For informational purposes only.

NYISO’s N.Y.C. LBMP hit $609/MWh in the evening, or 12.6x the 24-hour mean. That kind of move is not a noise print; it says the zone cleared against a thin margin and the market had to pay up for deliverability into Zone J.

The mechanics are straightforward even if the exact trigger chain is not yet public. In a constrained load pocket like NYC, a spike of that size usually means local demand, imports, and unit availability lined up badly for one interval, then eased as conditions normalized. The fact pattern points to a tight supply-demand balance rather than a broad system repricing. If the hour also came with binding transmission into Zone J, shadow prices would have done part of the work; if not, then the scarcity sits more squarely in local generation and load.

What I would watch today is whether the same setup repeats in the evening ramp. If load follows a similar profile and import capability stays limited, then the right comparison is not the day’s average but the marginal hour against nearby zones and the day-ahead strip. If real-time keeps clearing far above day-ahead in Zone J, that is the cleaner tell that the pocket remains tight and that traders should keep an eye on the spread into the rest of the state.

> A $609/MWh print in NYC is the market telling you the last megawatt in Zone J got expensive fast.

Not investment advice. For informational purposes only.

SPP just printed a $664.16/MWh real-time spread, with WR.VOLT.0174 at $621.21/MWh and WRKPAUGICN1 at -$42.95/MWh. That is not a modest basis move; it is a routing problem wearing a price tag.

The spread is mostly congestion, not energy. WR.VOLT.0174 carried a $605.96/MWh congestion component against a common $15.93/MWh energy adder, while WRKPAUGICN1 showed -$58.19/MWh congestion on the same $15.93/MWh energy component. FRONTIER was the second-highest node in the snapshot at $460.75/MWh, which tells you the pressure is not isolated to a single bus. The node stack is steep: the top end clears at more than 10x the energy component, while the low end is actively negative.

That shape usually means a binding constraint is forcing incremental MWh toward one pocket of the footprint and away from another. The anomaly log says SPP has seen 2,513.49 MW of wind redispatch curtailments and 28,471.4 MW of generation capacity on outage in the recent sample, with coal carrying 9,590.7 MW of that offline stack. I would not treat those numbers as a direct read on this 18:20 interval, but they are the kind of setup that lets congestion gaps open fast and stay open. The fact that WRKPAUGICN1, WRKPAUGICN2, WRKPAUGICN3, and WRKPAUGICN4 all printed the same -$42.95/MWh is another tell: this is a zone effect, not a one-node typo.

For traders, the key question is persistence. If the congestion component stays above the energy leg, then the basis between the western high-price pocket and the KP low-price pocket should remain wide, and any spread linked to WR.VOLT.0174 versus WRKPAUGICN1 will remain sensitive to operator action and constraint relief. If the next few intervals show FRONTIER and the WR.KP cluster converging, this was likely a sharp but transient constraint event. If not, watch for the spread to reprice as a session-wide congestion trade rather than a single 5-minute spike.

> When the energy component is the same and the congestion leg does all the work, the market is telling you where the wire is tight.

Not investment advice. For informational purposes only.

ERCOT’s AGUAYO_UNIT1 is reported at -$106.68/MWh. That is not a rounding error; it is the kind of print that usually means local supply is overwhelming the outlet, or a transmission bottleneck is trapping value on the wrong side of the node. The problem here is that the material does not give a primary-source confirmation for the Aguayo interval, timestamp, or component split, so the trade starts with a caution flag.

The surrounding ERCOT tape still looks noisy enough to support the thesis. ERCOT logged 100 unplanned outage events in the last 7 days and 100 SCED shadow price events in the same window, which says the system has not been idle. On the outage side, the stack includes 34 MW offline at SAMSON3, 38 MW at WHWIND, 25 MW at HILONE, 34 MW at SANTACRU, and a 75 MW forced extension outage at AMOCOOIL. On the constraint side, 6635_G carried a positive SCED shadow price of 73.863, while EASTEX showed a maximum shadow price of 5251. Taken together, that is consistent with a grid where dispatch is fighting both resource volatility and binding interfaces. It does not prove Aguayo is the pinch point, but it makes a negative node print less surprising.

For traders, the key question is whether the -$106.68/MWh came from a local congestion component or from a broader oversupply pocket. If it was congestion-led, then same-day or next-day node spreads should stay sticky and any adjacent wind or solar nodes with similar deliverability could remain weak. If it was just a transient oversupply print, the node should mean-revert faster than the constraint stack does. Watch whether Aguayo keeps repricing against nearby settlement points, and whether the shadow-price map around that interval shows a named limit that actually explains the node behavior. If not, the print is more likely a one-off market microstructure event than a durable curtailment signal.

> A -$106 node print matters only if the constraint behind it is real and repeatable.

Not investment advice. For informational purposes only.

A $373/MWh real-time spread is not background noise. It is the kind of print that tells you one node is stranded and another is clearing into scarcity, with intraday basis becoming the only trade that matters for the rest of the session.

The supporting tape in SPP is already noisy. In the last 7 days, the system logged 100 outage-capacity anomaly events and 100 curtailment anomaly events. One outage reading on 2026-04-15 20:00 showed 27,707.5 MW out, with coal at 8,644.8 MW; another showed 29,277.5 MW out, with coal at 9,099.7 MW and diesel fuel oil at 176 MW. On the wind side, redispatch curtailments were 334.93 MW at 2026-04-15 19:05 and 273.19 MW by 19:25, while manual curtailments were 0 MW in the sampled intervals. That combination is consistent with a system leaning on dispatch and topology, not just on fuel economics.

The problem: the source pack does not actually include the quoted $373/MWh spread, the exact interval, or the individual LMPs at WFECPEOPLOAD and WR.VOLT.0254. So the market read here has to stay disciplined. If the spread was driven by a binding constraint that stays active, basis should remain sharp and location-specific; if it was a one-interval routing or outage artifact, that premium can fade faster than the headline suggests. Watch whether the spread holds after the next dispatch interval, and whether any new curtailment or outage print lines up with the same corridor.

For traders, the useful question is not whether congestion exists. It does. The question is whether it is persistent enough to support repeated intraday dislocations, or merely a single-node blowout that mean reverts once the constraint relaxes.

> When SPP prints a three-hundred-plus dollar node spread, the trade is no longer power, it is topology.

Not investment advice. For informational purposes only.

$953/MWh at MISO Michigan Hub is a clean stress signal, not noise. The move was about 20x the 24-hour mean, which is the kind of ratio that usually comes from a local supply-demand mismatch, not broad system drift.

MISO’s outage tape was busy around the same window: 60 generation-outage events over the last 7 days, with forced-outage snapshots showing 6,566 MW, 12,639 MW, 2,319 MW, 3,754 MW and 7,041 MW at 2026-04-16 05:00. That is enough rotating and forced derate pressure to keep the dispatch stack brittle, especially if the problem clusters near Michigan or along export-constrained paths feeding the hub. The market also has a structural reason to stay twitchy: MISO expects load to jump 35% by 2035 on data center growth, which makes today’s tight hours easier to reproduce when outages and load stack on top of each other.

The key question for today is not whether the hub can print high again, but whether the same constraint set is still in place. If the next RT hour holds elevated compared with the 24-hour mean, then the recurrence case is intact; if prices mean-revert while outages persist, that points to a transient local issue rather than a broader scarcity regime. Watch any divergence between Michigan Hub and the wider MISO footprint: if the hub stays rich while system prices cool, the spread is telling you the bottleneck is local, not systemwide.

> $953/MWh says Michigan did not need a system emergency to clear like one.

Not investment advice. For informational purposes only.

TL;DR — MISO is investigating load-modifying resources that failed during the late January 2026 winter storm. This follows a parallel move by PJM to establish penalties for underperforming demand resources. With MISO load forecast to grow 35% by 2035, the scrutiny signals tighter future reliability rules.

On April 20, 2026, MISO launched a formal investigation into load-modifying resources (LMRs) that were unavailable during the late-January winter storm. The investigation's scope and the total MW of nonperforming resources are not yet public. Concurrently, PJM stakeholders are set to review a proposal on April 22 to establish penalties for underperforming load management and price-responsive demand resources.

LMRs are a critical reliability tool, contracted to reduce demand during system stress. Their failure during a winter event undermines resource adequacy assumptions and forces greater reliance on expensive, last-resort generation. MISO's probe, alongside PJM's penalty proposal, indicates a sector-wide shift toward enforcing performance guarantees for demand-side resources. This is especially critical as MISO expects system load to grow 35% by 2035, increasing strain on capacity margins.

Watch for rule changes that increase the cost and scrutiny of participating as an LMR. If penalties are adopted, the spread between capacity auction prices for traditional generation and demand-side resources could compress, as LMRs lose their "free option" value. The immediate risk is backwardation in forward capacity prices if the investigation reveals a material reliability shortfall. In ancillary services, expect upward pressure on Regulation and Synchronized Reserve prices as the ISO seeks more firm, fast-ramping capacity to backstop potentially unreliable demand response.

The chart shows a snapshot of significant forced and planned generation outages in MISO, highlighting the persistent background of resource unavailability the grid operator manages.

MISO's LMR probe is a warning shot: demand-side reliability is now a tradable risk.

TL;DR — Q3 2026 on-peak forwards imply a tight market, but the reserve margin does not tighten meaningfully until 2027. With 22 GW of solar and 4 GW of evening battery support expected, the scarcity premium looks overpriced.

Forward prices for ERCOT's summer 2026 strip are trading at levels that suggest a capacity-constrained year. This pricing persists despite analysis showing the grid's reserve margin remains comfortable until 2027. The market is assigning value to a scarcity event that fundamentals indicate is unlikely.

ERCOT's supply stack is transforming rapidly. Ascend Analytics expects 22 GW of solar to be added in 2026, materially easing afternoon strain. Batteries, which supplied an average of 4 GW during the critical 8:00 p.m. hour in summer 2025, will further blunt the evening ramp. This combination pushes meaningful tightening of the reserve margin out to 2027, making 2026 conditions look similar to the more relaxed 2024 environment.

The risk is being long the 2026 summer strip. The trade is to sell the Q3 2026 on-peak premium versus later years or versus expected realized prices. Watch the 2026 vs. 2027 calendar spread for widening as the market reprices the timing of the tightening cycle. The main break to this thesis is if projected data center load, which is driving a 14% demand growth forecast for 2026, materializes faster than modeled.

The chart illustrates the divergence between the forward curve's implied scarcity for 2026 and the fundamental projection of adequate reserves.

Sell 2026 summer forwards; the reserve margin doesn't bite until 2027.

For a decade, AI and power markets evolved on parallel tracks: software was eating the world while the grid was wrestling with decarbonization, aging infrastructure, and only modest demand growth.

That separation is ending fast.

This episode explores two forces that are now colliding:

AI-driven data center load is turning into a system-level demand shock in multiple U.S. regions.

A new class of “hardwired” inference chips—exemplified by Taalas—is aiming to make inference radically cheaper and faster.

The punchline: AI chips, data centers, and the grid are becoming one tightly coupled system. For U.S. power and gas markets, that coupling is likely to show up as higher volatility, wider structural spreads, and more cross-commodity complexity.

We cover:

Why Taalas matters and what it means when “the model is the computer,” with model-specific ASICs optimized for high-volume, stable workloads.

How the binding constraint is shifting from compute to power and interconnection, and why AI is a volatility multiplier rather than just “more load.”

Why more efficient chips will not simply ease grid pressure, as cheaper inference unlocks new always-on copilots, industrial optimization, and edge intelligence.

Four big changes for U.S. energy trading over the next 5–10 years: more scarcity-style hours and a fatter right tail in power prices, more valuable optionality in fast supply and flexible load, behind-the-meter generation becoming mainstream for large loads, and tighter gas–power coupling driven by gas-fired firming and LNG exports.

Where “hardwired inference” like Taalas can plug into the energy value chain: low-latency trading and bidding, autonomous price-responsive assets at the edge, and eventually utility and operator workflows once model families stabilize.

If you trade or invest in U.S. power and gas, this episode argues you should plan for persistent AI-driven demand growth, structurally elevated volatility, data centers behaving like energy-native counterparties, and a world where understanding both compute and the grid becomes a core edge.

Sources and further reading include: EIA on U.S. electricity demand growth, coverage of AI/data center load and power constraints, U.S. data center power demand projections, ERCOT queue and Texas reliability analysis, and academic work on rebound effects and AI energy use.