On November 2, Energy and Environmental Economics — E3 — released the second edition of its 2026 Core Case Electricity Market Price Forecasts, covering every major North American wholesale market: CAISO, ERCOT, SPP, MISO, PJM, NYISO, ISO New England, Alberta’s AESO, Ontario’s IESO, plus the Pacific Northwest and the Southwest. The forecast is a commercial product, delivered for the first time through a new E3 Forecast Portal, and every figure in it is E3’s — a single scenario, the Core Case, not a consensus. What makes it worth a market participant’s attention is not any single number. It is the shape of the next thirty years: where the forecast expects scarcity to live, and which of its own assumptions do the work of producing that answer.

What the Core Case is

E3 updates its Core Case forecasts twice a year. For this edition, the team refreshed every major input — load, fuel prices, near-term supply, resource costs, transmission limits, and policy targets — and modeled every market hour by hour across a 30-year horizon. Near-term prices are calibrated against recent market outcomes, capacity prices come with a range of scenarios, and each resource’s contribution to reliability is based on E3’s own analysis rather than a generic accreditation table. That calibration matters: the front years of the forecast are anchored to the markets as they actually cleared, which is why the near-term findings read as extensions of today’s conditions rather than leaps into a different world.

The delivery mechanism is itself part of the story. The Forecast Portal ties hourly prices to the load, dispatch, and renewable output behind them, in linked panels that share a single time window — so a price spike in a given week can be traced back to the net load, renewable output, and dispatch behind it. That is a different way of consuming a forecast than the traditional bound report: it invites the reader to interrogate the mechanism, not just accept the price path.

ERCOT: load doubles, and scarcity moves to winter

The headline finding for Texas is the scale of the load forecast. In the Core Case, ERCOT load roughly doubles by 2055. Baseload growth from population and electrification is the largest piece of that increase, while large flexible loads and data centers drive the near-term acceleration. The composition matters, because the two drivers imply different kinds of system stress. Population and electrification growth is broad-based and weather-driven: more people, more buildings, more electric heating and cooling, all rising and falling with the thermometer. Data-center growth is concentrated, lumpy, and around the clock — load that does not sleep, does not cool off at night, and does not respond much to price.

On the supply side, E3 expects solar to lead near-term additions in ERCOT, with wind and solar making up most long-run capacity. New gas is paced by turbine backlogs and cost — a constraint that appears nowhere in the physics of the grid and everywhere in the economics of building one.

The finding that matters most for anyone who trades or hedges ERCOT is the migration of scarcity. As heating load and around-the-clock data-center demand grow, E3 expects scarcity pricing to shift from summer evenings toward winter mornings and evenings. The summer-evening spike is the pattern every ERCOT participant knows by heart: solar fades, air conditioning holds, and the operating reserve demand curve — the ORDC, the administratively determined price adder that rises steeply as reserves thin — does its work in a handful of late-afternoon intervals. A winter scarcity pattern is a different animal. Winter demand peaks in the morning, when people wake and heat their homes, and again in the evening; solar contributes little in either window; and the failure mode is not a few hot hours but a multi-day cold event in which forced outages compound and gas supply itself tightens.

The winter problem, stated plainly

This is where the forecast connects directly to the scarcity anatomy we have covered before. A summer spike is usually a routine event: real-time stress, brief and localized, the system working as designed. A winter event is where sustained scarcity lives — reserves eroding over days, emergency procedures, and the kind of price persistence that tests balance sheets rather than just trading desks. The February 2021 freeze is the standing example of what winter scarcity costs when it arrives unprepared, and the market’s memory of it is one reason winter risk now commands a premium in forward curves.

E3’s forecast does not say a freeze is coming. It says something more structural and more demanding: that the routine scarcity in ERCOT — the scarcity that shows up in an ordinary year, not just in an extreme one — is drifting toward the season in which the system is least forgiving. A participant who has built a hedging book around summer-evening exposure, who owns storage sized for a four-hour evening discharge, or who holds demand response that can only be called in the afternoon, should read that finding as a prompt to re-examine the shape of their own exposure.

PJM: the shortfall is the story

If ERCOT’s finding is a migration, PJM’s is a collision. E3 forecasts that large loads in PJM will grow more than 10 percent a year from 2027 to 2035, and that the capacity market absorbs most of that pressure. The forecast puts the unforced capacity shortfall at a peak near 13 GW in the 2029/30 auction, with prices elevated but capped through 2031, before new supply closes the gap in the 2030s.

Start with the load number, because its compounding is what makes it formidable. Ten percent annual growth over eight years roughly doubles the large-load base, and it does so in a market where the capacity auction is the mechanism that is supposed to translate that growth into a build signal. The 13 GW figure is E3’s estimate of the unforced capacity shortfall — the gap between the capacity the market needs for reliability and the capacity that is actually committed — at its worst point, in the 2029/30 delivery-year auction. That is not a forecast of blackouts. It is a forecast of a market clearing short of its reliability target, with the consequences showing up in prices rather than in the lights going out.

The qualifier that follows — prices elevated but capped through 2031 — is doing as much analytical work as the shortfall number itself. A cap protects consumers and load-serving entities from the full scarcity price, but it also blunts the signal that would otherwise summon new supply faster. That is the same tradeoff we have described for energy price caps: the cap is a societal choice about who bears the risk of extreme conditions, and it is paid for in the currency of deferred investment.

That is the question the forecast cannot answer, because it is a question about institutions, not about load curves. Interconnection reform, state siting decisions, and the pace at which data-center developers energize the load they have announced all sit outside the model, or inside it only as assumptions. The forecast’s value is that it sizes the problem: a shortfall peaking near 13 GW is a problem of a different order than a shortfall of 2 or 3 GW, and it persists for several delivery years rather than one.

CAISO: the counter-example

The California findings in the same release are a useful counter-example, because they show the same load-growth story playing out with a different shape. In E3’s Core Case, electrification — mostly electric vehicles — drives California’s load growth, adding about 47 TWh by 2035, compared with roughly 4 TWh from data centers. Solar becomes the largest source of generation, and batteries shift the midday surplus into the evening, compressing daily price spreads. By 2035, E3 expects the evening peak in SP15 average hourly prices to have largely flattened.

Contrast that with ERCOT and PJM, and the lesson is that “load growth” is not one phenomenon. In California, the growth is diffuse and transportation-driven, arriving at hours when the system has a solar surplus to absorb it, and the flexibility response — batteries — is already scaling fast enough to arbitrage the spread it creates. In Texas and PJM, the growth is concentrated, around the clock, and arriving faster than the flexible supply that would tame it.

What a 30-year forecast can — and cannot — tell you

A forecast of this kind is most useful as a map of where the binding constraints will sit. It tells you, in E3’s view, that ERCOT’s binding hours move from summer evenings to winter mornings and evenings; that PJM’s binding constraint in the late 2020s is capacity rather than energy, and that it peaks around the 2029/30 auction; and that CAISO’s evening peak, the market’s defining feature for a decade, flattens as batteries scale. Those are structural claims, and they are the kind of claims a thirty-year model is equipped to make — because they follow from the shape of load growth and the shape of the resource mix, not from any particular year’s weather or fuel price.

What it cannot tell you is the price in any particular hour, week, or year. The hourly resolution is real — the model simulates each hour — but uncertainty compounds with the horizon. A load forecast for 2055 is a statement about demographics, electrification policy, and data-center buildout three decades out; a small error in the growth rate compounds into a large error in the level. It is a Core Case, one scenario among many, not a price guarantee.

The assumptions deserve the same scrutiny as the conclusions. Gas turbine backlogs pacing new gas is one: if the backlog clears faster than E3 assumes, or if developers substitute batteries and other resources for the missing gas, the PJM gap closes sooner and the ERCOT winter scarcity is milder. The pace of data-center energization is another: announced load and energized load are different things, and a forecast that takes the interconnection queue at face value will overstate near-term demand if projects slip. Policy targets are a third: the forecast refreshes them each edition, so the Core Case moves with the politics.

How to use it, and how not to

The disciplined way to use a forecast like this is as a stress test for decisions that are hard to reverse. A developer choosing between a battery and a gas peaker in Texas should ask how each performs under E3’s winter-scarcity world, and also under the world where that migration is slower than forecast. A load-serving entity in PJM hedging capacity exposure through 2031 should ask what its costs look like if the shortfall peaks at 13 GW and the cap binds, and what they look like if new supply arrives two years early. In each case the forecast is not the answer; it is the scenario against which the answer is checked.

The undisciplined way is to treat the price series as a forward curve — to lift the 2035 hourly prices out of the portal and plug them into a revenue model as though they were market prices. They are not. They are one forecaster’s simulation of one scenario, calibrated at the front and speculative at the back, and they carry no counterparty. A forward contract is a commitment; a forecast is an opinion, however well-modeled. The portal’s design encourages the disciplined use, because it makes the mechanism visible: the participant who traces a price spike back to the net load and dispatch in the same hours is using the forecast correctly.

Reading the forecast like an analyst

When a new forecast vintage lands, ask four questions. What changed since the last edition? E3 updates the Core Case twice a year and refreshed every major input for this one; the deltas between vintages are often more informative than the levels, because they show which assumptions the forecaster itself has revised. Where is the scarcity, and when? The location and timing of tight hours matter more than the average price — a forecast that moves scarcity from summer to winter is telling you something about the shape of the system, not just its level. Which assumption is doing the work? Turbine backlogs, data-center energization, policy targets: identify the two or three inputs that, if wrong, would change the conclusion, and track those inputs yourself. What would falsify it? A forecast that cannot be falsified cannot be used. If PJM’s 2029/30 auction clears without a shortfall, or if ERCOT’s next two winters pass without scarcity migrating, the Core Case needs revision.

E3’s second-edition 2026 forecasts will be superseded, as every forecast is, by the next edition’s refresh. The numbers will move. What is less likely to move is the underlying story: load is growing fast and unevenly, the flexible supply that would tame it is arriving on someone else’s schedule, and scarcity — the market’s rationing mechanism, the signal that pays for reliability — is migrating to the hours and seasons where the system is least prepared for it. The prices are E3’s; the questions they raise belong to everyone who participates in these markets.