Pennsylvania Generation Flat at 0.3% as PJM Core Diverges

EIA data through February 2026.

What did Pennsylvania generation actually do?

Almost nothing. Over the twelve months through February 2026, Pennsylvania generation grew 0.3%, which in a series this size is close to a flat line with a rounding error attached. That is the finding, and it is the one we kept coming back to while putting the mid-Atlantic supply work together.

The reason it stuck is scale. Pennsylvania produced 246 TWh over that same twelve-month window, more than half of everything the four-state core generated. So the state that moved least is also the state that matters most to the arithmetic of the region, and the two facts sit awkwardly next to each other in every chart we drew.

Meanwhile the national benchmark grew 1.9% on the same trailing twelve-month basis, and three of the four core states cleared it, with the top of the core’s range at 5.5%. Pennsylvania generation did not clear it. That is not a small gap in a market where everything is connected to everything else.

The biggest producer in the core is the slowest mover in the core. Everything interesting in this dataset radiates out from that one sentence.

The spread across the core is wide enough that treating “the mid-Atlantic” as a single supply story stopped being defensible somewhere in the middle of our review. Anyone using a regional average as a planning input is using a number that sits between four different trajectories, and Pennsylvania generation is the one holding that average down while holding most of the volume up.

Read the full report

Who does flat Pennsylvania generation land on?

It lands hardest on anyone who has to commit capital to a location before the data settles. That is a longer list than it sounds like.

Start with the people siting large interconnection requests. If your internal model of the region says “the mid-Atlantic is adding supply,” that statement is true in aggregate and misleading in particular, and the particular is where you sign the lease. The state you are looking at may be growing quickly or may be nearly flat, and the trailing twelve-month record through February 2026 puts real distance between those two cases. A siting team working off a regional average is working off a number that describes no single state especially well.

Next, state-level policymakers and the staff who write their briefing decks. A twelve-month growth figure of 0.3% takes nothing away from Pennsylvania generation on volume; the output is the output, and it is large. What it does is set a second question beside the first one, which is whether that output is still climbing. Those are separate claims and they now have separate evidence. Anyone whose slide asserts both at once is eventually going to have to pick one.

Then there is the procurement side: industrial buyers, hyperscaler energy teams, anyone negotiating a long-dated supply arrangement in PJM territory. The thing that changes for you is not the number itself, it is which number you cite in the room. A growth rate and an output level are different arguments. Bringing the wrong one to a negotiation is a self-inflicted wound. The distinction survives contact with a spreadsheet; it doesn’t always survive contact with a summary.

Finally, the analysts and reporters who cover this region on a monthly cadence. The honest version of this beat requires deciding, once, how you are going to read a seasonal series, and then holding that decision through the months when it produces a boring headline. That decision is a real piece of work, and it is the kind of thing that separates a durable read from a fast one.

What does this change for how you read the region?

It changes the unit of analysis. “PJM” is a market; it is not a growth rate, and after this pass we do not think it can be used as one.

Our read: the four-state core has stopped behaving like a bloc, if it ever did, and the useful question is no longer whether mid-Atlantic supply is rising. It is which parts of it are rising, by how much, and whether the parts that are rising are large enough to matter against the part that is not. Pennsylvania generation at 0.3% growth over the twelve months through February 2026 is the anchor in that calculation, and an anchor is exactly what it sounds like: heavy, slow, and decisive for where the whole thing sits.

This also changes what a headline year-over-year print is worth to you. A single month against the same month a year earlier is a number that moves for reasons that have nothing to do with anyone’s buildout, and the core states produce some genuinely eye-catching single-month prints in this dataset. We chose a consistent basis for the report and applied it to every series in it, including the national benchmark.

The choice matters. Not “matters” in the sense of a footnote. Matters in the sense that two competent analysts looking at identical EIA files can produce opposite characterizations of the same state, both defensible, and only one of them useful for planning. That’s an uncomfortable thing to say out loud about public data, and it’s still true.

Why might the obvious read of Pennsylvania generation be wrong?

Because a growth leaderboard and an output leaderboard are two different documents, and they do not rank the same way.

That is the trap sitting inside this data, and it is a polite trap: it does not look like a mistake while you are making it. The tempting move is to sort the four states by percentage growth, put the fast ones at the top, and write the story from there. Do that and Pennsylvania generation ends up at the bottom of the page, described as the laggard, while the states above it get the momentum language. Sort instead by how much electricity actually came out of the ground, and the page inverts.

Neither sort is wrong. Both are incomplete, and the incompleteness runs in opposite directions, which is why we spent most of our time on the reconciliation rather than on either list. The two documents also serve different readers. A developer choosing between sites wants the direction of travel; someone sizing what already exists wants the volume. Hand either of them the other’s list and you get a confident answer to a question they didn’t ask.

There is a strong instinct, in every conversation about this region, to reach immediately for the demand-side explanation, the one involving racks and cooling and a particular Virginia county. We understand the instinct. This dataset does not test it, and we did not use it. What we did instead, including the fuel-cost series we carried alongside the generation numbers as a control, is set out in the full report with its limits stated in plain language rather than buried.

The interesting number here isn’t the fastest state. It’s the slowest one, because of how much of the region it is.

Read the full report

What question does this leave open?

The one we would most like answered: does a flat twelve months in the largest producer in the core represent a plateau or a pause?

We cannot tell you from a trailing twelve-month figure, and anyone who says they can from this data alone is selling something. What a single window can do is establish the shape of the current position clearly enough that the next few prints become informative rather than noisy. If Pennsylvania generation is genuinely plateauing while the smaller states in the core keep adding, the composition of mid-Atlantic supply shifts over time even with no dramatic monthly headline anywhere. If it is a pause, the ranking that looks so striking right now compresses back toward the middle and this whole discussion ages badly.

That is a real fork, and it is why the level matters as much as the rate. A state producing 246 TWh in twelve months does not need a large percentage move to swamp a large percentage move somewhere smaller. Working out how much it would take, and what the answer implies for anyone planning around this region, is the sort of thing that belongs in a document with a data appendix attached to it. Either way, the people who have to commit before the answer arrives are the same people who had to commit last quarter.

The secondary question, the one that will decide how this series reads a year from now, is whether the divergence across the four states is a durable structural feature or the visible edge of a shorter cycle. We have a view. We put it in the report, with the evidence that supports it and the specific conditions that would make us abandon it, because a view published without its exit conditions is just a mood.

You can check the underlying series yourself at the EIA Electricity Data Browser, which is where the monthly net generation figures behind all of this live. We would encourage it. The files are public, the collection is well documented, and there is no substitute for having looked at the shape of a series before arguing about what it means.

Get the full breakdown

This brief gave you one state and one comparison. The report gives you the rest of it: every state in the four-state core with its own trailing and single-month figures side by side, the absolute-volume comparison that reorders the growth ranking, the Henry Hub fuel-cost series we carried as a control and what we did and did not conclude from it, the explicit statement of what state-level generation figures do and do not measure, and the forward view on which parts of this divergence we expect to persist.

Read the full report on mid-Atlantic and PJM generation growth