Rooftop Solar Has Flipped New York’s Morning Power Demand Negative and Tripled Its Evening Ramp

In spring 2018, metered electricity demand across the New York Independent System Operator rose by an average of 850 megawatts during the 8 to 11 a.m. window. In the same spring months of 2026, it fell by 923 megawatts.

The grid did not shrink. Roughly 2.8 gigawatts of small-scale solar, most of it rooftop systems under one megawatt, came online in New York between 2018 and 2026. Because those systems sit behind the customer meter, their output never registers as supply on the wholesale system. It registers as metered demand that disappeared.

That is the finding in a late-June Energy Information Administration analysis of NYISO load, comparing spring demand profiles across the two periods. Metered demand is what the grid operator sees after behind-the-meter generation has already been netted out. As rooftop output climbs through the morning, the load the system must serve falls below where it used to sit.

The evening. As small-scale solar generation falls off in the late afternoon, the demand it had been masking returns at once. The 4 to 7 p.m. ramp, which averaged a 681 MW increase in spring 2018, averaged a 2,221 MW increase in spring 2026. The evening climb more than tripled.

The net result is a load curve that sags through the middle of the day and then climbs steeply into the evening, concentrating the system’s effective peak in the hours after solar fades.

The duck curve. California’s grid operator, CAISO, named this shape the duck curve roughly a decade ago: a belly of suppressed midday demand, a steep neck of evening ramp. New York was expected to be different. It has less sun, a winter-weighted peak, and a dual-peaking daily profile that does not resemble California’s summer-dominated grid.

The NYISO data shows the same inversion arriving anyway. What distinguishes the New York version is the steepness of the evening ramp relative to the depth of the midday belly. The morning swing of roughly 1,800 MW between 2018 and 2026 is large, but the tripling of the evening climb is the part that strains the system, because it has to be met fast.

Where the kilowatts land. Commercial electricity bills in New York and most of the country are not dominated by how much energy a building consumes. They are dominated by its highest sustained draw, the demand charge, billed in dollars per kilowatt and frequently tied to a defined afternoon or coincident-peak window.

When the system peak sat near midday, a commercial building’s costliest kilowatts and the grid’s costliest hours roughly coincided, and on-site solar shaved both at the same time. As the coincident peak migrates toward 5 and 6 p.m., that alignment breaks. An office tower or a hospital is still fully occupied at 6 p.m. The rooftop array that had been trimming its midday peak has gone dark.

The kilowatts that set the bill now arrive in a window solar cannot reach.

This rewrites an assumption inherited from the solar era. A battery sized and scheduled to discharge into a midday peak is optimized for a curve that no longer describes New York. The value has moved to the evening ramp, and dispatch logic has to move with it. This reading is interpretation rather than something the EIA data states directly, but it follows from the same arithmetic: the peak that prices a commercial bill is now landing later in the day than the design conventions built around it assume.

The ramp. The steepest feature in the EIA numbers is the speed of the evening recovery. Meeting a 2,221 MW climb across three hours is a question of how quickly capacity can come online, not how much energy is consumed across the day.

Power, measured in kilowatts, is what demand charges price and what fast-responding resources address. A grid that has to climb 2.2 gigawatts in three hours values the ability to inject capacity on that timescale, whether the injection comes from a peaker, an import, or a building discharging behind its own meter.

Rate design. Commercial demand charges are frequently tied to a defined peak window. When a load curve’s most expensive hour moves later in the day, the windows drawn around an earlier peak no longer sit on top of it. Whether and how quickly tariff windows are redrawn to follow the migrating net peak is a regulatory question the EIA analysis does not address; the data establishes only that the underlying load shape has shifted. Any claim that existing Northeastern rate windows already lag the curve is inference, not a finding in the source.

What the data does support is directional. The states approaching the rooftop-solar penetration that bends the curve, across New England and the Mid-Atlantic, sit earlier on the same trajectory New York has now traveled.

There is a quieter point in the data for anyone reading it as a forecast rather than a snapshot. The 2018-to-2026 inversion happened with 2.8 GW of small-scale solar. New York’s distributed solar pipeline does not stop there, which means the belly deepens and the neck steepens from here, not the reverse.

In California, the duck curve became a generation story: oversupply, curtailment, and negative midday prices. In a grid like New York’s, where commercial bills turn on peak kilowatts rather than total energy, the same curve is a timing story. The system peak did not grow larger between 2018 and 2026. It moved roughly three hours later, into the part of the day when the sun has set but the building is still full.

Sources