TL;DR
- 786.0 TWh of U.S. nuclear net generation over the 12 months through April 2026, more than double solar.
- 313.1 TWh of solar over that same 12-month period, growing fast but still variable.
- 16.1% solar year-over-year growth in April 2026, a volume story for AI data center electricity, not yet a firmness story.
- $1,556.5B IT equipment and software investment at a seasonally adjusted annual rate in Q1 2026.
EIA data through April 2026.
The firmness question is the whole ballgame for AI data center electricity. A training cluster does not care that the sun is generous in July if it needs steady power through a December night. So when we look at the solar-versus-nuclear framing, we are not looking at a winner. We are looking at two very different shapes of supply, and the shape matters more than the headline growth rate.
Is nuclear really that far ahead on volume?
Yes. Over the 12 months through April 2026, U.S. nuclear net generation totaled 786.0 TWh, against solar’s 313.1 TWh over the same window. That is more than a 2x gap. Our read: the volume itself is less interesting than what sits behind it. Nuclear runs as near-constant baseload, so those terawatt-hours arrive in a flat, predictable stream. That predictability is exactly what a 24/7 compute load values, and it is not something a raw annual total captures on its own.
We want to be careful here. The dataset behind this comparison holds generation volumes and retail prices only. It carries no levelized-cost data. So nothing here is a cost verdict. It is a profile comparison, and the profile is the point.
Is solar’s growth real, or a seasonal illusion?
Both, in a sense. Solar net generation rose 16.1% year over year in April 2026, reaching 31,103.65 thousand MWh from 26,780.55 a year earlier. That is genuine, and it extends a steep multi-year climb. You can pull the underlying series straight from the EIA electricity data browser if you want to trace it yourself.
But then look at the range. Monthly solar output swung from a 36-month trough of 9.1 TWh in December 2023 to a peak of 33.4 TWh in July 2025. That is more than a 3x spread across the calendar.
Solar’s growth is a volume story. It is not yet a firm-capacity story for anything that needs power at 3 a.m.
That swing is the whole reason the firmness debate exists. A source that ranges this widely by season is doing something structurally different from one that holds near flat month to month. Neither is bad. They just are not interchangeable.
What does the AI data center electricity price actually look like?
The relevant tier for hyperscale facilities is industrial, and it sat at 8.71 cents per kWh in May 2026, up from 8.29 a year earlier. The 12-month rolling average was 8.84 cents per kWh, with a 36-month peak of 9.33 cents in July 2025. Our read: that is the number to watch for anyone modeling AI data center electricity costs, though actual facility costs depend heavily on power purchase agreement structure, so treat the retail rate as context rather than a bill.
And a house rule we hold to: retail price movement tracks with a stack of factors, general inflation, fuel costs, utility capital cycles, and possibly incremental load. We are not going to hang it on any single one.
Is the capital actually flowing behind all this?
It is, at least at the economy-wide level. Information processing equipment and software investment hit a $1,556.5B seasonally adjusted annual rate in Q1 2026, up 18.7% year over year and 6.1% from the prior quarter. This is a macro capital signal, not a data-center-specific line item, so we read it as backdrop rather than proof. The FRED series sits behind this figure if you want the full history from the Federal Reserve’s FRED database.
Zoom out and the grid context tracks with the buildout narrative. U.S. total net generation reached 4,456.3 TWh over the 12 months through April 2026. Texas generation rose 7.2% year over year in April 2026, totaling 594.8 TWh over the trailing 12 months. The load is growing, the capital is flowing, and the two generation profiles at the center of the siting debate look nothing alike.
So where does that leave the solar-versus-nuclear call?
Undecided, and honestly that is the correct answer. Our read: solar’s 16.1% growth is real and worth respecting, but the 9.1-to-33.4 TWh seasonal swing keeps it in the volume column, not the firmness column, for now. Nuclear’s 786.0 TWh arrives flat and predictable, which is what around-the-clock compute wants. Without cost data, we will not pretend to render a verdict on AI data center electricity economics. We will just say the shapes matter, and the shapes are not converging.
—
The full report walks through the national baseline, the industrial price context, the Virginia and Texas load lens, and the IT investment signal behind the buildout. Read the full comparison of solar vs nuclear for data centers.