> But they are being used, every night. And then you can also use them during the day on the minority of days when it's cloudy.

But they're not being used every night. Night load is mostly carried by combined-cycle, nuclear and wind. Peaker plants are expensive to operate compared to combined-cycle, so they will get turned off as soon as possible, so that the grid relies on less expensive sources for base load operations.

Okay, so let's go to the interactive view of that map at https://cleanview.co/power-projects/operating/solar-farms/us

It's basically just a whole bunch of single-colored dots on a map. That doesn't tell you when a given solar farm was installed, just by looking at it. And it doesn't tell you a whole lot about the capacity. If you zoom in, you can see that most of those dots are under 50MW in size. When I zoomed in and looked at a sample, most of them seemed to be single-digit MW ratings. What we really need to count is capacity, not number of dots.

Unfortunately, this page doesn't actually give us a lot of detail. But let's look at the detail it does give us. First off, let's look at the ten most recent solar farm installations that they show and where they are:

Five are in Texas, in the counties of Yoakum, Ector, Milam, Kaufman, and Upton. Two are in Illinois, in the counties of Pulaski and Jefferson. One is in Arizona. One is in Colorado. One is in Michigan. Of all of those, the ones in Colorado and Michigan are at about 40 degrees north latitude or higher, so that would qualify as "in the north". Pulaski and Jefferson counties in Illinois are below 40 degrees north latitude, so I count them "in the south".

Okay, so now let's look at where the ten biggest solar farms are:

Three are in Texas, in the counties of Jack, Swisher, and Webb. Two are in Arizona, in the counties of Maricopa, and La Paz. One is in Nevada. One is in California. One is in Illinois. One is in Ohio. And one is in Utah. Of those, the three in Illinois, Ohio, and Utah are all about 40 degrees north latitude, so they can be considered to be "in the north".

So, while we don't have the full dataset available to us to show the actual distribution of when and where grid-scale solar farms are being installed, the tiny amount of data we do have from this website clearly shows that the majority of the new grid-scale solar farms that are installed or being installed are going into the South. Texas alone has 50% of the newest solar farm installations, according to that list.

According to EIA, Texas alone has 40% of the planned new solar farm installations (see https://www.eia.gov/TODAYinENERGY/detail.php?id=67205), out of the 43.4GW of planned new solar farm capacity to be installed across the country. That would give us another ~17.36GW of solar to be added in Texas alone.

There's also 24GW of battery capacity to be added this year, of which 53% is planned for Texas alone (12.9GW), and 80% of the total across just three states: Texas, California, and Arizona. Solar is happening, yes. But so are batteries.

All of which is in line with what I said before.

But all that said, we should be looking at total average amounts of sunshine, and not north versus south. Anything at or above about an average of five hours of sunshine per day is considered "sunny", and the order here is Arizona being the sunniest, followed by Utah, Colorado, and Texas. Illinois, Michigan, and Ohio are in the "low-sun" group.

By that criteria, of the ten largest solar farms, only two of ten are in low-sun states: Illinois and Ohio. For share of total solar farm capacity of the top ten, that would be about 20% (1,170 of 5,786 MW).

Of the ten most recent solar farms to be installed, only three of them are in low-sun states: Illinois twice and Michigan. For share of capacity of the ten newest solar farms, that would be about 27% (692 of 2,569 MW).

So, we're basically back to where we were, but now on more solid footing. Instead of north/south, we're looking at expected average sunny hours per day (a much more reliable measure), and we're also now measuring the actual capacities of each group across the sets.

If you want to run these counties through the NREL data (https://developer.nlr.gov/docs/solar/pvwatts/v8/), then you find they break out into three groups. The sites in Arizona are the best, with West Texas, California, and Nevada not far behind -- 95% or better, compared to the best. Central and East Texas, Colorado, and Utah come in behind that, from 80-90% of the best. The Midwest is 75% of the best or lower.

For the ten largest sites, 51% of the capacity is installed in the top tier sites, 29% capacity in the second tier sites, and only 20% in the third tier sites.

For the ten newest sites, 47% of the capacity is installed in the top tier sites, 26% capacity in the second tier sites, and only 27% in the third tier sites.

So, we get back to my belief that capacity skews toward sunnier sites, for a reason. Yes, you can build in sites that are not as sunny, if you're willing to pay the price. But the incentives are strong.

Now, getting back to the duck curve -- it is predictable enough that the grid operators can get nuclear and combined-cycle natural gas turbine powerplants started far enough in advance that they will be ready to take over when the solar farms are giving out. Depending on how good their prediction is, they might need to bring up a few peaker plants to cover the intermediate term, but then again maybe not.

They don't run peaker plants unless they have to, because they are more expensive than combined-cycle plants, even though they're both turbine-based systems running on natural gas. The problem with combined-cycle plants is that they take two to three times longer to turn up compared to peaker plants, and longer to curtail, and that's why they need peaker plants to fill in gaps.

For solar and wind plants to take a bigger percentage of the grid production, they need to operate more reliably, and more like base load plants. Which means that they need buffers. The advantage is that we may also be able to use them like peaker plants as well, if we're willing to pay to build bigger buffers than we would otherwise need.