Moore's law will eventually stop, too. But people have been predicting the end for 50 years and it still looks like it has at least another 10 to go. It seems quite clear that solar costs will continue to drop exponentially until the end of the decade, which is all the author's assumptions need.
Land in a desert can cost as little as a few hundred dollars per acre, so it's far from being a dominant factor. And in other places solar is often a secondary use of the land.
The article suggested local production, good luck buying land anywhere close to NYC at 100$/acre.
Mores law also failed several times. More revised his initial 1965 estimate for a doubling every year in 1975. He predicted various doubling rates with the weakest form being transistors on a single chip every 2 years starting in 1980.
So, it’s current form isn’t an exponential increase in density but in terms of transistors on a single chip. If density actually doubled every 2 years then starting from 1971 an Intel 4004 @ 188 t/mm then we should have hit 6 million t/mm in 2001 and 197 t/mm in 2011. Except chips only broke 6 million in 2012 (11 years late) and are nowhere near 197 million t/mm in 2022.
I don't think I've posted any land cost estimates, elsewhere in this thread I mentioned how many warehouse roofs in NJ, Staten island and long island have no solar on them.
You're definitely not going to power all of NYC by rooftop solar, but also look at the size of the warehouse roofs (empty, no solar) in NJ just across the river, and out on long island. And Staten island. Etc.
Empty land is not necessarily a requirement.
Parking lot per-row shade structures that integrate PV panels on top are also a thing now, and there are sure a shitload of parking lots in NJ and out on long island.
I'm not from the area, but isn't Manhattan well known for its tall buildings? How would those buildings' shade affect solar panels closer to ground level, e.g. warehouse roofs and parking shade structures?
Rooftop solar is far more expensive than solar put on land. The cost of rooftop solar (particularly homeowner solar) is usually hidden by implementing a big subsidy to wealthy consumers that is paid for by less wealthy consumers - sort of a reverse robin hood scheme.
We are well past the point where anyone can argue we should subsidize rooftop solar because we need to help build a fledgling industry - yet I see solar advocates continue to advocate for this wealth transfer from the poor to the well off.
Rooftops are unused space, but the land use is generally a small part of the cost. Utility scale solar has real economies of scale compared to one-off small installations on someone's roof.
Moore's law already stopped - since Intel's 10nm process was delayed by 5 years. The exponential grown continues, but a lot slower than predicted by Moore.
Even that "exponential" growth is a not a thing anymore. There are factors which make it less obvious (growing TDPs of modern CPUs and GPUs; going wider, i.e. increasing the number of CPU cores, instead of faster; making chiplets and building cpus out of them instead of big single dies), but these factors only postpone the inevitable (there's only so many levers you can pull and most of them have already been pulled). We're a couple years away from the silicon limit, the 2nm situation already looks grim.
Land in a desert can cost as little as a few hundred dollars per acre, so it's far from being a dominant factor. And in other places solar is often a secondary use of the land.