Solar Is Cheapest Electricity in History, U.S. DOE Aims to Cut Costs 60% by 2030

> These will start to be important problems when the quantity of power produced by solar panels is about 100 times larger than current world marketed energy consumption. I expect that this will happen in about 30 years.

You predict energy needs will increase 100x in 30 years? Surely you mean just solar energy production?

Production and consumption are objective. Needs are subjective. The only objective necessity for a living thing is its death — everything else is optional. To think otherwise is wishful thinking.

The behavior of living systems is to expand when not constrained by resources; the human economy has been constrained by fossil fuels for 250 years, due to its inability to take advantage of solar energy, much as it was constrained by agricultural production for the preceding 12000 years, with occasional exceptions like petroleum-drilling-fueled salt refining in the Song dynasty. Since about 02015, solar energy has been brought within within the scope of what the human economy can effectively consume directly, rather than through agriculture.

Very few people have noticed this yet or understand what it means; it's still common to hear foolish remarks like https://news.ycombinator.com/item?id=26220534 "I don't actually see where solar and wind are actually powering a modern economy. I see a lot hope and handwaving." The early stages of exponential growth are indistinguishable from the early stages of sigmoid growth that's an order of magnitude or more from its asymptote; you can't simply extrapolate the growth empirically. You need to understand the underlying dynamics of the system. And so it's very easy to fool yourself, whether out of wishful thinking, vulnerability to manipulation by others, or simple random error. And so far solar energy is under 10% of world electricity generation and under 3% of the IEA's world marketed energy consumption.

So I could be mistaken. Although the solar resource is three orders of magnitude larger than current world marketed energy consumption, maybe there's some limiting factor that will choke off the consumption of solar energy through photovoltaic cells. The most ignorant have suggested that rare-earth metals are such a limiting factor, unaware that solar panels do not use any rare-earth metals. Less absurd is silver: current silicon solar cells use screen-printed silver-paste electrodes, which accounts for some 10% of the cost of the cell and some 10% of world silver mining, so the next order-of-magnitude increase in solar-panel production will probably require the substitution of abundant copper, which will reduce the cells' efficiency.

But the most plausible limitation is storage — a solar power plant is not a direct replacement for a coal power plant unless it's coupled with some kind of utility-scale energy-storage system, which considerably reduces its cost advantage relative to thermal generation stations.

But this is only a limitation insofar as scalable consumers of such intermittent power fail to appear. Traditionally, for example, people would work during the day, leaving their tools idle at night, but this becomes less economically appealing for more capital-intensive forms of production, because they increase the capital cost of leaving your capital goods idle one-third or two-thirds of the time, increasing capital inputs per unit of production by respectively 50% and 200%. Solar-powered industry without enough energy storage to last it through the night and through cloudy days will thus have to pay higher costs of capital per unit of production.

But it seems implausible to me that no profitable and scalable industries exist for which the cost savings from near-zero-cost energy would exceed the cost savings from 24/7 productivity.

So, are there other limiting factors I don't know about?

It may be hard to imagine what humans will use 100 GW or 1000 GW on. But in 01800 it was hard to imagine what we would use 1 GW on (if we don't count agricultural production, which the IEA doesn't). Steam-engines were stationary machines, used mostly to pump water out of mines, and in some cases to drive looms in manufactories; the steam locomotive hadn't been invented yet. Steam-ships had been conclusively shown to be impractical by the disastrous experiments of Papin, Allen, Hulls, Henry, and Fitch; Henry's boat had sunk when he tried to put a steam-engine in it. Fitch's boat at least didn't sink, but his fares couldn't pay the heavy expenses required by the steam-engine. Doctors expressed skepticism about whether the human body could withstand the unbelievable velocities some of the wilder "engineers" were talking about, such as 30 miles per hour or even more. Fulton had met Henry, but hadn't yet seen a steamboat, much less built one. Steam-engines were also notorious for exploding, killing people en masse, and filling their surroundings with poisonous fumes; many doubted their use would ever be widespread.

Yet in 01830 the B&O Railroad was running the 1-kilowatt Tom Thumb steam locomotive down its 23 miles of track (37 km in non-medieval units) at 18 mph (8 m/s) https://en.wikipedia.org/wiki/Baltimore_and_Ohio_Railroad#Ea..., and similar lines were running in England and France. Steam-ships were starting to cross the Atlantic, cutting the transit time to a mere month, and paddle-powered steam-boats plied the Thames, the Seine, the Ohio, the Mississippi, and the Great Lakes. Such is the impact of the advent of a new source of energy.

Remember that in the 01950s von Neumann reprimanded one of his graduate students for writing a compiler, saying that a valuable scientific instrument like the computer should not be wasted on clerical work. What would von Neumann have thought of https://hackaday.com/2021/03/26/nixie-shot-timer-adds-useful..., where a computer runs 16 million instructions per second to detect when a pump has turned on in an espresso machine? Could even such a great mind as von Neumann have imagined such a thing, much less condoned such an irresponsible waste of precious computation?

So we should expect that in 02051 people will be using cheap solar energy for innumerable purposes that today would seem absurdly profligate.

It's also possible that world wars, pandemics, global dictatorships, or other civilization-collapsing events will slow or stop the growth in human use of solar energy. But it seems probable that, barring such calamities, solar energy production will continue to grow until it's a significant percentage of total terrestrial insolation, which is the point at which the plant-shading and heat-retention effects start to become significant.

That was a lot to read to figure out that you did mean you think energy consumption will increase 100x in 30 years.

That's crazy. You do realize electricity consumption has actually declined in the US in 7/10 years over the last decade, despite a growing population and economy [1]. Worldwide as standard of living catches up with the US, electricity consumption will increase - but it would take a massive unforeseen demand to make your prediction come true. I don't see it happening.

[1] https://www.eia.gov/energyexplained/electricity/use-of-elect...

"That's crazy" isn't a rebuttal; it's just a statement of surprise. But I already know that what I said was surprising. As I said, in 01800 it was "crazy" to contemplate people safely traveling at 30 miles per hour (13 m/s in non-medieval units) or crossing the Atlantic under steam, and in 01830 it was being done, if not yet commonplace. In 01954 it was "crazy" to use a computer for "clerical work" like compilation; in 01984 children played with computers in schools, running BASIC interpreters and paint programs. How could we go about figuring out what's really going to happen in 30 years? It certainly isn't by simply believing the least surprising ideas. Instead, I suggest we explore the underlying dynamics of the system.

You didn't answer my question. Are there other limiting factors I don't know about? That wasn't a rhetorical question. I'm very interested to know what other possible limits we might encounter between here and Kardashev Type 1.

— ⁂ —

You seem to have shifted from discussing energy consumption to discussing electricity consumption; perhaps you aren't aware that there's a difference. Marketed energy consumption in the US has actually increased over the last 10 years by about 15%, from about 90 EJ/year (2.9 TW) to about 105 EJ/year (3.3 TW)—again, excluding agricultural products. Only about a third of current energy consumption in the US is electrical.

The page you linked says US electrical energy consumption is only 450 GW or so, which would be under 15% of the total, but that's after the efficiency losses from electrical generation and distribution. It also points out that US electrical energy consumption has increased in 59 of the last 69 years, with a total increase of 13× during that time. Total US energy use, however, has only increased by 3× during that time; it doubled roughly from 01950 to 01970 (from about 1 TW to about 2 TW), and has increased by about 50% since then, four times slower. What might account for this slowdown?

Well, energy prices have been relatively high and volatile since 01970. Looking at the last ten years from 02011 to 02014, WTI was in the US$70–100/bbl range, although since then it's dropped below US$50 and very briefly to US$20 at the beginning of the covid pandemic. By contrast, in the 01945–01970 period before the energy crisis, it was fairly stable in the US$20–30 range (adjusted for inflation), and since then it's mostly been US$40–100. So it's unsurprising, in retrospect, that in the last 50 years there's been strong pressure to conserve energy: energy cost about four times as much, so energy use grew about four times slower.

There are places, such as Germany, that have been shifting their economic production to lower-energy-intensity sectors of the economy and higher-efficiency ways of using energy, with the consequence that their energy use has actually declined. This is partly driven by energy prices, but more by strong conservation efforts led by the Green Party, which have been critical to the current dramatic reduction in the cost of photovoltaic energy.

And precisely what we're discussing here is that solar photovoltaic energy is now dramatically cheaper than fossil-fuel sources of energy, even in some non-tropical countries like the US. And the total solar resource is, as I said, about a thousand times larger than total world marketed energy consumption. So we should expect dramatic growth in energy consumption in the next decades. Maybe not in Germany, though.

— ⁂ —

As for your plaint about the length, that was under 1000 words! It takes literally under three minutes to read. If you think that's a lot to read, I'd hate to see how you react to a so-called book. I know this is probably hard to imagine, but I regularly spend not just three consecutive minutes reading a text written in one of these "books", but literally multiple hours. Sometimes they don't even have pie charts in bright primary colors! Terrifying, I know. But it's important, because when I read things and carefully consider them, making calculations, I find out when I'm wrong, which is very often. (So-called "research papers" are even more important, because they're more up to date, but they take longer to read even though they're shorter.) Then I change my opinion to be less wrong. I recommend trying it!

However, it does have the disadvantage that it forces me to abandon popular beliefs that can't possibly be correct in favor of "crazy" ones that are backed up by evidence and reasoning. Sometimes, to my surprise, the popular beliefs turn out to have been correct after all. More often reality turns out to have been just as crazy as I thought.

— ⁂ —

So, I don't know what will happen over the next 30 years, but I think a 100× increase in world marketed energy consumption due to PV is eminently plausible. Maybe it'll only be 3×, maybe 1000×. Maybe there will be another world war, worse than the last two, and world marketed energy consumption will drop 10×. But it is not plausible that things will continue as they have for the last 50 years.

> I think a 100× increase in world marketed energy consumption due to PV is eminently plausible.

It's not plausible, and despite your verbosity you offered no hypothesis as to why energy needs might increase 100x in just 30 years.

It's not impossible, but there is no reason to expect that to happen - and you offer none, which I think is the minimum required to support your argument.

That demand will increase 100x simply because solar energy is cheaper is not an argument.

> you offered no hypothesis as to why energy needs might increase 100x in just 30 years

Right, because I was un-asking the question. There's no such thing as "energy needs", just energy conversion and dissipation, so your question is nonsensical. Up a couple of levels I linked to Marco Schulte, who's using an Arduino Nano, with maybe half a million transistors in it, to detect when his espresso machine turns on and count up the seconds of espresso brewing on Nixie tubes.

Have "transistor needs" thereby increased by half a million transistors? No, clearly this could have been done with 50 transistors or less. Or not done at all; plenty of espresso machines get by without timers protruding from the top, and plenty of people get by without coffee. There was no need, just consumption. But at this point we have something like 15 or 20 sextillion transistors in the world, so speaking of "transistor needs" is nonsense. Transistors are not rationed like covid vaccines in a backward country, where you only get to use as many as you can prove you need. They're still not free, though—though you can use half a million transistors as easily and cheaply as you can use one, a chip with sixteen billion transistors costs a little more than a chip with half a million, and a machine with a hundred billion transistors (16 GiB of RAM or Flash, for example) costs more than that and also takes up space and uses a significant amount of power. And once you're up into the trillions of transistors the cost goes up linearly with transistor count.

So, in short, Marco Schulte is using half a million transistors to detect his espresso machine turning on because transistors are cheap enough that he doesn't get any advantage he cares about by using less.

This phenomenon is not limited to transistors. Demand for just about anything will increase if it gets cheap enough. Forgive me if I point out that this assumption, that demand curves are generally downward-sloping, is fundamental to economic theory; it's not something I need to offer an argument for in a particular case, because the exceptions are few and far between.

— ⁂ —

Of course, that doesn't tell us how much the demand will increase. The 100× ballpark comes from taking, not the 39% yearly exponential growth in installed photovoltaic capacity over the period 01993 to 02018 https://commons.wikimedia.org/wiki/File:PV_cume_semi_log_cha..., but the much slower 23% yearly exponential growth we've seen over the last few years; extrapolating it 30 years into the future; and then dividing by 5 to make the estimate "conservative." Of course it's only really conservative if we don't hit some limiting factor before that point, as we did, for example, with oil in the 01970s. That exponential growth will surely level off at some point, at which point such empirical extrapolations become nonsense.

So what's the limiting factor? I don't know.

— ⁂ —

What will people be using abundant solar photovoltaic energy for in 30 years? Probably mostly things that are too absurd for us to imagine now, like Marco Schulte's espresso machine, whose coffee timer has roughly the computational power of the 24 AN/FSQ-7 computers that made up NORAD's missile defense system in 01965. I can come up with lots of uses for lots of energy that aren't so absurd, though they aren't likely to be the real answer:

- Mining Bitcoin. There's no limit to how much energy you can spend on that, because it's purely competitive, like soccer. It takes all the hashing you can do just to stay in the same place with respect to the other miners.

- Making more solar cells. The raw materials (except silver) are abundant, and the processing is highly automated, but it takes a lot of energy. So having a lot of solar cells sitting around producing cheap electricity makes it cheaper to make more solar cells, which in turn makes the electricity they produce even cheaper.

- Smelting aluminum. Currently 20%–40% of the cost of smelting aluminum is just the cost of the energy, but it's only that low because aluminum smelting pots are designed to be efficient and not waste too much energy. You can always trade off some efficiency for other desirable attributes of the design, like cheapness. Probably this would reduce the cost of aluminum per unit of strength below the cost of steel.

- Smelting ferrosilicon, which is used as a feedstock for, for example, magnesium. More generally all kinds of mining and smelting processes use a lot of energy, and have even energy-hungrier alternative processes that we don't use because they're more expensive than the ones we do use.

- Desalination for irrigation. The Sorek reverse-osmosis plant produces drinking water at a total cost of US$0.58/kℓ at 70 atmospheres, which is 7.1 kJ/ℓ; energy is something like a third of the cost, and as with aluminum, you can presumably make other aspects of the plant cheaper if you can afford to waste more energy. Suppose this less efficient design uses 20 kJ/ℓ. Growing rice needs about 5 feet of water (acre-feet per acre) per crop, because a rice field is basically a swamp; with 2 feet of water per year, you can get pastureland or vineyards instead of a swamp. 5 feet of water (1.5 m) per year at 20 kJ/ℓ is just under 1 W/m². So turning the Sahara (9.2 million km²) into rice fields would cost 9.2 terawatts, which is about half of total world marketed energy consumption as of 02021. At current prices, the requisite solar panels (which would themselves occupy about ½% of the Sahara) would cost US$1.4 trillion.

- Direct air capture of CO₂ to reverse global warming, which requires minimally about a gigajoule (250 kWh) per tonne on entropic grounds, and maybe 10 GJ/tonne if we can't figure out how to approach the theoretical efficiency. We need to remove about 1.29 × 10¹⁶ kg to get back to pre-industrial levels (see Derctuo for the calculation) and if we do that over 25 years at this 10%-efficient 10 GJ/tonne level, it will take 160 TW, about 9 times current world marketed energy consumption.

— ⁂ —

Now, right now, most of these projects would be uneconomic, because energy is so expensive, so it's tempting to reject them out of hand as implausible. No doubt von Neumann would have done the same if you'd told him a prophecy of Hypercard, which shipped 30 years after his death in 01957. How much more so your ignorant USan man on the street, whose knowledge of computers in 01957 was limited to DO NOT FOLD, SPINDLE, OR MUTILATE, a UNIVAC trying to take over the world in The Invisible Boy, and breathless newspaper articles about "giant electronic brains"?

On the flip side, what would someone in the 01957 USA have thought if you told them that in 01987 all the US manufacturers of small planes would be bankrupt, leaded gasoline outlawed, and the interstate speed limits reduced to 55 mph?

The future is not only stranger than we imagine; it is stranger than we can imagine.

— ⁂ —

So, I ask you, a third time: Can you think of any other limiting factors I don't know about? What other possible limits do you think we might encounter between here and Kardashev Type 1?