In 1976 a solar panel cost about $100 for each watt it could make. A watt is the unit that power is measured in. In 2023 it cost about 15 cents. That is a fall of more than 99 in 100, after taking inflation out. The sun did not change. What changed is how many had ever been made.
After 1 doubling80
After 264
After 351
After 533
After 1011
The price, with the start set to 100, if each doubling takes a fifth off.
Here is the rule that fits the whole curve. Every time the total number of panels ever made doubles, the price falls by about a fifth. Make twice as many in all, pay 80 in 100 of the old price. Factories learn: workers get quicker, machines get bigger, waste gets found. People call this a learning curve.
Before the doubling100
After one doubling80
The price is 100. The total number of panels ever made doubles. Roughly what is the price now?
Still 100. Making more does not change a price.
Not yet. For panels it has, for fifty years, and the line in step 1 is the result. It does not work for everything, which is the second half of this lesson.
About 50. Twice as many, half the price.
Not yet. A doubling takes a fifth off, not half. Halving the price takes about three doublings. The curve is gentler than that at each step, and steeper than that over fifty years.
About 80.
Right. A fifth off. Halving the price takes about three doublings, not one. That is why the fall looks slow at first and then, decades later, looks like a cliff.
Move the control to set how many panels have ever been made, compared with the start. Each doubling of the total takes about a fifth off the price.
Price at the start100
Price now100
Each doubling of the total ever made takes about a fifth off.
Panels ever made, compared with the start: 1× (the start)
The start. The price is set to 100.
Price at the start100
Price now51.2
Each doubling of the total ever made takes about a fifth off.
Panels ever made, compared with the start: 8× more
3 doublings. The price is 51.2: about a fifth off, 3 times over.
Price at the start100
Price now26.2
Each doubling of the total ever made takes about a fifth off.
Panels ever made, compared with the start: 64× more
6 doublings. The price is 26.2: about a fifth off, 6 times over.
Price at the start100
Price now10.7
Each doubling of the total ever made takes about a fifth off.
Panels ever made, compared with the start: 1,000× more
10 doublings. The price is 10.7: about a fifth off, 10 times over.
Price at the start100
Price now1.2
Each doubling of the total ever made takes about a fifth off.
Panels ever made, compared with the start: 1,000,000× more
20 doublings. The price is 1.2: about a fifth off, 20 times over.
Which takes more money off the price: the first three doublings, or the next three?
The first three. 100 to 51 is 49 off; 51 to 26 is 25 off.
Right. The share is the same each time, about a fifth. But a fifth of a smaller number is a smaller number. The early doublings do the visible work; the later ones make it cheap enough to matter.
The next three. The factories are bigger by then.
Not yet. The factories are bigger, and the rule already counts that: it is why each doubling still takes a fifth off. But a fifth of 51 is less money than a fifth of 100.
The same. A fifth is a fifth.
Not yet. The same share, not the same money. Twenty off a hundred and twenty off fifty are different amounts, and the second is what happens by the sixth doubling.
Move the control.
095190
19002020
The price of a tonne of coal with 1900 set to 100, inflation removed, rounded. The shape is the claim, not the exact points.
Coal does not do this. Its price has bounced with wars and booms, but a tonne in 2020 cost about what a tonne cost in 1900, once inflation is taken out. Mining learns too. But each new mine digs deeper or further than the last, and the saving from learning is eaten up by the harder digging.
Which of these gets cheaper the more of it we make, and which does not, because we dig it or grow it?
One card at a time. Tap the pile it belongs to.
Card 1 of 6
Solar panels
Coal
Lithium batteries
Natural gas
Wind turbines
Crude oil
Solar panels → Made in a factory
Not yet. Solar panels are made in a factory. The same panel made a million times gets a little cheaper each time.
Coal → Dug from the ground
Not yet. Coal is dug from the ground. Each new mine has to find its coal again, and the easy layers went first.
Lithium batteries → Made in a factory
Not yet. Batteries are made in a factory. Every doubling of how many are made has cut the price, year after year.
Natural gas → Dug from the ground
Not yet. Gas is dug from the ground. Drilling learns a little, but each new well goes deeper or further than the last.
Wind turbines → Made in a factory
Not yet. Turbines are made in a factory. Blades and towers built again and again get cheaper with each batch.
Crude oil → Dug from the ground
Not yet. Oil is dug from the ground. A barrel today costs about what it did decades ago, because the finding never gets easier.
Made in a factorycheaper each doubling
vsvs
Dug from the groundno such rule
Right. Factory-made things follow the learning curve: the same thing made again and again, each time a little better. Dug things do not. Each mine has to find its coal again, and the easy layers were dug first.
New solar farm$60
New wind farm$50
New gas station$75
New coal station$118
Cost of a thousand units of electricity from a NEW plant, US, 2024, middle of the range, with no government help. One unit is what a small electric heater uses in an hour.
The result: in most of the world, a new solar or wind farm now makes electricity cheaper than a new coal or gas station, before any carbon price. That was not true in 2010. Nothing about coal got worse. Panels got cheap, and the sums flipped.
In 2010 a car battery holding one unit of energy cost about $1,400. What did it cost in 2023?
Pick one. Nothing is scored.
2010$1,400
2023$140
For one unit of battery, rounded.
About $140.A fall of nine tenths in thirteen years. It is the same rule as panels, about a fifth off per doubling of batteries ever made, and electric cars doubled the total fast. Forecasts made in 2010 priced electric cars with 2010 batteries, which is why they missed.
Today100
After one doubling80
After two0
The third bar is the question.
A panel costs 100 today. The total ever made doubles, then doubles again. What does it cost?
Take a fifth off, then take a fifth off what is left.
Right. 80 of 80 is 64. Two doublings of the world's panels have lately taken about six years, so 100 to 64 is roughly the pace the line in step 1 runs at now.
Nuclear power stations have been built for seventy years, yet in most rich countries a new one costs more than the last. Why does the learning curve not work here?
The rule only works for small things.
Not yet. Wind turbines are enormous and follow it closely. What matters is not the size of the thing but how many times the same thing is made.
Nuclear fuel is dug from the ground, like coal.
Not yet. The uranium, the metal fuel, is a small part of the cost. The plant is the cost, and a plant is built, not dug. The rule fails here for a different reason: almost nothing is repeated.
Few are built, each one differently, over a decade. There are almost no doublings, and little is repeated.
Right. The rule needs the same thing made many times. A country that builds one plant every fifteen years, to a new design each time, never gets a doubling. South Korea, which built the same design over and over, did get cheaper.
Lesson complete
Each doubling of panels ever made cuts their price by about a fifth. Coal has no such rule.