Wright's Law

The law

Theodore Wright observed in 1936 that the cost of building an aircraft fell by a predictable fraction each time cumulative output doubled. The fraction is empirical and specific to the technology. For solar panels it has run about 20 percent per doubling for nearly five decades. For lithium-ion batteries it is about 23 percent, which is how pack prices fell more than 90 percent after 2010.¹ The mechanism is unglamorous: factories that make more of a thing get better at making it.

The forecasting corollary is the sharp part

The interesting claim is not about cost curves but about who gets them wrong, and for how long. Institutions that modeled solar the way they model traditional energy, as a resource with a rising marginal cost of extraction, rather than as a manufactured product closer to a flat-screen television than a power plant, were wrong in the same direction for twenty consecutive years. Every year the International Energy Agency projected solar growth leveling off, and every year the exponential continued.¹

Vaclav Smil's argument that energy transitions are inherently slow fails on the same distinction. Coal took 35 years to go from 5 percent to 25 percent of the market, oil 40, natural gas 55. But coal, oil, and gas never got 20 percent cheaper every couple of years. The historical base rate is a fact about extracted commodities and says nothing about manufactured ones.

An error sustained for two decades by competent institutions is worth more attention than the curve itself. It is the clearest case in this archive of a category mistake surviving repeated falsification because the category, not the data, was doing the predicting.

Riding it on purpose

A learning curve is normally something that happens to an industry. It can also be built into a company deliberately. Zach Dell's Base Power is designed as a vehicle for exactly this: vertical integration and technology lower cost, lower cost permits lower prices, lower prices bring demand, demand brings scale, and scale lowers cost again.¹ The visible output of a flywheel like that is not a margin but a product. Base's first battery was 25 kilowatt-hours; three years later the 39.2 kilowatt-hour Core shipped at the same price.

Pairing solar with storage stacks two learning curves, since the battery repairs the one defect in the cheaper of the two, which is that the sun sets.

Where it breaks

Wright's Law is an observed regularity, not a guarantee. Doubling cumulative production gets harder in absolute terms every time, and input constraints in lithium, land, or tariffs can dominate the learning effect. There is also a geopolitical complication the curve does not care about: China holds more than 80 percent of battery cell production and a commanding share of every stage of solar manufacturing, so the learning is accruing to a strategic competitor, and the tariff response raises prices in the opposite direction to the curve.¹

The reader is directed to the file on the interconnection queue, which is where the cheap output of this law goes to wait.