The paradox of faster learning and slower cost reduction as technologies mature
This paper analyzes 145 clean technology series to reveal a paradox where, as technologies mature, learning efficiency increases while annual cost reduction slows due to decelerating production growth, necessitating policies that both catalyze early-stage innovations and sustain the deployment of mature ones.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you are watching a new video game console hit the shelves. At first, it's expensive, rare, and a bit buggy. But as more people buy it, factories get better at making it, engineers figure out how to cut corners without breaking it, and suddenly, the price drops like a stone. This isn't magic; it's a well-known pattern in science called "technological learning." Think of it like a student getting better at a subject: the more they practice (or the more units are produced), the cheaper and easier it becomes to make. Scientists have long used two main rules to predict this. One rule, often called Wright's Law, says costs drop every time production doubles. The other, Moore's Law, says costs drop every year that passes. For decades, most computer models used to plan our energy future assumed these rules were like a steady metronome: click, click, click, always at the same speed. But what if the rhythm changes? What if the music speeds up in some parts and slows down in others? Understanding this is crucial because if we guess wrong about how fast clean energy (like solar panels or electric cars) will get cheaper, we might either panic unnecessarily or, worse, fail to act fast enough to save the planet.
Now, let's dive into the new study by Jessica Jewell, Huaxuan Wang, and their team, which turns the old metronome idea on its head. They gathered a massive library of data—145 different stories of technologies growing from tiny experiments to global giants, covering everything from solar panels to nuclear power. They organized these stories into three chapters: the messy "formative" start, the fast "accelerating growth" middle, and the "steady growth" mature phase.
Here is the twist they discovered, and it's a bit of a paradox. As technologies get older and more mature, they actually get better at learning. Imagine a factory worker who starts out clumsy but eventually becomes a master craftsman. The study found that "learning rates" (how much cheaper things get every time production doubles) jump up significantly. For brand-new tech, the learning rate is about 15%. But for mature tech, it rockets to over 30%. The factories are learning faster than ever!
However, here is the catch: even though the factories are learning faster, the actual price of the technology stops dropping as quickly. In the early days, costs might fall by 10% every year. But once the technology matures, that annual drop slows down to just 3%. How can you get better at learning but get worse at cutting costs? The authors explain this using a simple metaphor: running a race. In the beginning, you are running on a short track, so every step you take (every doubling of production) happens quickly, and your speed (cost reduction) is high. But as the race goes on, the track gets longer and longer. Even if you are a faster runner (higher learning rate), it takes you much longer to complete the next lap (the next doubling of production) because the track is so huge. The "experience" is piling up, but it's piling up more slowly because the market is getting saturated.
The team didn't just guess this; they tested it with computer simulations. They created 10,000 fake technologies to see what would happen under different rules. They found that the only way to recreate the real-world pattern—where learning speeds up but annual cost drops slow down—is if two things are happening at once: the technology is getting better at learning-by-doing, and it is also improving just because time is passing (autonomous progress). If you only had one or the other, the numbers wouldn't match reality.
So, what does this mean for us? It suggests that the old way of modeling the future—assuming costs will drop at a steady, predictable rate forever—is likely wrong. The "self-reinforcing" cycle where cheap tech leads to more sales, which leads to even cheaper tech, starts to weaken as technologies mature. The authors suggest that we can't just sit back and wait for prices to crash. Instead, we need smart policies. For new technologies, we need to push them hard to get them into the "accelerating" phase. But for mature technologies, we can't rely on the old magic of rapid growth; we need specific policies to keep the deployment going and overcome the new barriers that slow things down. The learning is still happening, but the race has changed, and we need to run it differently.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.