This is a segment from the Blockworks Daily newsletter. To read full editions, subscribe.
When people hoped that more efficient steam engines would help conserve England’s limited reserves of coal, William Jevons warned of just the opposite: Greater efficiency, he predicted, would lead to even greater consumption.
He was right, of course, and his prediction became known as the Jevons paradox, which is often invoked whenever efficiency gains inspire people to use more of something.
When cars got more fuel-efficient, people drove more; when lightbulbs got more efficient, people installed more of them; when HVAC got more efficient, people built bigger houses.
But these are better described as examples of a “rebound effect,” where some, but not all, of the expected savings from efficiency are offset by increased usage.
It wasn’t until the digital revolution that the Jevons paradox truly came into its own: As fast as Moore’s law has decreased the cost of computing, our use of computers has increased even faster.
As a result, things like semiconductors and cloud computing — where costs are in perpetual decline — have been both great businesses and great investments.
The current craze for investing in AI is similarly predicated on more efficient GPUs enabling a disproportionately large expansion of usage.
Microsoft’s Satya Nadella is confident that it will: “Jevons paradox strikes again,” he posted on X in response to investors briefly worrying that the advent of DeepSeek would reduce demand for both GPUs and cloud services.
So far, he seems to be right — the efficiency gains from DeepSeek and other more efficient LLMs seem to have been more than offset by increased demand.
But any investor in either airline or telecommunication stocks can tell you that it doesn’t always work out that way.
In both cases, efficiency rose, costs fell and usage increased disproportionately — but investors still lost.
Airline traffic, for example, has increased exponentially since people started flying commercially in the 1950s (and with only the smallest of cyclical downturns):
That looks like a highly investable chart and the Jevons paradox-related numbers are even better: Since 1970, the real cost of air travel has fallen by 1.7% per year, on average, while the underlying rate of air traffic has grown 4.4% per year.
Anyone who had those numbers in advance would likely have started an airline or at least invested in airline stocks — but they’d have come to regret it: Warren Buffett calls airlines the “worst sort of business” and airline stocks a “deathtrap for investors.”
Similarly, lots of people who correctly predicted the explosion of internet traffic in the 1990s — or underpredicted it even — lost money buying stock in the companies that hosted that traffic.
Internet traffic grew 127% per year between 1997 and 2003 — at the end of which period investors in the likes of Level 3 Communications, Global Crossing and WorldCom were more or less wiped out.
Now consider crypto.
Much of crypto investing is predicated on the assumption that more efficient, lower-cost blockspace will lead to an exponential increase in demand for that blockspace.
The results have been mixed so far.
Cheaper blockspace has led to an explosion of activity on blockchains like Solana and Ethereum layer-2s like Base.
But that’s mostly been thanks to memecoin trading, which is presumably not a decades-long growth story — it’s hard to imagine that even cheaper blockspace will lead to even more memecoin trading.
(Let’s hope not, at least.)
For what the elasticity of blockspace demand looks like without memecoins, we might only have to look at Ethereum.
Early hopes that moving execution off-chain would lead to Ethereum collecting more fees in settlement and data availability than it lost in execution have been disappointed.
On Ethereum itself, the cost of transacting has fallen over 90% from its 2021 peak — but, per data from Blockworks Research, the number of transactions is close to unchanged.
Worse still, the combined fees paid on all Ethereum-aligned layer-2 blockchains does not offset the loss of fees on layer-1 Ethereum — so the ecosystem as a whole does not appear to be an instance of the Jevons paradox.
