Whilst he may have died almost a century and a half ago, 2025 has seen a sudden revival of interest in William Stanley Jevons.
Jevons, born way back in 1835, was one of the great economists of the nineteenth century. And while he may not have quite the enduring fame of some of his contemporaries, he is still widely recognized as a pioneer in the use of mathematics in economics. The reason for the sudden revival of interest in his work though lies in what is known as Jevons Paradox.
In 1865 Jevons published a book commonly now called The Coal Question – although actually entitled, in the somewhat over-elaborate style of the Victorian publishing industry – The Coal Question; An Inquiry Concerning the Progress, and the Probable Exhaustion of Our Coal Mines.
Jevons observed what seemed to be a contradiction in the market for coal. Inventions such as Watt’s steam engine in the decades before he penned his work had made coal an increasingly efficient source of power. But even as more power was extracted from each unit of coal, the demand for coal seemed to increase.
Whereas one might have assumed that if, say, twice as much energy could be extracted from each lump of coal then demand for coal might half, this was incorrect. As Jevons wrote:
“It is a confusion of ideas to suppose that the economical use of fuel is equivalent to diminished consumption. The very contrary is the truth.”
What was actually happening was that the increased efficiency of coal was making it an ever more attractive source of fuel and so, it was being adopted for more and more purposes.
The Jevons paradox – sometimes called the Jevons effect – can come into play following a technological change if certain conditions hold. If a new technology or process increases the productivity of something and the price of that good falls as a result and demand for that good is highly price elastic (that is to say, demand for that good is highly responsive to changes in prices) then the paradox will hold. Increasing efficiency in this case will lead to lower unit prices but rising demand may mean even more units are demanded.
The sudden resurgence of interest in Jevons and his paradox has come about thanks to Microsoft CEO Satya Nadella and his response to the Chinese AI model DeepSeek.
Not only was the model trained on the cheap, running it costs less as well. DeepSeek splits tasks over multiple chips more efficiently than its peers and begins the next step of a process before the previous one is finished. This allows it to keep chips working at full capacity with little redundancy. As a result, in February, when DeepSeek starts to let other firms create services that make use of v3, it will charge less than a tenth of what Anthropic does for use of Claude, its LLM. “If the models are indeed of equivalent quality this is a dramatic new twist in the ongoing LLM pricing wars,” says Simon Willison, an AI expert.
When equity markets came to terms with what Deepseek meant, the initial reaction was a sharp sell-off in the shares of AI producers and, in particular, of Nvidia – the maker of the high-end chips that underpin much of the industry.
It was at this point that Mr. Nadella brought up the Jevons paradox. Posting on X that:
“Jevons paradox strikes again! As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can’t get enough of”.
One can see why Mr. Nadella and other Big Tech bosses might find the idea of the Jevons paradox appealing. As On Global Markets has previously noted, US equity returns have become increasingly narrow in recent years with increasing concentration amongst the biggest tech firms. Much of that investor enthusiasm results from a belief that AI is the wave of the future and that the largest tech firms will reap large profits as its use spreads.
As the violent initial market reaction to the existence of Deepseek demonstrated, anything that threatens that picture has the potential to pull valuations lower.
In that regard then, the Jevons paradox might seem appealing. If it were to hold then even as more efficient LLMs appeared then the falling price of AI services might mean even more AI is utilised by firms and households and even more Nvidia chips are demanded.
But is this comforting story (or comforting for Big Tech firms anyway) true?
That may be more of a stretch. The hope for Big Tech is that the Jevons paradox means that whilst unit prices will fall, they will make up for that in revenue terms by selling even more units. But not all goods are like coal in the mid-nineteenth century. The paradox does not always apply. What really matters is the demand response and the notion that demand for AI will soar to the extent required is, at best, questionable.
But there should be a deeper concern for tech titans too. The valuations of their equities are not only based on already quite punchy assumptions about the future uses and values of AI but also on these firms being able to reap some more monopoly-like profits from this growing market.
DeepSeek not only raises questions about pricing but also suggests the AI market of the future will be much more competitive.
