Energy emergence at the frontier
Jevons Paradox, the Khazzoom-Brookes Postulate, and DeepSeek
Welcome to Carbon Risk — helping investors navigate 'The Currency of Decarbonisation'! 🏭
If you haven’t already subscribed please click on the link below, or try a 7-day free trial giving you full access. By subscribing you’ll join more than 4,000 people who already read Carbon Risk. Check out what other subscribers are saying.
You can also follow on LinkedIn and Bluesky. The Carbon Risk referral program means you get rewarded for sharing the articles. Once you’ve read this article be sure to check out the table of contents [Start here].
Thanks for reading Carbon Risk and sharing my work! 🔥
Estimated reading time ~ 10 mins
The ‘rebound effect’ is experiencing something of a renaissance.
Last week Microsoft CEO Satya Nadella took to X to argue that the emergence of Chinese artificial intelligence model DeepSeek - thought to be 95% more energy efficient than other AI models - would help AI use to “skyrocket, turning it into a commodity we just can’t get enough of.”
Beginning his post, “Jevons paradox strikes again!,” Nadella makes reference to an obscure economic theory from 19th Century Britain.1
In 1865 William Stanley Jevons, an English economist and mathematician, warned that Britain’s economic prosperity was at risk because it would soon run out of coal. The late 19th Century was a period of profound change where electricity, railroads, and other technological advancements dramatically altered the economy. Coking coal was in high demand to manufacture the enormous amounts of steel required by industry.
Jevons suggested that technological advances would make it possible for a blast furnace to produce iron with less coal. The resulting rise in profits would attract new investment in iron production, leading to a fall in the price of iron that would stimulate additional demand for coal. As an example Jevons observed that Britain’s consumption of coal soared after James Watt first introduced the Watt steam engine in 1712.
As Jevons concludes in his essay The Coal Question, “It is wholly a confusion of ideas to suppose that the economical use of fuel is equivalent to a diminished consumption. The very contrary is the truth.” The essay a rebuttal to critics who suggested that as steam engines and steel production became more efficient, demand for coal would fall.
The Jevons Paradox argues that rather than leading to lower energy consumption, an improvement in energy efficiency (perhaps due to a new technology or way of doing things) will lead to an increase in energy consumption. The paradox is also known as the ‘rebound effect’ when it merely erodes some of the savings, or the ‘backfire effect’ when the increase in energy consumption overwhelms the initial saving (see Energy efficiency is undervalued and misunderstood).
The micro reverberations
Direct effect (consumer): An energy efficiency gain reduces the effective price of energy, causing consumers to use more of it. For example, the installation of more efficient household heating causes the household to heat more rooms or turn the temperature up.
Direct effect (producer): Energy efficiency gains reduce the effective price of energy, meaning producers are potentially able to increase their output. It may also allow firms to reconfigure the inputs of new capacity to use more energy profitably. The greater the flexibility to reconfigure new productive capacity the greater the direct rebound effect.
Indirect effect (consumer): Households that saved money off their heating bills after installing a more efficient boiler, then spending it to purchase other goods and services that require energy to provide.
Indirect effect (producer): Energy efficiency gains reduce the output price of products causing consumers to shift toward them (and away from more energy intensive products), but the shift in demand places higher energy demands on the more efficient producer.
The macro implications
While the Jevons Paradox was born out of Britain’s Industrial Revolution, the Kazzoom-Brookes Postulate emerged in response to developments following the 1970’s oil crisis.
In the 1980s, economists Daniel Khazzoom and Leonard Brookes independently put forward a theory that increased energy efficiency paradoxically tends to result in increased energy consumption. It was only later, in 1992, that US economist Harry Saunders dubbed this hypothesis the Khazzoom-Brookes Postulate.
If the Jevons Paradox can be thought of as what happens to energy consumption on a micro level, the Khazzoom-Brookes Postulate takes it one big step further and shows what could happen to energy consumption at a macro level. In short, the postulate states that "energy efficiency improvements that, on the broadest considerations, are economically justified at the microlevel, lead to higher levels of energy consumption at the macrolevel."
The dramatic spike in the price of oil in 1973 and then again in 1979 spawned huge improvements in energy efficiency, in particular more fuel efficient vehicles. The knock-on impact of the improvement in energy efficiency was an increase in demand for road transport (in terms of miles travelled), and higher oil consumption.
The Khazzoom-Brookes Postulate goes further and argues that energy efficiency is what enables economic growth to take place. Energy efficiency leads to lower energy prices and enhances energy availability. This then leads to new products and services being developed, which in turn results in greater demand for energy.
Going deep
One of the largest assessments of peer-reviewed journal literature was published in 2011 by the Breakthrough Institute in which they reviewed almost 100 published journal articles assessing evidence of a rebound effect. The report finds that direct rebound effects “can erode 10-30% of projected technical energy savings”, even before considering the potential indirect or macroeconomic impacts.2
Direct rebound effects also appear to be more significant in energy intensive sectors of the economy (20-60%), where energy services are more easily substituted for other factors of production. The degree to which demand for a particular energy service has been fulfilled is also important, according to the Breakthrough Institute. The evidence suggests that larger rebounds (~40-80%) are more likely when there is pent-up demand for energy, e.g., demand for electricity and steel in less developed economies.
In contrast to the direct rebound effects, very little research has been undertaken trying to quantify the magnitude of the indirect effects - either in terms of spending by consumers, or investment by firms. The struggle to quantify it stems in part from the inability to observe it in action, and the potential for long-lead times. Overall the message from the Breakthrough Institute report is that these indirect effects are likely to be very small.
Complex analysis involving computable general equilibrium (CGE) and integrative modelling have been used to explore the evidence of a rebound at the macroeconomic scale. Whatever the model used, the majority of the research published tends to indicate a macroeconomic rebound effect in the 30-50% range, both on the national and global scale.
The frontier effect
But could the estimated level of rebound effect be too conservative? The Breakthrough Institute believes that modelling is “likely to have underestimated the potential for economy-wide rebound, given that they are currently restricted by exogenous assumptions about the scale of direct rebound”, and are limited to modelling “energy productivity improvements without considering the potential for multi-factor productivity improvements from energy-saving technologies.”
There’s reason to heed their warnings, especially when it comes to technologies for which efficiency improvements may unlock unforeseen new energy-consuming applications and products, or even spawn whole new industries - the ‘frontier effect’ dynamic. Sorrel, one of the academics cited in the report, notes that this is most likely to arise for ‘general-purpose technologies’ that have “a wide scope for improvement and elaboration, have potential for use in a wide variety of products and processes, and have strong complementarities with existing or potential new technologies.”
Historical examples of this ‘frontier effect’ dynamic include the widespread new applications of Watt’s improved steam engine, which was at the heart of Jevons’ original treatise. The introduction of more fuel efficient internal combustion engines (ICE) in the aftermath of the 1970’s oil crisis is another. Other examples put forward by the report (remember that it was published in 2011), include “gas turbines, semiconductors and computing technologies, lasers, robotics, radio transmitters.” Clearly these emergent ‘frontier effect’ dynamics are particularly difficult to model and to predict the effect they could have on energy demand.
Seeking the next frontier
The world has recently undergone another energy crisis, this time driven by higher natural gas prices following the Russian invasion of Ukraine. We’re only beginning to see the impact that this latest energy crisis will have on energy efficiency. The frontier technology today is arguably semiconductor chips and AI.
The former fits Sorrell’s classification of ‘general-purpose technology’, and one that has “a wide scope for improvement and elaboration”, while the latter has “potential for use in a wide variety of products and processes, and have strong complementarities with existing or potential new technologies.”
The extent to which we can observe this particular frontier technology is really up to our imagination, and that’s what makes it so difficult to understand the potential impact on energy demand. Microsoft’s Nadella is right that as AI comes down in price and becomes more commoditised, the rate at which it reaches broad adoption will accelerate, until it hits the steepest part of the S-curve and we can’t get enough of it.
For investors, the kneejerk reaction to DeepSeek was to question the aggressive power demand forecasts. If more energy efficient, open source AI models could be developed then the power crunching demands of data centres would most probably go down. That’s the first order knock-on effect. However, history tells us that the direct rebound effect means consumers will use more, and businesses will up their investment - especially so with frontier technologies.
Lets consider the indirect rebound effects. The evidence indicates these are negligible and difficult to quantify. For AI and data centres the second-order effect probably means more of what is now business-as-usual will be outsourced to the cloud. Although that means we are likely to consume more, the migration towards hyperscale, hyper efficient datacentres is likely to counter the increase in energy demand.
The really big knock-on impact is likely to be macro. This third order knock-on effect will pivot on the degree to which chips and AI result in a reconfiguration of the global economy, the birth of unimaginable new business models, and potentially, the dawn of a new period of burgeoning demand for energy.
Nevertheless, history tells us that we should err on the side of caution. We’ve been here before. In 1999, an article was published in Forbes Magazine called 'Dig more coal -- the PCs are coming'. The opening paragraph begins, "Somewhere in America, a lump of coal is burned every time a book is ordered on-line." At the time internet traffic was doubling every three months.3
The authors of the article make the astounding prediction: “It’s now reasonable to project that half of the electric grid will be powering the digital-Internet economy within the next decade.” But a decade later data centres powering the internet consumed only a couple percent of America's electricity, nowhere near the 50% prediction made in the article. And so, whether it’s steam, gasoline, or electricity, the history of frontier technologies suggests that it may take longer than we expect for them to have a transformational impact on energy demand.
Unlike previous periods in history, this latest frontier need not result in higher greenhouse gas emissions. The opposite may be true. The boom in data centre capacity amplifies both the need and the opportunity for sustainable power sources.
Rather than simply being seen as hogging power, data centres are a valuable customer for renewable energy developers. Huge concentrated energy demands, coupled with the willingness and ability of data centres to pay top dollar is likely to continue to drive investment in renewable energy.
Finally, for Daniel Khazzoom and Leonard Brookes, long cast under the shadow of William Stanley Jevons, the idea that energy efficiency ultimately drives progress is what it’s always been about. For that the latest frontier technology should be something to celebrate, not feared.
👋 If you have your own newsletter on Substack and enjoy my writing, please consider recommending Carbon Risk to help grow this amazing community of readers! Thank You!👍
https://x.com/satyanadella/status/1883753899255046301
https://thebreakthrough.org/articles/new-report-how-efficiency-can
https://www.forbes.com/forbes/1999/0531/6311070a.html?sh=768327272580