Jevons' paradox

I had not heard of Jevons’ Paradox before the DeepSeek AI scare in late January 2025. So why did I hear then? Well, Jevons' paradox is that energy efficiency improvements can lead to increased energy consumption (Sorrell, 2009): i.e. efficiencies lead to more resources being available, so we use more (York & McGee, 2016). These two ideas - Al and Jevons' paradox - come together for me because suddenly it was possible that the more AI power we have, maybe the more AI power we might want to use (Boyle, 2025). And with DeepSeek, AI was free.

Where is the economic driver in that model, I hear you ask? Well, that is a good question. We will come back to that. 

Let's begin with Jevons' paradox. The brain-child of William Stanley Jevons, an economist way back in 1865, the idea is that efficiencies may inadvertently lead to profligacy - running counter to many economic sustainability theories (Sorrell, 2009). Jevons' is a subset of the rebound effect (York & McGee, 2016) where energy efficiency improvements in early-stage, energy-intensive technologies are more likely to backfire (Sorrell, 2009). Jevons looked at steam-engine coal consumption, showing that efficiency improvements lead to increased energy use (Sorrell, 2009), with similar patterns in steel production where energy efficiency gains spurred greater demand; and lighting efficiencies leading to significant energy consumption increases. We do more with the energy we have, so we don't reduce consumption. Think cars: fuel efficiency meant we could now have a heater, air conditioner, heated seats, electric windows; and China's energy efficiency improvements since the 1970s with per capita energy use increasing at the same or greater rate (York & McGee, 2016). Fuel-efficient cars may entrench a car-centric economy, over public transport (York & McGee, 2016). Capitalism rewards profit: efficiency may reduce build costs, driving production and consumption (York & McGee, 2016). 

A more refined take on Jevons’ paradox, called the Khazzoom-Brookes (K-B) postulate (Sorrell, 2009) is where high-quality energy inputs are a primary driver of economic growth, focusing on energy quality and efficiency (Sorrell, 2009; York & McGee, 2016). More conventional economic theory downplays energy in favour of capital, labour, and technological change (Sorrell, 2009). So, if energy plays a more significant role in productivity improvements than mainstream economics acknowledges - i.e. the K-B postulate/Jevons' paradox - energy efficiency improvements might (a) increase energy consumption, (b) improve productivity, and (c) be a better driver of economic growth (Sorrell, 2009). But. There isn't a consensus on efficiency as a driver of environmental reform or degradation: and it seems likely that efficiency may drive economic growth AND resource consumption, over conservation (York & McGee, 2016).

The more we have, the more we use. Is it bad to want more of a good thing? Perhaps the rebound idea really comes down to an opportunity cost judgement call. We humans are a switching predator, after all, so efficiencies seem to lead to over-consumption, not to greener practices. 

Using historical correlations between energy efficiency and economic output, Brookes (of the K-B postulate; as cited by Sorrell, 2009) highlighted the importance of energy quality and the potential for energy efficiency improvements to stimulate demand for energy services. He thought that energy efficiencies could lead to higher energy use ...via the mechanisms productivity and economic growth. Further, Saunders (as cited by Sorrell, 2009) thought that energy efficiencies might backfire, leading to energy use rises, not falls; challenging conventional energy-economic models, and potentially underestimating rebound effects. Basically, we don't know which way any system will go: perhaps it is a continuum with Brookes on one end and Saunders on the other.

Jevons’ paradox brings up questions about how energy efficiency, productivity, and economic growth relate to each other; and question whether rebound effects will cause unforeseen problems in general-purpose technologies with broad economic impacts (Sorrell, 2009). We need to be awake to that backfiring potential (Sorrell, 2009), and to realise that environmental practices require other drivers, as the market is likely to consume all it can (York & McGee, 2016).

That may mean that, if we have more AI, we will use more AI. If AI is free due to China's AI models actually being open source (Boyle, 2025) - providing DeepSeek et al are of appropriate quality - we may consume all that we can eat. And that seems likely to make the economics of the USA 'Tech Bros' AI models precarious (Boyle, 2025). 

Watch this space.

Sam

References:

Boyle, P. (2025, February 2). DeepSeek - How a Chinese AI Startup Shook Silicon Valley [video]. YouTube. https://youtu.be/xUzzWUlSk98

Sorrell, S. (2009). Chapter 7: Exploring Jevons' Paradox. In H. Herring, S. Sorrell (Eds.), Energy Efficiency and Sustainable Consumption: The Rebound Effect (pp. 136–164). Palgrave macmillan.

York, R., & McGee, J. A. (2016). Understanding the Jevons paradox. Environmental Sociology, 2(1), 77–87. https://doi.org/10.1080/23251042.2015.1106060