The Jevons paradox is a strange little thing, partially because it isn’t actually a paradox but more because of how it keeps finding ways to be relevant.
William Stanley Jevons, who is the father of the mathematical approach to economics, observed in 1865 that the increased efficiency with which coal could be used actually led to an increase in its usage. That may seem obvious today but the reasonably popular view at the time was that the technological developments that improved efficiency would in turn reduce usage.
There have been plenty of other examples of it in the 160 years since. Some of you may remember the paperless office being discussed for years, but right now artificial intelligence is creating a scenario where it is worryingly relevant.
It all comes down to code. Every line of code before the AI revolution was written by someone. A human had to do the work to input it. That put some natural barriers on how much code could be produced and also put a premium on the service.
With AI, code has become a lot easier to produce. The human limit on it has been reduced substantially as software developers can now use it to vastly increase their output.
Now consider the Jevons paradox. It’s not just that the existing software will be produced and updated faster.
[ Replicant rock: How AI-generated pop stars are killing musicOpens in new window ]
Thousands, more likely tens or hundreds of thousands, of projects that didn’t previously make sense from a resource perspective are now viable. Everything is going to become a software project.
The growth of agentic AI is already showing this as small businesses can now create tools that previously required a development budget. They may need some human skills but these can be brought in on a contractor basis rather than requiring the investment in an entire team.
The world will still need people who can write code, it’s just that access to them won’t be as scarce. That will cause some market disruption for developers but there’s a very good chance they end up as well or better off from it. Demand for coal miners increased substantially following the original observation by Jevons.
Demand for code will soar as it becomes more efficient to produce. This will also lead to far more projects that turn out to have no real value being executed. Budget constraints tended to be a rather good way of vetting a project adequately before giving it the green light. There will be follies and no shortage of them.
The issue is that this is going to require an awful lot of energy. The impact of data centres on the grid is already a hot topic and the increased ease of access to code is only going to make that increase.
Even as the energy usage of data centres becomes more efficient – market forces indicate this is likely, as it’s in everyone’s interests – that doesn’t mean the energy usage will drop. Increased convenience will only make demand rise, meaning that more total energy usage is effectively certain.
The International Energy Agency found that AI-related energy use in data centres jumped by 50 per cent in 2025 and expects that to jump by another 75 per cent this year. Consider that data centres accounted for 22 per cent of electricity demand in 2024, with their share forecast to rise to 31 per cent by 2034.
Even with pressure mounting on those running data centres to be more energy efficient and those tech giants largely working to respond in kind, the ease of access to code presents a problem.
Every data centre business in Ireland, indeed globally, could do its absolute utmost to improve energy efficiency and usage will still most likely go up. It’s not quite Catch-22: a more radical shift to renewable energy would solve the bulk of the environmental issues that come when the Jevons paradox is applied to code.
[ Failed shoe company’s pivot to AI is for the birdsOpens in new window ]
Putting that context on the matter would surely help the push towards greater development in the renewables sector, as it could be presented as enabling industry rather than just ensuring the planet is liveable. That may be glib but the carrot is usually a faster-acting motivator than the stick.
The downsides of this increased access to code can also be far more direct to businesses. It may be easier to create a wide range of new features to use but each of these must be tested and integrated into business operations.
The more companies lean on automation, the less they will understand about what actually makes their systems work. One scarcity gets replaced by another.
The supply challenge will be in people that ensure a tool is suitable for your business and, most importantly, is and remains safe to use.
In 1865 it’s unlikely that even a mind as great as Jevons could have conceived the AI we have today. He did however realise that the increased efficiency of coal use would be replaced with other scarcities.
In the AI era, it may be that knowing what code not to use is the most valuable asset of all.
