Fifteen years ago, Jeremy Rifkin published The Empathic Civilization and promised, among other things, that smart grids would transform the global energy system. The intelligent networks were “ready”. Technology would resolve what politics had failed to resolve. We know how that ended.
Today the term “AI” enters the same slot and sounds new. The argument is identical.
The recycled narrative
The reasoning runs as follows: there is an enormous problem (climate change), there is a promising technology (AI), therefore the technology will solve the problem. Energy optimisation, demand forecasting, leak detection in electrical grids, high-resolution climate models. The list of applications grows every week.
The applications are not the difficulty. The conclusion drawn from them is.
Prediction is not control. Efficiency is not reduction. Unbounded digitalisation is not sustainability; it accelerates the same trajectory that produced the current situation.
An AI system that forecasts peak electricity demand with greater precision does not thereby lower total consumption. Where efficiency gains are redirected into greater output, the net effect on emissions may be neutral or adverse. Economists have a name for this: the rebound effect, or Jevons paradox. It is not a hypothesis. It is a pattern documented repeatedly whenever energy efficiency increases without constraints on volume.
The data centre problem
A further point rarely surfaces in discussions of “green AI”: AI systems themselves consume resources at growing scale. Data centres, cooling water, critical minerals for chips. Every query to a large language model carries a real energy cost, invisible to the user.
This does not make AI intrinsically unsustainable. It makes the equation more complex than “AI equals less carbon”. The outcome depends on what the system is used for, at what scale, with which energy sources, and under which regulatory counterweights.
COP30 and the technological temptation
This reasoning is already permeating forums such as COP30. The appeal is understandable: faced with the scale of the crisis, technology offers the illusion of an exit that requires no change to models of production or consumption. Optimise, forecast, digitalise.
Serious discussion of sustainability begins elsewhere, with the imposition of limits. Without them, the narrative of “AI to save the climate” functions as a diversion: it displaces attention from what matters, which is reducing, repairing and halting where halting is required.
The effects of ecological collapse do not wait for a technological narrative to mature. When they arrive, the question will not be whether the tools existed, but whether the corresponding decisions were taken.