The true cost of prompts
With the rise of artificial intelligence (AI), the consumption of water and energy resources by data centers has significantly increased. According to an UN report, data centers consumed 4.5 trillion litres of water in 2025, equivalent to the needs of 600 million people in Sub-Saharan Africa. This trend is expected to double by 2030. In other words, this situation is not sustainable.
Immateriality Rhetoric
For several years, the AI industry has played on the rhetoric of immateriality, presenting AI as a miracle solution without considering the real and physical resources that support it. As philosopher Marcello Vitali-Rosati highlighted in his work “In Praise of Bugs”, this rhetoric allows the true value and meaning of things to be hidden, leaving a minority to exploit what remains invisible. The image of the “cloud” is a good example: it suggests that the data produced vanish like smoke, while they actually accumulate in tangible data centers. Entrepreneurs in AI have exploited this idea by promising miracles to users, inspiring both their admiration and their anxiety.
Example of “cloud” and hidden costs
For example, an AMD executive noted a regression in Claude, the AI assistant’s, responses due to exponential computing costs. This observation was debated but revealed a new kind of tension. Could this be the beginning of the end for the era of free and infinite AI promises?
Increasing costs and tokenwar
As water, electricity, and the minerals necessary for digital production consumption increases, the costs associated with AI become increasingly high. Businesses, who might have hoped for cost reductions through AI, may instead face higher costs due to excessive token expenses. An Uber executive acknowledged that their company had exhausted its token budget in just three months, highlighting the need to maximize the number of tokens to boost productivity.
The tokenwar and the AI adult
If the era of infinite promises of AI may be drawing to a close, we may be entering a new era - that of mature AI, characterized by limited capabilities. This phase will require individuals, businesses, and governments to identify the needs that justify these considerable costs. The war for tokens is far from over.
Finite Economics
The AI is prompting us to reconsider our relationship with the physical world. As economist Arnaud Orain has shown, the economy of scarcity involves searching for resources in a limited and constrained world. And we are indeed entering such a phase. Strategic choices surrounding AI, deeply rooted in material value chains and the real economies of the involved countries, are transforming our naive relationship with this technology.
Cet article est rédigé par Blaise Grenier, Journaliste reporter d'images et auteur. Découvrir le parcours de l'auteur et la charte éditoriale »