Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence: Summary & Key Insights

When I speak of the Earth in relation to AI, I begin with the literal material on which all computation depends: the minerals, metals, and resources extracted from the planet to make circuits, sensors, and servers. The myth of AI as cloud-based or weightless dissolves when you stand on a lithium mine in Chile or a rare-earth processing plant in Inner Mongolia. Every chip begins with excavation, every neural network depends upon the carbon costs of production and operation. To create vast computational infrastructures requires enormous energy ecosystems, and with them come pollution, labor exploitation, and geopolitical tensions.

I trace the environmental footprint of AI from its roots in resource extraction to its towering data centers that demand continuous energy flows. Google’s and Amazon’s centers consume the power of small cities while promising efficiency and intelligence. What often goes unseen are the degraded landscapes left behind and the communities living amid residues of industrial consumption. The physical scale of AI infrastructures mirrors the vastness of the data they absorb, revealing a paradoxical truth: an industry seeking to predict and optimize the future is destroying its very conditions of possibility.

In seeing AI as an ecological system, we confront a crucial question: can intelligence built on destruction truly be intelligent? Recognizing this interconnection doesn’t merely evoke environmental concern—it requires rethinking ethics itself. Intelligence cannot be separated from responsibility; computing cannot be detached from ecology. AI’s planetary reach compels us to understand it as part of a continuum of extractive capitalism stretching from mining to modeling.

Behind the sleek automation that defines AI lies the labor of millions of people whose work remains hidden. In Atlas of AI, I ask readers to think about all the hands that shape and maintain machine learning systems—the clickworkers labeling images, the moderators cleaning toxic content, the engineers scripting automation routines. AI depends upon vast hierarchies of labor, particularly invisible, precarious, and underpaid forms of human work.

One striking example comes from data annotation portals in Venezuela and the Philippines, where workers tag objects for self-driving cars and facial recognition datasets. Their fine-grained attention to detail trains algorithms to perceive the world with seeming intelligence, even though that intelligence is actually borrowed human labor. The paradox of automation is that the more we celebrate systems as self-sufficient, the more invisible the human work beneath them becomes.

In telling these stories, I show that AI’s labor economy is globalized and stratified. The promise of flexibility and scale often conceals exploitation. Even within high-tech firms, the rhetoric of innovation can hide the burnout and moral distress experienced by those managing biased or abusive content. AI is not a replacement for human intelligence; it is a machine that consumes it.

By making labor visible, we gain insight into the politics of value creation in technology. The dream of automation is never purely technical; it is historical, tied to efforts to abstract and control human activity. Understanding these labor conditions allows us to ask critical questions about justice, dignity, and the future of work in the age of so-called intelligent machines.