The Paradox at the Heart of Circular AI
Artificial intelligence is often portrayed as the technology that will finally make the circular economy real, the system that can track materials, redesign products, sort waste flawlessly, and optimise industrial processes so value stays in circulation. In this vision, AI is the invisible coordinator that turns a fragmented material world into a regenerative one.
But The AI Linearity Paradox reveals a quieter truth beneath the optimism: AI’s own lifecycle is still profoundly linear. The very infrastructure that powers circular solutions i.e. data centres, GPUs, water‑cooled server farms, and rapidly obsolete hardware...is expanding at a pace that undermines the circular gains AI helps create.
Electricity demand from data centres is rising sharply, projected to reach 945 TWh by 2030. Generative‑AI hardware turnover could generate up to 5 million tonnes of e‑waste this decade. Water withdrawals strain local basins, and efficiency improvements are swallowed by rebound effects that erase most of the savings. AI, in other words, advances circularity downstream while reinforcing linearity upstream.
To resolve this contradiction, the paper proposes the Net‑Circularity Equation, a rule that requires any claimed circular benefit of AI to be weighed against the full lifecycle burden of the infrastructure that produced it. When the authors attempt to run this equation for a real-world case, they find that the necessary data simply isn’t available. The gap in evidence becomes part of the paradox. The narrative ends with a clear message: AI can support a circular economy, but only if the circular economy applies its own rules to AI. Without lifecycle accountability, transparency, and governance of hardware, water, and energy, circularity becomes performance rather than practice.
Access the full report here:
