While AI’s promises are often wrapped in narratives of innovation and transformation, these technologies are grounded in deeply material realities: server farms, energy-hungry computation, rare earth minerals, and the human labour behind system maintenance (Brevini, 2021; Crawford, 2022; Saenko, 2023). Large language models (LLMs) quantify their outputs in “tokens”—a benign-sounding metric that obscures the real-world costs behind each interaction. As we design learning experiences powered by AI, we must also ask: what are we truly spending, and on whose behalf?
Drawing on metaphors from data extraction to climate catastrophe (Weller, 2022; Holmes & Tuomi, 2022), this gasta considers alternative ways of visualising the cost of AI use. Imagine if, alongside each AI-generated response, learners were confronted with the image of a rare earth mine labourer, a coal energy plant, or a melting glacier? Could such visual interventions foster a sense of ethical intentionality and digital restraint? This gasta proposes a rematerialisation of the environmental toll of AIED within our digital learning spaces—making energy consumption visible and accountable. Rather than perpetuating dominant, extractive AI narratives focused on scale and speed, this talk advocates a pedagogy of digital degrowth (Selwyn, 2023), where critical use, slowness, and sufficiency become design values.