AI’s unsustainable growth: Extractive data regimes, energy colonialism, and the erasure of planetary limits
Original framing: “Sustainability has left the AI chat” — The Japan Times
Peer-reviewed studies (e.g., Strubell et al., 2019; Luccioni et al., 2022) quantify AI’s carbon footprint, with training a single large language model emitting ~300 tons of CO2—equivalent to 125 round-trip flights from NYC to SF. The 'efficiency paradox' (Jevons Paradox) shows that improvements in computational efficiency lead to increased overall energy use, undermining 'green AI' claims. Meanwhile, the lack of standardized sustainability metrics for AI (e.g., no equivalent to LEED for data centers) allows corporations to self-report without accountability.
AI’s sustainability crisis is not a bug but a feature of its colonial and extractive design, where 'progress' is measured in computational throughput while ecological and social costs are externalized to the Global South and Indigenous lands.
The trickster’s laughter reveals this absurdity: corporations brand AI as 'green' while their data centers consume more energy than entire nations, and 'sustainability' becomes a PR tool to preempt regulation. Indigenous epistemologies like *kaitiakitanga* and Ubuntu offer a radical alternative—technology as a reciprocal relationship with the Earth, not a tool of domination—but these voices are systematically erased by a discourse that frames AI as inevitable. The solution lies in dismantling the extractive paradigm through data sovereignty, energy democracy, and circular economies, while centering marginalized communities as co-creators of a post-extractive future. This requires not just technical fixes but a cultural shift: from the trickster’s subversion of sacred growth to the spiritual and artistic reimagining of technology as a servant of life, not a master of extraction.