Decoupling Thesis

Definition

The decoupling thesis argues that with enough efficiency, grid decarbonisation, and AI-driven optimisation, GDP can grow while resource use and emissions fall. It distinguishes relative decoupling (resource use grows more slowly than GDP) from absolute decoupling (resource use falls in absolute terms as GDP grows). The strong version claims absolute decoupling is achievable at global scale.

Why it matters

The thesis is internally consistent, and it underwrites most mainstream sustainability policy. Its weakness is not logical but empirical: it holds only while the total scale of the system remains roughly linear. Against exponential compute and demand growth, the promise of decoupling is perpetually deferred by fresh waves of demand before any saturation point is reached.

How it appears in AI systems

In climate policy that pairs decarbonisation targets with unbounded GDP growth assumptions; in AI vendor claims that model efficiency alone will offset infrastructure expansion; in national inventories that record relative decoupling but omit imported emissions and induced demand.

Examples

• Application-level AI efficiency reported while training and inference infrastructure emissions grow faster.
• Grid decarbonisation announced alongside new fossil generation contracted for datacenter siting.
• Sectoral 'green growth' figures that stop at the border and ignore embedded carbon in imports.