AI depends on computing infrastructure. The International Energy Agency’s Energy and AI report examines the relationship between data centres, electricity and AI deployment. It provides a useful starting point for treating energy as part of the story of intelligence.

When comparing systems, ask which resource is constrained. Training, everyday use and physical deployment have different requirements. A cheaper calculation is not automatically a cheaper overall service if other parts of the system grow more demanding.

Consider the whole chain: hardware, software efficiency, cooling, power availability and useful work delivered. Progress can come from better algorithms as well as additional resources. Claims about future scale are easier to assess when their assumptions about this chain are explicit.

Sources & further reading

Links point to original reporting or primary research.

  1. International Energy Agency · Energy and AI (2025)
  2. Vaswani et al. · Attention Is All You Need (2017)
See an error? Read our editorial and corrections policy.