27/05/2026: POLICY BRIEF – AI Decarbosation (UK Focused):
The Hidden Climate Problem: Why the software running AI might be a bigger problem than you think.
| Summary: – Britain wants to be an AI superpower in the global race while also achieving its Net Zero goals. As of now, those two goals, though achievable, present challenges that require a unifying methodology: the government’s own estimates of emissions from new AI computers increased more than 100-fold after public pressure to disclose the information. – Most attempts to make AI “greener” focus on the chips it runs on and the size of the models currently being operated. The code itself, which is the software telling the chips what to do, is mostly ignored. – Research findings (Banerjee, 2026) show that rewriting the same AI model in more efficient software can reduce its electricity use by 4 to 24 times, with no real loss in accuracy. The same model, running on the same chip, with minimal code changes, can achieve more efficient and greener results. – These savings are widely achievable using existing knowledge. They require no new power stations, no new chips, and no new planning permission. -The POLICY CLUB recommends that UK policymakers require AI systems to disclose the computational energy used by their software, in the same way buildings and other physical assets publish energy ratings. |
| POLICY BRIEF – UK | Download |
27/06/2026: POLICY BRIEF – Value Transfer for Industrial Decarbonisation – Forthcoming