Virtuous Cycle Between AI and Distributed Renewable Energy
Discussion
< "Cheap clean energy can be a public abundance or a private monopoly; AI can serve citizens or surveil them; precision fermentation can break the cruelty of industrial livestock or hand it to three patent-holders; autonomous transport can free cities from the car or lock them into platform rents. The technology does not decide. The rules around it do." >
Nafeez Ahmed:
"A standard Google search uses about 0.3 watt-hours. A generative AI query uses 10 times that. The Electric Power Research Institute projects that AI-related US power demand could rise tenfold by 2030. Frontier training costs have grown 2.4 times every year since 2016. Amazon, Google, Meta and Microsoft will spend $700 billion on capital expenditure in 2026, most of it on AI data centres – more than the GDP of all but the top 20 national economies.
Such demand cannot be met by a centralised, fossil-fuelled grid running on the dwindling energy returns described above: the cost of electricity from a degrading fossil system rises while the cost of the AI on top of it is forced to fall.
A different physical configuration works. Solar panels, wind turbines and batteries are themselves semiconductors – close cousins of the chips the AI itself runs on – and they share the same exponential learning curves. Pair the solar revolution with the AI revolution and the result is a virtuous cycle: cheap electrons train the AI, and the AI in turn coordinates a grid of millions of intermittent renewable sources – rooftop panels, wind turbines, electric-vehicle batteries, smart appliances – at a speed no human-managed grid could match."