🌱 Is AI sustainable?

There is often much talk of the AI bubble when considering the sustainability of AI. (⏰ Expand on this)

But this video claims that despite the extraordinarily large planned capital investment in infrastructure, it is likely to be insufficient given the rapid development of AI capability such as agents which, in 2026, are predicted to move out of experimentation into production.

Unlike AI training, which is intensive but front loaded and diminishes upon implementation, AI agents do not merely replace or mimic human behaviour (occasionally querying a chatbot) but run persistently and operate at machine pace. This will require 'always-on' capacity that far exceeds historical demand.

The implication is that AI growth does not depend solely on model capability but on the infrastructure, and can be easily constrained by the latter.

The sustainability of AI has also been highlighted by the introduction of 🌱 Ads within AI models