I think the solution is clean baseload power- advanced nuclear (SMRs), next-gen geothermal, and true 24/7 renewables paired with massive storage. Also, "carbon-aware computing" that routes workloads to where/when green energy is abundant. Efficiency helps, but the industry must build new clean energy at an unprecedented scale. Maybe we should ask AI for the solution.

Agreed on all counts. The post isn't arguing against AI infrastructure, just for more transparency about the current costs and more intentionality in how we use it while that clean baseload buildout happens.

The carbon-aware computing angle is actually in the full piece. It's one of the more promising near-term approaches since it doesn't require waiting for new plants to come online.

The irony of asking AI to solve its own energy problem isn't lost on me. Though to be fair, this post did use Claude for research, which I estimated at roughly 26 watt-hours and 8 gallons of water. Figured I should practice what I preach and disclose that.