AI's cost just shifted—not down, but from training to inference.
Training happens once. Inference runs forever. Google: 9.7 → 1,300T tokens/mo in 18 months. ~134x.
Unit cost fell to 1/280th, yet total spend explodes. Jevons, not Moore.I wrote the axis.
https://t.co/Wa9e4We5sE
Irreversible tipping points loom, yet AI infrastructure booms overshadow the threat of climate change.
AI inconvenient truth ignored by AI labs, government, and media alike.
My new open-source book tackles this reality head-on.
https://t.co/StUt8syTSI
@WSJ The IPO is the headline. The trajectory is the real story.
ARR $14B to $30B+ in under two months.
Dario: "We planned for 10x, we saw 80x."
I decoded this $1T trajectory across 6 layers — open-source, free:
https://t.co/u9qb6hN9WE