Why Smarter AI Models Could Drive Up Compute Prices 10x
Source
Dwarkesh Patel
Author
Dwarkesh Patel
Date
Terms in this piece · Glossary
inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
Why it matters
If better models make compute more expensive rather than cheaper, the cheap-tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition → assumption behind most AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → designs breaks. The argument is worth testing against your own inferenceRunning a trained model to get answers — the phase where AI is actually used, as opposed to trained.Full definition → budget and retry-heavy loops.
