infrastructureScore 60/100Watch
Andrej Karpathy identifies LLMs as utility-like infrastructure requiring massive CAPEX for training, comparable to electricity grids
Andrej Karpathy· Independent· AI· 2026-06-03· about OpenAI, Gemini (Google), Anthropic
“LLM labs, like OpenAI, Gemini, and Property, et cetera, they spend CAPEX to train the LLMs. And this is kind of equivalent to building out a grid. And then there's OPEX to serve that intelligence over APIs to all of us.”
Why it matters
LLM labs are committing significant capital expenditure to train models, similar to utility infrastructure build-outs. This signals sustained heavy spending on compute, chips, and data centers.
Investment implication
Companies supplying compute infrastructure (NVIDIA GPUs, cloud services, cooling systems, power management) will see sustained demand as LLM labs continue CAPEX cycles to build and maintain model training infrastructure.