LLM coding costs may exceed $1,000 per $100 in revenue, raising affordability questions

Summary

This article analyzes the economics of coding with Large Language Models (LLMs) like Claude Code and OpenAI Codex, arguing that the cost structure makes LLM-assisted coding unaffordable at scale. The author suggests that AI companies like Anthropic and OpenAI may be spending over $1,000 in compute/infrastructure costs for every $100 of revenue they generate from users. The piece is a long-form analysis questioning whether LLM-coding can be a viable "killer app" given the massive infrastructure costs involved, and casts doubt on the narrative that AI will build itself or become self-sustaining.