A Theory of Training Profit-Optimal LLMs

In three linesEconomic model combining scaling laws and microeconomic theory to characterize profit optimization in LLM training. Analyzes how model size, token budget, and computational costs interact. In compute-bound regime, optimal spending tracks hardware efficiency (FLOPs/$) near-linearly. In data-bound regime, it scales as D²/E.

Summary generated by Claude — human-verified