Early report based on the official source The Decoder (DE): According to an analysis of data from Ornn, Silicon Data and Bloomberg (as of August 2026), cited by a16z, the AI market shows a classic Jevons paradox. Token prices, the billing unit for AI models, keep falling. Rental prices for H100 GPUs, the hardware behind them, stay high or rise.
Quick Facts
- Cheaper tokens enable AI agents, automation and new applications, so volume grows faster than per-unit costs fall.
- It remains unclear how much demand comes from human usage and how much the systems generate themselves; agentic AI consumes tokens at a particularly high rate.
- The system rests on the assumption that AI usage grows fast enough to offset falling token prices. If demand flattens, chip makers, memory suppliers, energy providers and cloud companies would be affected.
- Financial markets react sensitively: US stocks fell on reports that OpenAI's annualized revenue may be lower than previously reported.
Analysis: For German companies this means falling model prices do not automatically mean lower infrastructure costs. Those planning their own workloads should budget computing costs separately from token prices and keep open questions about future demand in view.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
