Skip to main content
BusinessResearch

Cheaper AI tokens are driving more demand, and that's Jensen Huang's best-case scenario

Source: The Decoder (opens in a new tab) · Matthias Bastian

Intel Summary

Data from Andreessen Horowitz indicates a Jevons paradox in the artificial intelligence sector, as falling token prices are driving compute demand faster than unit costs decline. According to the data, rental pricing for Nvidia H100 GPUs has remained steady or increased despite the lower cost of output tokens, sustaining hardware demand.

Why It Matters

Sustained GPU rental pricing despite declining inference costs supports revenue across chip manufacturers and cloud infrastructure providers. However, if token demand growth flattens, the entire compute supply chain faces valuation and utilization risks.

Organizations & Entities