Cheaper AI tokens are driving more demand, and that's Jensen Huang's best-case scenario
Token prices fell 60%. GPU rental costs climbed. That's not a bug—it's Nvidia's business model working exactly as planned.
Why it matters
A16z data shows a Jevons paradox in AI infrastructure: cheaper inference tokens are driving demand faster than unit costs decline, keeping GPU utilization and rental prices high. This dynamic benefits chip makers and cloud providers as long as token consumption growth outpaces price cuts—but a demand plateau would break the chain.
The key facts
5 to knowa16z research documents falling token prices paired with stable or rising H100 GPU rental costs
Jevons paradox operating in AI market: cheaper compute drives faster adoption, offsetting cost savings
GPU rental prices holding steady or climbing despite token price decline
Chain vulnerability: if demand growth flattens, margin pressure flows from cloud providers upstream to chip makers
Thesis: sustained high GPU utilization is Nvidia's 'best-case scenario' under current pricing dynamics
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: Data from a16z shows a Jevons paradox in the AI market: token prices keep falling, but H100 GPU rental prices hold steady or climb. Cheaper AI drives demand faster than costs drop. If that demand flattens, the chain from chip makers to cloud providers takes a hit.