AI token spending jumps 122% to 160% as prices fall 67%, a16z data shows

AI just got a lot cheaper, and companies responded by spending a lot more on it. Venture firm a16z reports that average AI token prices have fallen 67% since May 2026, while total token spending has climbed more than 120%, with estimates ranging from 122% to 160%.

Token Usage Surge

In this context, a "token" is the billing unit for AI model inference - a small chunk of text a model reads or writes. The data shows a stark divergence: per-token prices dropped 67% while total spend rose between 122% and 160%.

The biggest driver appears to be AI agents, which consume nearly five times as many tokens as human users. Agent token usage has grown roughly 14 times since February 2026, pointing to a shift toward larger, more automated workflows.

Jevons Paradox

a16z ties the trend to the Jevons paradox, where when a resource gets cheaper, total consumption can rise instead of fall. Cheaper inference makes new use cases economical, with tasks that were too expensive to automate at May prices now feasible.

GPU Costs Rising

While token prices dropped, GPU rental prices moved sharply in the opposite direction. Rental costs for Nvidia's B300 rose 106% since mid-July 2026, while B200 rental prices climbed 66% over the same period. This creates a "rack spread" - the gap between falling output prices and rising hardware costs.

Market Implications

For AI model providers, the data presents a mixed blessing: volume is surging, but revenue per token is collapsing while GPU costs rise. For GPU owners and cloud operators, the picture looks friendlier with compute demand outrunning supply. For companies building on AI, individual calls are cheaper, but agent-heavy products can quickly accumulate high bills due to token consumption.

The concentration among buyers also deserves attention, as a relatively small group of top users accounts for a significant share of total token spending, making the market sensitive to their decisions.