Crypto Briefing은 a16z 데이터를 인용해, AI 토큰 지출이 122%에서 160%까지 증가한 반면 토큰 가격은 67% 하락했다고 보도했다.

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 between 122% and 160%. The data appeared in the firm's Charts of the Week, published October 31, 2026.

In this context, a "token" is the billing unit for AI model inference: a small chunk of text a model reads or writes. Providers charge by the token. Per-token prices dropped 67% since May, but total spending over the same period rose between 122% and 160%. The unit got cheaper, but the total bill grew, because usage expanded fast enough to outweigh the price cut.

The biggest driver appears to be AI agents. According to the a16z data, agents consume nearly five times as many tokens as human users. A person asks a chatbot a question and reads the answer. An agent runs a longer task, carrying more context and taking more steps, and each step consumes tokens. Agent token usage has also grown roughly 14 times since February 2026, pointing to a shift toward larger, more automated workflows rather than one-off prompts.

A16z ties the trend to the Jevons paradox, the idea that when a resource becomes more efficient or cheaper to use, total consumption can rise instead of fall. Cheaper inference makes new use cases economical. Tasks that were too expensive to automate at May prices now pencil out at current rates.

The supply side adds a twist. While token prices kept falling, GPU rental prices moved sharply higher. According to the data, rental costs for Nvidia's B300 rose 106% since mid-July 2026, and B200 rental prices climbed 66% over the same period. The gap between falling output prices and rising hardware costs has been labeled a "rack spread."

The a16z charts also show concentration among buyers. A relatively small group of top users accounts for a significant share of total token spending.

For AI model providers, the data is a mixed blessing. Volume is surging, but revenue per token is collapsing, and the cost of the GPUs underneath is rising. For GPU owners and cloud operators, the picture is friendlier. A 106% jump in B300 rental rates since mid-July suggests demand for compute is outpacing supply, at least for the newest chips.

For companies building on AI, each individual call is cheaper than it was in May. Agent-heavy products, however, can generate bills quickly, since agents use nearly five times the tokens of a human user. The concentration point also matters. When a small group of heavy users drives most of the spending, the market becomes sensitive to their decisions.

The a16z charts tell a straightforward story with a counterintuitive ending. AI inference got two-thirds cheaper in roughly five months, and the response was to use so much more of it that total spending more than doubled.