AI Token Spending More Than Doubles as Prices Fall
AI token spending has more than doubled since May 2026, even as average AI inference token prices fell 67%, according to a16z data published on 31 October. Estimates put the increase in total spending at 122% to 160%.
The growth is linked to AI agents, which use nearly five times as many tokens as human users. Agent token usage has risen about 14-fold since February, suggesting more automated, multi-step workloads are driving demand. The trend reflects the Jevons paradox: lower costs can encourage enough additional use to raise total spending.
However, compute costs are also rising. Nvidia B300 rental prices climbed 106% since mid-July, while B200 prices rose 66%. This squeeze between falling token prices and more expensive GPUs could pressure AI inference providers, while benefiting GPU owners and cloud operators. A relatively small group of heavy users also accounts for a significant share of spending, making demand sensitive to their decisions.
Neutral
The report describes AI usage and compute economics, not a direct change to cryptocurrency prices, regulation, or blockchain activity. Its immediate effect on broad crypto trading is therefore likely limited, supporting a neutral classification. Traders may still watch AI-related crypto tokens for sentiment spillover, particularly if investors connect rising AI adoption with demand for decentralised computing or AI projects. But the article provides no evidence of increased blockchain use or capital flows into those tokens.
In the short term, the figures could reinforce interest in AI and semiconductor themes, while higher GPU rental costs may raise concerns about margins for AI service providers. Crypto assets may react mainly through broader risk appetite or sector-specific speculation rather than a clear fundamental catalyst. As in past episodes when technology-investment narratives lifted AI-linked tokens, any move could be volatile and sentiment-driven.
Over the longer term, expanding AI workloads and demand for compute could support narratives around decentralised GPU networks and AI-related crypto projects. That would depend on measurable adoption, revenue, and token utility; rising AI spending alone does not establish those benefits. The concentration of spending among a small group of users and the widening gap between token prices and hardware costs are also risks. Overall, the news has an indirect and uncertain connection to crypto market stability, rather than a clearly bullish or bearish signal.