Cheaper Tokens, Bigger Bills (www.apollo.com)

The Silicon Data Token Expenditure Index has revealed a doubling of total spending on large language model (LLM) usage since late 2025, despite the cost of individual tokens plummeting over 90% since 2023. This phenomenon exemplifies Jevons paradox, where a decrease in the price of a resource leads to increased consumption rather than reduced expenditure. As LLM tokens become more affordable, companies are not cutting back; instead, they are deploying more AI agents, automating a wider array of workflows, and generating greater amounts of code, resulting in surging aggregate costs. This trend is significant for the AI/ML community as it underscores the profound impact of economic accessibility on AI deployment strategies. As organizations embrace cheaper tokens, they’re able to expand their AI operations without facing financial constraints. This growth could accelerate innovation and productivity across various sectors, as firms leverage the technology to optimize processes and enhance efficiency. The implications of this shift suggest a pivotal moment for AI adoption, where accessibility might drive widespread integration of sophisticated AI solutions into everyday business practices.