Enterprises Scale Back AI Spending as Token Costs Soar

Enterprises are scaling back artificial intelligence spending as the era of subsidized pricing ends and operational costs skyrocket. The financial pressure is driven largely by the adoption of AI agents, which consume significantly more tokens than standard chatbots to execute complex tasks such as coding and file management. In some instances, firms have seen AI token costs exceed employee salaries within months, a phenomenon described as tokenmaxxing. This trend is further compounded by global shortages in computer chips and data center capacity.

Meta and Uber have emerged as prominent critics of indiscriminate AI adoption. Meta's Chief Technology Officer Andrew Bosworth issued a memo advising staff against using AI tools without a clear purpose, reversing previous internal encouragement of high token usage. Simultaneously, the Chief Operating Officer of Uber stated that AI spending has not yielded a noticeable increase in productivity.

To mitigate these expenses, businesses are shifting toward free open-source models, smaller industry-specific models, or breaking complex tasks into smaller steps to utilize cheaper mini-models. Meanwhile, leading AI providers including OpenAI and Anthropic are raising prices to achieve profitability as they prepare to go public later this year to attract new investors.