The initial free-for-all phase of AI adoption, characterized by a practice called "tokenmaxxing" where employees were encouraged to use as much AI as possible, is rapidly coming to an end as companies face significant cost overruns. Despite the price of individual AI tokens decreasing, overall AI spending has soared due to increased usage and the proliferation of "agentic tools" that make numerous calls to AI models. Some firms, like Uber, blew through their entire 2026 AI coding budget by April, while Meta and AT&T have begun limiting employee access to AI tools such as GitHub Copilot and Anthropic.
Companies such as Walmart and Amazon have also implemented caps on internal AI agent use, with Amazon even discontinuing an internal leaderboard that ranked employees by AI usage after it led to inflated compute costs. Microsoft found some individual engineers spending between $500 and $2,000 monthly on Claude Code tokens alone. This shift has led to a new buzzword, "tokenminimizing," which reflects the corporate drive to control runaway AI expenses.
This cost-cutting trend has also spurred a new market for AI spend management solutions. Companies like -i are emerging to track, measure, and optimize GenAI investments, while existing tech giants are integrating AI cost management features into their platforms. The Tokenomics Foundation, backed by Accenture, IBM, Oracle, and JPMorgan Chase, is working to standardize AI budgeting metrics and define a framework for "tokenomics." However, even with these efforts, Goldman Sachs projects global token usage to multiply 24-fold by 2030, meaning companies need immediate solutions to manage their current budget woes.
While some companies like Databricks maintain unlimited AI budgets for engineers, confident in their efficient use, many others are exploring alternatives. This includes swapping expensive frontier models for cheaper or open-source options for simpler tasks to reduce bills without necessarily cutting back on AI use entirely. The sentiment is encapsulated by Microsoft CEO Satya Nadella's argument for swappable AI models, suggesting a future where less reliance is placed on a few dominant, potentially costly, models.
