Just a year ago, corporations were racing to provide access to AI models, encouraging employees to use them to the fullest. Tokens were seen as a free corporate bonus, and internal rankings rewarded the most active users. Today, the situation has changed dramatically: executives are hastily cutting off access, imposing strict limits, and frantically revising budgets. Tokens have transformed from free fuel into the main line item of unforeseen expenses.
So what went wrong? It all started with monetization. Until March 2023, public AI projects like OpenAI operated on a subscription basis. But with the release of GPT-4, Sam Altman's startup introduced token-based pricing: $30 per 1 million input tokens and $60 per 1 million output tokens. From that moment, the token became the unit of measurement for the cost of AI tools. The period from 2023 to 2025 was the "era of subsidization"—corporate developers paid, but prices remained soft, stimulating mass adoption. However, since the spring of 2026, all major providers—Anthropic, GitHub Copilot—have fully transitioned to monetizing requests, citing high data processing costs. The era of the "tokenapocalypse" has begun.
Jevons Paradox: Prices Fall, Costs Rise
The paradox is that over the past three years, the cost of tokens has dropped by approximately 98%. Yet, total enterprise AI spending has surged by 320%—from $1.2 million in 2024 to $7 million in 2026. The reason is an explosive increase in usage volumes. Agentic technologies are particularly straining resources. A simple query in 2023 cost $0.04, while a query to a full-fledged agentic system already costs $1.2—30 times more. For complex workflows, costs increase by 50 to 500 times. This is a classic Jevons paradox: a decrease in the price of a resource leads to an increase in its overall consumption, and in the case of AI, the effect has been hypertrophied.
Case Studies: From Uber to Microsoft
Real-world examples vividly demonstrate the scale of the problem. At Uber, where AI tools were actively deployed, the CTO personally spent $1,200 on tokens during a single two-hour demo. The company, which had encouraged employees to use AI "as much as possible," was forced to set a limit of $1,500 per month per employee. Microsoft, after granting thousands of developers access to Anthropic's Claude Code, shut down the experiment six months later. Internal surveys showed that engineers preferred the competitor's tool, but its excessive use hit expenses. GitHub Copilot even suspended new user registrations, as the costs of payment agents exceeded the value of subscription plans.
Particularly telling is the case of an anonymous company that forgot to set limits on Claude Opus. At the end of the month, the firm received a bill for $500 million. One developer spent over $40,000 on tokens in a single month. As Priceline's CIO figuratively noted: "It's like a crack cocaine epidemic. They give you a taste to get you hooked, and now you're addicted."
The New Reality: Discipline and Standards
Large corporations like Amazon and Microsoft will undoubtedly survive the "tokenapocalypse" thanks to direct investments in AI projects. But for more modest firms, the situation is critical. It is against this backdrop that the Linux Foundation introduced the Tokenomics Foundation project—a new standard designed to instill "discipline" in managing AI costs, analogous to FinOps for cloud services.
My expert conclusion: Tokens have become a new budget line item, and ignoring them is an unaffordable luxury. Companies that fail to learn how to forecast and control these expenses risk facing financial turmoil. As AI evolves, such spending will likely become as routine as cloud subscription costs, but the path to this stability will be rocky. Rational consumption and the adoption of standards are the only way to survive the era of the "tokenapocalypse."