Just a year ago, corporations were pushing employees to use AI to the fullest: tokens were handed out as bonuses, and rankings rewarded those who "burned" the most. Today, the situation has changed dramatically. Managers are panicking, cutting off access to models and imposing strict limits while revising budgets. Tokens, once free fuel for innovation, have become the main line item for unforeseen expenses. How did corporations miss this crisis?
What is a token and why do you need to pay for it
In the context of artificial intelligence, a token is the smallest unit of memory that a language model processes. One token can correspond to several characters or an entire word. Models use them to process queries and generate responses: on average, 100 words are broken down into 130–140 tokens. They are divided into input (prompt) and output (response).
For a long time, tokens were perceived as an abstract unit of information until March 2023, when OpenAI released GPT-4 and set public pricing: $30 per 1 million input tokens and $60 per 1 million output tokens. From that moment, the era of monetization began. From 2023 to 2025, companies subsidized AI usage, but by spring 2026, all major providers (Anthropic, GitHub Copilot) had fully switched to pay-per-query models, citing high data processing costs. This period is already being called the beginning of the "tokenapocalypse."
Price drops, demand rises: Jevons paradox in action
The paradox is that over the past three years, the cost of tokens has fallen by about 98%. However, according to industry analyses, corporate spending on AI has skyrocketed by 320% — from $1.2 million in 2024 to $7 million in 2026. It all comes down to a multiple increase in usage volumes, especially due to agentic technologies. A simple query in 2023 cost $0.04, while a query to an agentic system already costs $1.2 (30 times more). For complex workflows, costs increase by 50 to 500 times. Economists call this the Jevons paradox: a decrease in the price of a resource leads to an increase in its consumption, but in the case of AI, the effect has been hypertrophied.
Who miscalculated: real cases
Uber: limits after euphoria
Uber actively implemented tools like Claude Code and Cursor. In May, the CEO boasted that agents were writing 10% of the code. But within a month, the company set a limit of $1,500 per month per employee. The CTO admitted that he personally spent $1,200 on tokens in a single two-hour demo, exhausting the annual programming budget. Developers now control their own spending.
Microsoft: failed experiment with Claude
In December 2025, Microsoft gave thousands of employees access to Anthropic's Claude Code. The tool was so good that it was used constantly, extending beyond engineering teams. Six months later, Microsoft ended the experiment, ordering a migration to its own Copilot. According to analysts, this is a signal: the economics of corporate AI coding at current token prices don't work. The tools are so effective that their consumption spirals out of control.
Amazon: new contract and hidden risks
Amazon restructured its partnership with Anthropic, switching from paying for compute hours to paying for tokens. This could increase costs for Claude models used in Alexa, the Kiro tool, and other services. For giants like Amazon and Microsoft, the "tokenapocalypse" is merely an inconvenience, but for smaller firms, it could be fatal.
Costly mistakes
One anonymous company forgot to set limits on Claude Opus usage and received a $500 million bill. Another developer spent $40,000 on tokens in a single month. According to analysts, token consumption per developer grew 18.6 times over nine months, and most companies still cannot measure the return on these costs.
Not the foggiest future
The term "tokenapocalypse" exaggerates the problem. Large corporations will pay the bills, and small companies will adapt through rational consumption. The Linux Foundation has already introduced the Tokenomics Foundation project — a new standard for managing AI costs, similar to FinOps for cloud services.
My view as an analyst: We are on the cusp of AI market maturity. Tokens are the new currency, and companies that fail to learn how to manage them risk being left behind. Those who implement discipline in accounting and optimization will gain a tremendous competitive advantage. In the next two years, we will see consolidation: only those who can predict and control their AI spending as easily as they currently manage basic corporate subscriptions will survive.