Just last month, the most important metric in Silicon Valley was tokens burned—the units of measurement for the computing power being used by AI models. CEOs were giving employees the Matthew McConaughey “those are rookie numbers, you gotta pump those numbers up” speech from The Wolf of Wall Street. Now, they’re asking their staff to pump the brakes.

According to a report from the Wall Street Journal, corporate leaders have realized that it actually costs money to burn AI tokens, and doing almost exclusively for the sake of doing it with no other goal in mind is actually not a great business strategy. Good thing these guys get paid tens of millions of dollars a year to figure these things out.

Earlier this week, Uber CEO Dara Khosrowshahi said that it was “getting harder to justify” the cost of AI initiatives within the company because the output was not keeping up with the token burn rate, while acknowledging that part of the reason that they went so ham on token burning in the first place is that it “can feel” like AI is free.

It’s not, as it turns out.

An anonymous AI consultant told Axios that one of its clients accidentally spent half a billion dollars in a single month because it never bothered to put a usage limit on employee access to Anthropic’s Claude. That is… a lot. Like to the point of straining credulity. The Journal didn’t find anything quite that egregious, but did hear about a financial institution that saw its employees burn through hundreds of thousands of dollars worth of tokens per month, with employees using premium-tier models to ask basic questions and have inane back-and-forth conversations.

That was pretty much always going to be how this very dumb era of justifying AI expenditure was going to go. Corporations have already spent tons of money to embrace these systems, and they need to justify the spend. To do so, they encourage their employees to use it as much as possible. In turn, the employees do—even when there is no point in using AI for a task.

Meta killed its token-burning leaderboard last month after it leaked, revealing that the top “Token Legend” managed to burn 281 billion tokens in a month—more than the amount of compute it would take to reproduce the entirety of Wikipedia 33 times over. Amazon joined in that rollback this week, according to the Financial Times, axing its scoreboard of employees who were using the company’s internal AI tools the most—a decision that reportedly was made after it became obvious that employees were giving AI agents pointless tasks just to hold their position on the leaderboard.

It’s clear the corporate world is willing to frivolously light money on fire in an effort to justify their existing cash burn pits. Turns out they do have limits, though. You can only use “tokens burned” as a metric at so many quarterly earnings calls before shareholders start to wonder what the price tag of all those tokens adds up to.