The most important indicator in Silicon Valley last month was tokens burned – units of measurement for computing power used by AI models. CEOs were giving employees Matthew McConaughey.these are new numbers, you need to increase these numbers” exit from He founded Wall Street. Now they are asking their employees to hit the brakes.
according to A report from the Wall Street Journalcorporate leaders have realized that burning AI tokens actually costs money, and doing it almost exclusively for the sake of doing it is not really a great business strategy. Good thing these guys get paid tens of millions of dollars a year to figure this out.
Earlier this week, Uber CEO Dara Khosrowshahi said, “it becomes difficult to justify” value of AI initiatives within the company, as the output was not commensurate with the token burn rate, while admitting that part of the reason they went so ham on the token burn in the first place was because they could “feel” that AI was free.
Apparently not.
An anonymous AI consultant told Axios one of his clients accidentally spent half a billion dollars for a month because it never bothered to put a usage limit on the employee’s access to Anthropic’s Claude. I mean… a lot. As much as it strengthens the credibility. The magazine didn’t find anything too egregious, but it did hear about one financial institution that saw its employees burning hundreds of thousands of dollars worth of tokens a month, as employees used high-level models to ask basic questions and engage in inane conversations.
It’s almost always how this very stupid cycle would go to justify the cost of AI. Corporations have already spent tons of money to adopt these systems and they need to justify the cost. For this, they encourage employees to make the most of it. In turn, workers do it – it doesn’t even make sense to use AI for a task.
Meta killed the token burning leaderboard After being leaked last month, it was revealed that the top “Token Legend” managed to burn 281 billion tokens in one month, which was more than the computational amount needed to multiply the entire Wikipedia 33 times. Amazon joined the retreat this week, According to the Financial Timesaxed the company’s scoreboard of employees who used the company’s internal AI tools the most – a decision reportedly made after it was revealed that employees were giving AI agents meaningless tasks to take their place on the leaderboard.
It is clear that the corporate world is willing to frivolously burn money to justify existing cash burning pits. It turns out they have their limitations. You can only use “tokens burned” as a metric in many quarterly earnings calls before shareholders start wondering what the price tag of all those tokens has resulted in.
