AI Consultant Reveals: Unknown Company Accidentally Burned $500 Million on Claude in One Month Due to Failing to Set Usage Limits for Employee Access

[ref]

According to a May 28 report by Axios, an AI consultant revealed that one of its clients inadvertently spent approximately $500 million on Claude services in a single month, because the Claude licenses issued to employees had no usage caps, leading to runaway consumption. The company’s identity has not been disclosed, but Tom’s Hardware noted that overspending on this scale could only occur at one of the world’s largest enterprises. Quoting broader industry observations, the Axios report noted that corporate executives are beginning to question whether ever-rising AI spending can deliver real returns. Some companies find that employees use AI mainly to automate tedious, trivial tasks rather than doing valuable work—some even use AI to check the weather. Meanwhile, token consumption by AI agent tools can reportedly be up to 1,000 times that of a standard LLM query, further raising the risk of runaway costs. This incident clusters with several other AI overspending events in the same period: OpenClaw’s creator burned $1.3 million on OpenAI API fees in a single month; Uber’s CEO warned there is no clear link between token consumption and actual product delivery; and Amazon shut down Kirorank after employees “gamed token usage.” A Goldman Sachs research report previously estimated that the proliferation of AI agents could increase token demand by up to 24 times, making the cost of rapidly rolling out AI licenses without governance mechanisms increasingly clear. Axios | Tom’s Hardware https://www.tomshardware.com/tech-industry/artificial-intelligence/mystery-company-accidentally-blew-usd500-million-on-claude-in-a-single-month-failed-to-put-usage-limit-on-licenses-for-employees