An unnamed enterprise customer ran up a roughly $500 million bill on Anthropic's Claude in a single month after setting no usage limits or spending caps for its staff.
Key Points:
- One company spent about $500 million on Claude in a single month with no spending caps in place.
- The bill traced back to unrestricted employee access and token-heavy automated workflows.
- Microsoft, Uber and Amazon have all moved to curb their own AI costs.
Claude Bill Tops $500 Million
An AI consultant revealed the figure on May 28, describing what may rank among the costliest IT governance lapses in recent memory. The client had handed thousands of employees open access to the platform, with no spending caps, no usage dashboards and no alerts to flag a runaway bill. Enthusiastic adoption turned, within a few weeks, into a burn rate few finance teams have ever had to explain.
Costs climbed fastest among the heaviest users.
Engineers who run agentic workflows or feed long-context prompts can rack up hundreds or thousands of dollars apiece each month, since every request, prompt and follow-up draws on a fresh pool of metered tokens.
Spread that across an unrestricted workforce of thousands, and the meter simply never stops.
Also Read: Cardano Whales Seize 67.5% Of ADA Supply, A Six-Year High
Microsoft, Uber Tighten AI Budgets
The episode is no outlier across corporate America. Microsoft capped most of its internal Claude Code licenses after per-engineer costs ran between $500 and $2,000 a month, and it is shifting work toward in-house tools.
Uber exhausted its entire 2026 AI budget by Apr. after an aggressive rollout of AI coding tools across its teams. Its chief operating officer said the spending had grown harder to justify against everyday priorities. Amazon, meanwhile, shuttered an internal usage leaderboard that staff had gamed with low-value prompts.
Many firms treated these tools like flat-fee software through 2024 and 2025. They badly underestimated how sharply usage-based pricing scales with model choice, context length and around-the-clock autonomous agents.
Anthropic Revenue Keeps Climbing
Anthropic does sell a full set of enterprise controls, including admin dashboards, per-user limits and compliance tools. Those tools need switching on first, and in this case they plainly were not. The episode is now steering more buyers toward hard caps, role-based access and live monitoring.
The bill lands during a steep climb for the company.
Anthropic closed 2025 near a $9 billion annualized revenue run rate, with about 80% of that revenue drawn from enterprise clients rather than consumers. Microsoft alone is on pace to spend roughly $500 million a year on its models, a sign of how fast that base is widening.
Read Next: Cisco Research Shows Frontier AI Models Failing Under Multi-Turn Attacks