Amazon accidentally spent $1.8 million using Claude for menial coding task, went 860% over budget — 'catastrophically expensive' coding blunders discovered in internal Amazon AI usage metrics

According to the Financial Times, several internal Amazon AI projects ran well over budget.

A failed Claude Sonnet deployment reportedly reached $1.8 million, about 860% above budget.

Other examples included an auditing tool with $541,000 in extra costs and a logistics system with $134,000 more.

Amazon said it is still experimenting with AI and that the cases do not reflect companywide usage.

The report underscores how per-token pricing and agent-based workflows can turn routine tasks into expensive failures.