Amazon Blew $1.8M on Claude: A Costly Lesson in AI Agent Oversight

Amazon recently attempted to use Claude Sonnet to automatically populate author information on its website, but the seemingly simple task ended up costing $1.8 million—exceeding the budget by 860%—and was only discovered after five months, with the deployment ultimately failing. According to Anthropic's public pricing of $3 per million input tokens and $15 per million output tokens, $1.8 million could have burned up to 600 billion tokens, roughly twice the size of GPT-3's entire training corpus. This incident is not isolated; similar cases have occurred inside Amazon multiple times, often going unnoticed for long periods. Despite these failures, Amazon continues its aggressive automation push, with CEO Andy Jassy announcing a planned 2026 capital expenditure of about $220 billion, mostly for AWS, custom AI chips, and power infrastructure. In its Q2 2026 earnings, AWS reported net sales of $42.2 billion, up 37% year-over-year, contributing about 60% of Amazon's total operating profit. Meanwhile, Amazon has cut about 30,000 jobs since October 2025 and aims to automate around 75% of warehouse operations by 2033, which could reduce U.S. employment by over 600,000. External observers, including Nobel laureate Daron Acemoglu, warn that Amazon could become a 'net job destroyer.' The issue of runaway token costs is not unique to Amazon; Meta, Uber, and others have faced similar overages. Meta even built an internal leaderboard called Claudeonomics, tracking token usage among 85,000 employees, with a monthly bill equivalent of $221 million, leading to token usage caps and a centralized monitoring platform. Sam Altman revealed that OpenAI's internal heavy users consume up to 100 billion tokens per month, and one employee used 210 billion tokens in a single week. A survey found only 26% of companies have full visibility into AI costs. The article also cites the historical case of Knight Capital, which lost $440 million in 45 minutes due to a flawed automated trading system, leading to its collapse, as a cautionary tale about the dangers of unregulated automation.