Amazon AI Project Costs Exceed Budget by 860 Percent
The cases show that AI adoption can dramatically increase costs for tasks that were cheap with traditional coding, and Amazon is now building guardrails to prevent such overruns.
Reporting from 1 source: GIGAZINE.
Amazon reported internal cases where AI adoption led to runaway spending, including a project using Anthropic's Claude Sonnet that cost $1.8 million and exceeded its budget by 860 percent before failing. Another project incurred $541,000 in unexpected costs, and a logistics project overspent by $134,000. Amazon is building guardrails to monitor and limit AI spending.
Amazon told employees at a late July meeting that switching from traditional programming to AI models caused unplanned spending. One project used Claude Sonnet to match product listings with author information on the e-commerce site, spending $1.8 million before the effort failed. The cost exceeded the project budget by 860 percent and took five months to surface.
Other cases included $541,000 in unexpected costs for a financial audit tool and $134,000 for a logistics project aimed at improving delivery speeds, the latter taking over two weeks to detect. Amazon said it is building guardrails to automatically monitor and limit AI usage. The company argued that these are small, isolated cases and that the excess AI costs amount to less than 0.1 percent of one month's revenue.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.