{
  "id": 17983,
  "title": "Amazon spent $1.8 million on a failed AI project, and didn't notice the overrun for five months",
  "url": "https://urgent.news/2026/07/31/amazon-spent-1-8-million-on-a-failed-ai-project-and-didnt-notice-the",
  "topic": "ai",
  "section": "AI",
  "published": "2026-07-31T15:39:00.000Z",
  "source": {
    "name": "TechSpot",
    "slug": "techspot",
    "url": "https://www.techspot.com/news/113310-amazon-spent-18-million-failed-ai-project-didnt.html"
  },
  "original_language": "en",
  "account": "Amazon, in its pursuit of artificial intelligence for various operations, has encountered a growing issue: AI projects can quickly become expensive, sometimes without anyone noticing until the bill is substantial. Internal documents, as reported by the Financial Times, reveal that several AI-driven initiatives have exceeded their budgets, with one case involving a $1.8 million expenditure on a project using Anthropic's Claude Sonnet model to link author information with product listings. This project ultimately failed, and the overspend remained undetected for five months. Engineers describe these budget overruns as \"catastrophically expensive,\" contrasting them with the \"trivially cheap\" mistakes of traditional software systems. The root cause lies in the unique pricing structure of AI systems, which charge based on usage, measured in tokens, unlike the fixed costs of conventional software. This change introduces variable expenses that can escalate rapidly when systems are misconfigured or left unchecked. Multiple projects have suffered similar fates, with costs ranging from $134,000 to $541,000. Engineers are now implementing safeguards, such as automated controls and real-time spending tracking, to prevent such issues. Amazon acknowledges the learning curve associated with AI and claims it's actively working on cost efficiencies. However, the company emphasizes that these isolated incidents shouldn't be generalized to suggest widespread problems across the organization. Nonetheless, these cases underscore a broader industry trend as companies shift from fixed pricing to usage-based billing for AI, making cost management more complex and unpredictable. Amazon has already faced related setbacks, including AWS outages linked to AI coding tools, prompting the company to restrict their functionalities and reassess its heavy reliance on AI. Despite the financial impact being relatively minor for a company of Amazon's scale, the incidents highlight the need for new approaches to managing AI costs, diverging significantly from traditional software management practices.",
  "summary": "In one case, Amazon spent $1.8 million on a project that used Anthropic's Claude Sonnet model to match author information with product listings. The system ultimately failed, and spending exceeded the original budget by 860%. The issue went undetected for five months. Read Entire Article",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}