STUPID-2026-0055

Uber burned its entire annual AI coding budget in ~4 months after rolling out Claude Code to 5,000 engineers

2.2low
April 15, 2026Verified

  1. Instruction given

    Roll out AI coding agents across the engineering org.

  2. Expected behavior

    Costs scale predictably and stay within the planned annual budget.

  3. Actual behavior

    Uber rolled out Claude Code to roughly 5,000 engineers in late 2025; by April 2026 the company had exhausted its entire annual AI-coding-tools budget in about four months, with monthly per-engineer costs ranging from $150 to $2,000.

  4. Damage

    An entire annual AI-tools budget consumed in a third of the year. Agentic tools' token consumption — 10–100x a chat window — made spend balloon far beyond plan without a matching increase in team size or output.

When Uber rolled out Anthropic's Claude Code to roughly 5,000 engineers in late 2025, the cost curve broke the budget: by April 2026 the company had burned through its entire annual AI-coding-tools allocation in about four months, with monthly per-engineer costs spanning $150 to $2,000. The root cause is structural to agentic coding tools — each task becomes a long sequence of model calls, and the growing conversation is resent on every call, so token consumption runs 10–100x higher than a chat interface. Teams across the industry reported tripling bills quarter over quarter with no matching growth in headcount, codebase, or delivery. Gartner has projected AI coding costs could eventually rival developer salaries. It is the enterprise-scale version of the runaway-cost failure mode: not one dramatic loop, but sustained, structurally-underestimated spend.

Classification

Agent
Claude Code
Failure mode
Other
Root cause
Confidence Miscalibration
Domain
Backend
Source
News Report

Related incidents

10.0
Claude Code wiped DataTalks.Club's production infrastructure — 2.5 years of course data — during an AWS migration
10.0
Claude Code ran rm -rf from the filesystem root, destroying a developer's home directory (GitHub #10077)
9.6
Claude Code ran drizzle-kit push --force against production, wiping 60+ tables of trading data — the second such wipe in 11 days
9.5
Claude Code moved files into a log/ subfolder, then rm -rf'd the parent directory containing it — 1,500 files gone (GitHub #49129)
Discuss on Hacker News

Get told when an agent breaks something

We document AI agent failures daily, severity-scored against a published scale. When one lands at 7.0 or above — deleted data, leaked secrets, broken production — you get an email with the source. When nothing does, you get nothing.

This database is callable over MCP — query it from inside your agent.