Home / Blog / Amazon Races to Fix an AWS Billing Bug That Showed Charges in the Billions
Cloud

☁️ Amazon Races to Fix an AWS Billing Bug That Showed Charges in the Billions

By Dillip Chowdary • Jul 17, 2026 • Source: TechCrunch

Amazon spent Thursday and Friday scrambling to fix a bug in the AWS billing portal that showed some customers charges ranging from the merely alarming to the frankly impossible — including one account that briefly displayed a month-to-date estimate of around $2.5 billion.

Amazon traced the problem to “a recent change” in its billing computation subsystem and stressed that the erroneous numbers “do not reflect actual usage and charges.” In other words, nobody is actually on the hook for a nine- or ten-figure cloud bill; the meter that renders those figures simply broke.

The company was cautious about specifics. It did not say how many customers saw inflated totals, whether any accounts were suspended or throttled on the basis of the bad numbers, or whether corrections would be applied automatically. Amazon estimated the fix would take “several more hours” and declined to comment further.

For finance and platform teams, the incident is a reminder that a billing dashboard is production software too. When cost-anomaly alerts, spend caps, and automated budget kill-switches all read from the same subsystem, a display bug can trigger very real downstream automation — paging on-call engineers, freezing deploys, or shutting off workloads — even when no money has actually moved.

Key details

  • What broke: The AWS billing portal showed inflated month-to-date charges to some customers on Thursday and Friday.
  • Worst case seen: One customer reported an estimate of roughly $2.5 billion.
  • Root cause: Amazon attributed it to a recent change in its billing computation subsystem.
  • Amazon's line: The charges “do not reflect actual usage”; a fix was expected within several hours.

Why it matters

A cloud bill is only as trustworthy as the pipeline that computes it. Teams that wire budget alarms and automated spend controls directly to the billing API should treat this as a live-fire drill: build in sanity bounds so a single upstream glitch can't cascade into a self-inflicted outage.

Source: TechCrunch. Reporting cross-referenced by Tech Bytes on Jul 17, 2026.

🔎 More interesting news

Stay Ahead

Free Tools