Some Amazon cloud customers experienced a massive shock late Thursday and Friday morning. They logged into their Amazon Web Services (AWS) accounts and found jaw-dropping bill estimates. Suddenly, these users owed millions, hundreds of millions, or even $2.5 billion for a single month of cloud computing.
Fortunately, the astronomical numbers are completely fake. Amazon confirmed the issue and assured customers they do not actually owe these massive sums. The staggering billing estimates simply do not reflect actual cloud usage or charges.
What Caused the Billion-Dollar AWS Error?
The crisis started early Friday at 7:38 AM PKT. Amazon began displaying highly inaccurate billing data inside its AWS Billing and Cost Management Console. Consequently, panicked customers quickly took to platforms like Reddit to share evidence of their ruined budgets.
Amazon eventually identified the root cause as a unit pricing glitch hidden within its estimated billing computation subsystem. Initially, engineers attempted to roll back a recent system change to fix the problem. However, this early mitigation failed entirely. To stop the fake bills from inflating further, Amazon completely paused all estimated billing computations globally.
Despite the sheer scale of the panic, Amazon kept its official communications strictly limited. Amazon spokesperson Aisha Johnson simply pointed to the AWS status page. She outright refused to answer questions regarding the technical bug. Furthermore, Amazon declined to confirm whether the system automatically suspended any customer accounts due to the massive, fake negative balances.
The Fix & What Customers Must Do
Currently, Amazon engineers are backfilling corrected data across the platform. The company expects to fully recover all affected AWS accounts by July 19 at 12:00 AM PDT / 12:00 PM PKT.
Meanwhile, most customers do not need to take any action. The system will automatically resolve the fake charges. However, there is a major exception for specific enterprise setups. Some users configure their Cost and Usage Reports with a “Create new report versions” option. These users currently have corrupted, inaccurate reports sitting in their S3 buckets.
Once the backfill completely finishes, Amazon will send a corrected report using a brand-new assemblyId. Therefore, these specific customers must manually update their downstream processes. This includes Athena tables, Redshift pipelines, custom ETLs, or Amazon QuickSight. These processes must reference the new assembly ID. Finally, they should delete the stale, impacted report to prevent processing inaccurate data.
