A billing error in AWS resulted in users receiving estimated charges that reached as high as $2.5 trillion.

      TL;DR: An AWS billing error sent users estimated bills in the billions and trillions. Amazon confirmed a pricing mistake and is recalculating all estimates.

      AWS users awoke on Friday to find estimated billing emails displaying charges from hundreds of millions to $2.5 trillion. Amazon confirmed that a unit pricing error in the estimated billing computation subsystem of the AWS Billing Console was responsible. One Reddit user, who had charges of $0.19 last month, received an estimated bill close to $2.5 billion. Others reported monthly estimated charges ranging between $126,000 and $2.5 trillion.

      “The billing estimates shown do not reflect actual usage and charges,” AWS stated. The company identified the root cause within 90 minutes, describing it as “a unit pricing issue within the estimated billing computation subsystem.” AWS halted updates to the estimated bills to prevent further increases in the inflated figures and stated that it is now recomputing all billing data. Corrected amounts should be available by Saturday, July 18, at noon Pacific time.

      The error caused alarm before users realized it was incorrect. “I nearly had a heart attack when my two S3 buckets with a few MBs of data resulted in a forecast of half a billion dollars,” one Reddit user shared. Another commented, “Needless to say, I panicked and deleted everything on this account.” Last week, AWS allocated $1 billion for forward-deployed AI engineers, and the billing bug serves as an unwelcome reminder that the infrastructure supporting AI workloads worth billions monthly is also susceptible to generating billing errors of a similar scale.

      This incident follows a separate AWS CloudFront outage that resulted in error pages instead of websites. As AI demand drives unprecedented spending, the cloud infrastructure is managing monthly compute bills that can reach hundreds of millions for individual customers, making the reliability of billing systems a more critical issue than when the largest bills were in the thousands.