AWS customers worldwide woke up to astronomical bills on July 17, 2026, as a billing system glitch displayed estimated charges ranging from $1.7 billion to over $190 trillion for accounts that normally cost pennies to operate. The bug, which affected the AWS Billing Console and Cost Explorer, sent system administrators into a panic and sparked a massive discussion on Hacker News that accumulated over 1,100 points and 660 comments within hours.
The issue was so widespread that even accounts with minimal or no usage reported seeing mind-boggling invoices. One user on Hacker News reported their normal bill of around $0.19 had been replaced with a projected charge of $2.5 billion. Another shared a screenshot showing a balance of $87,967,679,887,258.36. “Help, what is this number?” they asked. The response: “That is 87 trillion, 967 billion, 679 million, and so on.” Users joked about depleted rainy day funds and having to take out mortgages to pay their cloud bills.
AWS acknowledged the incident on its Health Dashboard, listing it under “Inaccurate Estimated Billing Data.” The company confirmed that the glitch affected estimated billing data displayed in the console and Cost Explorer, though it noted that actual invoices would not reflect these inflated numbers. The root cause was linked to S3 billing estimates, and AWS engineers worked through the night to resolve the issue. Many affected customers reported that corrected billing data began appearing within 12 to 18 hours.
What Actually Happened
The glitch appeared to stem from an error in how AWS calculates estimated billing data. Users reported seeing charges at levels that were clearly impossible. One customer who had set a $7.50 budget received an alert saying their actual cost had exceeded $3 billion. Another saw a bill of $155 billion and growing, expressing concern that they might “deplete their rainy day fund.”
The Hacker News thread became an impromptu support group as users shared increasingly absurd screenshots. Numbers quoted included $190 trillion, $87 trillion, and $35 trillion for RDS services. The widespread nature of the glitch, affecting accounts of all sizes across multiple services, suggested a systemic issue rather than a problem isolated to a specific customer or region. Users debated whether the error was a bug or a new “vibe billing” feature, a reference to the recent trend of AI-powered pricing models.
Some users noted that the absurdity of the numbers ironically helped. “It is a good job it was off by such a large amount,” one wrote, “or I might have panicked instead of writing it off as a phishing attempt.” Others reported receiving the alerts and immediately locking down their accounts, deleting old resources, or canceling their AWS accounts entirely in response to the shock.
The Root Cause: A Simple Unit Error
A former AWS employee familiar with the billing system provided one of the most insightful explanations of the glitch. “It is a unit error,” they wrote on Hacker News. “In my case we meant to charge like 5 cents per GB, but missed the unit (GB), and then the billing system defaults to bytes. 5 cents per byte of data transferred meant some customers were seeing multimillion-dollar bills within hours.”
The explanation reveals a fundamental fragility in how cloud billing systems operate. AWS services emit metering values that are not directly tied to prices. Each SKU or line item is defined in a pricing plan with a unit type, region, and price per unit. When metering records are joined to a pricing plan based on account ID, region, and SKU, a mismatch in the unit type causes the metering data conversion to fail. The result: astronomical numbers that have no basis in reality.
Current and former AWS employees confirmed that the process for correcting such errors is well established. Corrections and amendments are issued, apology emails go out, and the system is patched. But the incident raises uncomfortable questions about how such a basic error could slip through.
Why No One Caught It
The deeper story here is not about a unit conversion bug. It is about the systemic gaps in testing and monitoring at even the largest technology companies. As one Hacker News commenter with experience at multiple big tech firms explained, there are typically separate testing teams for the metering service and the billing service. “Test 1 will verify that the new system emits billing entries in some expected way,” they wrote. “Test 2 will be in the billing system. But they will not test the two things together because it will be harder to do and the teams will have different management chains.”
This disconnect between teams is a classic organizational failure. Unit tests pass in isolation. Integration tests do not exist because the cross-team coordination required is expensive. And end-to-end testing with real billing data is often blocked by legal and accounting concerns. “Somebody will have said we should actually have the tests charge money and somebody else will have said well we cannot have the tests actually charge money, that is a legal accounting problem, it might even be a crime and then nobody would have asked what the next best thing was,” the commenter continued.
Even basic anomaly detection should have caught this. A single account whose estimated bill jumps by 10 million percent should trigger an automated review before it reaches the customer dashboard. The fact that it did not suggests that AWS, despite its reputation for operational excellence, has significant blind spots in how it validates billing data before displaying it to customers.
What This Means for Cloud Customers
The AWS billing glitch is a reminder that even the most reliable cloud providers can have catastrophic failures in their billing systems. For startups and solo founders who rely on AWS, the incident highlights the importance of setting budget alerts, using billing alarms, and maintaining a paper trail of normal usage patterns. It also underscores the value of diversifying cloud providers to avoid single points of failure, not just in infrastructure but in operations and billing.
AWS has since resolved the inaccurate billing estimates and issued credits to affected accounts where appropriate. But the psychological damage may take longer to heal. As one user put it: “Who is going to compensate us for the years taken off our lives when we received the alerts?” For an industry built on trust in infrastructure, incidents like this erode confidence in ways that no credit can fully restore.