Coinbase said a localized cooling failure inside an AWS data center set off a May 7, 2026 outage that halted nearly all trading across the exchange and took multiple hours to recover. The company’s postmortem says the problem began in a single availability zone in US-East-1, where multiple cooling units failed at the same time and forced thermal shutdowns across affected racks.
The timing matters because Coinbase published the postmortem now, after users were left with a trading platform that was effectively frozen for hours. What looked at first like a cloud incident became a much broader failure once Coinbase traced the outage through its own systems: EC2 instances and EBS volumes went offline, then the exchange’s matching engine, which was tightly coupled to the affected zone, and its messaging infrastructure kept the recovery from moving quickly.
That distinction is the heart of the report. The AWS thermal event started the chain reaction, but Coinbase said its own architectural dependencies significantly prolonged the disruption. In other words, the cloud failure was the spark, but the exchange’s design helped the fire spread. Nearly all trading activity across Coinbase was halted, which is a far more serious outcome than a brief infrastructure hiccup and one that cuts straight to the reliability of a major trading venue.
The postmortem leaves one crucial question unresolved: what Coinbase will change so a failure in a single zone does not cascade through the rest of its trading stack again. The company has now explained how the outage unfolded, and why it lasted for multiple hours, but it has not publicly laid out the specific design changes that would break that zone-level dependency the next time a cooling system fails.
