Coinbase has released a postmortem on the service disruption that shut down trading and interrupted deposits, withdrawals, and transfers for roughly eight hours on May 7, 2026. The exchange said the incident began with a cooling-system failure at an Amazon Web Services data facility, then lasted longer because separate problems surfaced inside Coinbase’s own recovery path.
According to the company, the outage started at 7:20 p.m. ET when multiple chiller units failed at an AWS data hall in the us-east-1 region. The loss of cooling triggered thermal shutdowns, taking affected servers and storage systems offline. By 7:48 p.m. ET, Coinbase trading had halted, while buying, selling, deposits, withdrawals, and transfers were unavailable across several products.
Matching engine and Kafka issues slowed the restart
Coinbase said AWS availability zones are built to contain failures, but the event exposed weaknesses in its own resilience design once the disruption spread beyond a single facility. During recovery, the exchange found that its matching engine lost quorum after AWS terminated servers within a cluster placement group.
The company said that system did not have automated cross-zone failover. Engineers had to push emergency code changes and rebuild infrastructure before operations could come back. A second problem then appeared: AWS Managed Kafka Service did not automatically reassign partition leaders after the outage. That failure disrupted event-streaming services tied to quote systems, fee calculations, payments infrastructure, and parts of the ledger pipeline.
Markets reopened in stages early on May 8
Coinbase said engineers worked with AWS teams to manually reassign Kafka partitions and restore functionality. Trading returned in phases. The exchange first enabled cancel-only mode at 2:25 a.m. ET on May 8, then fully reopened markets at 3:49 a.m. ET.
After the incident, Coinbase said it is adding warm cross-zone standby capacity for the matching engine and will run regular failover exercises in production environments. The company also plans to expand one Kafka deployment from two availability zones to three, while adding new tools, tests, and operating procedures for similar failure scenarios. Coinbase said it has notified relevant regulators and is still completing formal impact assessments tied to the outage.