- A cooling-system failure at an AWS data facility triggered server shutdowns that disrupted Coinbase trading, deposits, and withdrawals.
- Recovery was delayed by matching engine failover problems and a Kafka service issue that affected critical backend systems.
- Coinbase has introduced new redundancy, failover testing, and infrastructure upgrades to improve resilience against future outages.
A major infrastructure failure inside Amazon Web Services disrupted Coinbase trading, deposits, and withdrawals for hours on May 7, 2026, according to the exchange’s newly released postmortem. The outage affected retail and institutional users after cooling systems failed at an AWS data facility, while recovery stretched well beyond the initial disruption due to additional system failures inside Coinbase’s infrastructure.
AWS Failure Knocked Core Systems Offline
According to Coinbase, the incident began at 7:20 p.m. ET on May 7 when multiple chiller units failed simultaneously inside an AWS data hall in the us-east-1 region. The cooling loss triggered thermal shutdowns that forced affected servers and storage systems offline.
Soon after, Coinbase users experienced widespread service interruptions. Trading halted at 7:48 p.m. ET, while buying, selling, deposits, withdrawals, and transfers became unavailable across several products.
The exchange stated that AWS availability zones are designed to isolate failures. However, Coinbase discovered weaknesses in its own recovery systems when the disruption spread beyond a single facility.
Recovery Delayed by Two Separate Failures
As engineers worked to restore services, a second issue complicated recovery efforts. Coinbase said its matching engine lost quorum after AWS terminated servers within a cluster placement group.
The company lacked automated cross-zone failover for that system. Consequently, engineers deployed emergency code changes and rebuilt infrastructure to restore operations.
However, another obstacle emerged. According to Coinbase, AWS’s Managed Kafka Service failed to automatically reassign partition leaders after the outage.
That issue disrupted event-streaming services, affecting quoting systems, fee calculations, payments infrastructure, and parts of the ledger pipeline. Engineers ultimately performed manual partition reassignments alongside AWS teams to restore functionality.
Trading resumed in stages. Coinbase restored cancel-only functionality at 2:25 a.m. ET on May 8 before reopening markets fully at 3:49 a.m. ET.
Coinbase Details Infrastructure Upgrades
Following the incident, Coinbase outlined several infrastructure changes aimed at improving resilience.
The company plans to strengthen matching engine redundancy through warm cross-zone standby systems. Additionally, it will conduct regular failover exercises across production environments.
Coinbase also plans to expand one Kafka deployment from two availability zones to three. Furthermore, the exchange is developing new tools, tests, and operational procedures to address similar failure scenarios.
According to Coinbase, the company has already notified relevant regulators and continues completing formal impact assessments related to the outage.
