Coinbase Suffers Hours-Long Outage: Single AWS Availability Zone Failure Disrupts Trading

On May 7, 2026, cryptocurrency exchange Coinbase experienced a large-scale outage that disrupted trading services for more than five hours, forcing the platform to temporarily enter “cancel-only” mode, during which users were unable to trade through either the web or mobile applications.

The incident came at a difficult time for Coinbase, as the company was already facing weaker-than-expected quarterly earnings, a declining stock price, and a 14% workforce reduction, further intensifying industry concerns about its infrastructure resilience and operational stability.

AWS Multi-Availability Zone Failure Triggered Trading Interruption

According to an official Coinbase statement, the outage was caused by a “thermal event” in the AWS US-EAST-1 region in Northern Virginia, which led to hardware failures across multiple availability zones.

Several Coinbase core components were affected, including:

• FIX order gateways
• Trading matching engines
• Amazon Managed Streaming for Apache Kafka (AWS MSK)

Rob Witoff, Head of Platform at Coinbase, explained:

“Our systems were designed to tolerate single-region failures, but this event involved failures across multiple AWS availability zones, resulting in prolonged disruption of core trading services.”

To minimize latency, Coinbase had deployed its matching engine within a single availability zone, which amplified the impact of the outage and exposed single-point-of-failure risks.

Although Coinbase maintained replication across multiple regions, vulnerabilities within the MSK cluster prevented automatic failover, forcing engineers to manually intervene to restore services.

During the outage, Coinbase temporarily switched the platform into “cancel-only” mode to prevent abnormal trading activity before gradually restoring market operations.

The company later stated on X:

“The primary issue has now been fully resolved. We appreciate our users’ patience as we continue investigating the incident alongside AWS.”

Industry Comparison: The Importance of Multi-Region Redundancy

The incident highlighted the limitations of centralized exchange architectures when facing infrastructure-level failures.

In contrast, some fintech companies have already adopted multi-cloud and multi-region redundancy strategies:

• UK digital bank Monzo can switch to a lightweight banking service running on Google Cloud Platform (GCP) when AWS services fail.
• Payment platform Dojo simultaneously operates across two Google Cloud regions and one AWS region, allowing all three regions to process traffic in parallel.

Coinbase CEO Brian Armstrong acknowledged the issue on May 8:

“Our centralized exchange architecture did not fully withstand availability zone failures. In light of this incident, we will reevaluate these trade-offs to ensure the best possible trading environment for our users.”

Technical Lessons: The Limits of Kafka and Managed Clusters

Coinbase relies heavily on AWS MSK to build its distributed event streaming architecture, enabling ultra-low-latency transmission and processing of terabytes of trading data.

Although the system had operated reliably for years, the simultaneous failure of both the matching engine infrastructure and the MSK cluster during extreme hardware failures prevented automatic recovery.

Witoff emphasized that even with multi-region replication, such extreme scenarios can still occur, exposing the limitations of managed clusters under severe infrastructure events.

The failure spread along two major paths:

• Multiple underlying hardware components supporting the matching engine failed simultaneously.
• The MSK cluster could not maintain availability and required partition migration to new broker nodes.

Coinbase has committed to publishing a detailed post-incident analysis report to provide operational lessons for the broader industry.

Cloud Disaster Recovery Lessons for the Industry

This outage serves as a critical reminder for the fintech industry: while low latency and performance optimization are important, centralized architectures can significantly amplify single availability zone failure risks.

Multi-region and multi-cloud deployment strategies, combined with extreme failover testing and disaster recovery drills, are becoming essential for ensuring platform resilience and business continuity.

The Coinbase incident provides a vivid real-world example of why modern cloud disaster recovery strategies must evolve beyond traditional assumptions and prepare for increasingly complex infrastructure failure scenarios.