AWS outage in Virginia disrupts CME Group and Coinbase services

A rise in temperature inside a single Amazon Web Services data center in Northern Virginia was enough to jolt two of the most closely watched financial platforms in the country, interrupting trading on CME Group and Coinbase and exposing how much digital market infrastructure is concentrated in one cloud region.

AWS said the problem began Thursday when temperatures climbed inside one data center and the company brought additional cooling online while shifting traffic away from the affected Availability Zone. The disruption hit one zone in the US-EAST-1 region, identified in reporting as use1-az4, a reminder that even a region built with multiple zones can still be vulnerable when too much critical activity depends on the same cluster of infrastructure in Virginia.

The impact moved quickly into financial markets. CME Group, the world’s largest derivatives marketplace, reported technical and latency issues on its platforms. Coinbase said its own platform was having performance problems tied directly to the AWS outage. The sequence was stark: a cooling problem in one building in Northern Virginia cascaded into trading friction for firms that sit at the center of global finance and crypto markets.

AWS said it had begun seeing early signs of recovery, then later said restoring the remaining affected systems was taking longer than expected and that there was no timeline for full recovery. By Friday, AWS services were largely back online, and Amazon said a full recovery would take several hours. Coinbase later said all markets were re-enabled for trading. CME said users could log back into its CME Direct platform after essential maintenance work.

The outage also highlighted how cloud concentration shapes risk across the broader economy. AWS says each region has at least three Availability Zones, and AWS Health is its authoritative source for service events affecting customer resources. Yet the event showed that formal redundancy does not erase the consequences of a localized failure when exchanges, apps and websites are built on the same infrastructure backbone.

CME’s public Global Command Center alert archive and client-support channels underscore how market operators now manage disruptions through tightly controlled incident systems, but those safeguards still depend on the resilience of the cloud beneath them. The episode added to a growing concern in the industry: overheating in data centers is becoming a more common threat as cloud and AI servers draw massive power and generate intense heat.