AWS US-EAST-1 Thermal Event Forces Coinbase Into Manual Disaster Recovery as Cooling Failure Damages EC2 Hardware

Overview

A cooling system failure inside a single Amazon Web Services data center hall in Northern Virginia triggered a cascading outage on May 7–8, 2026, that physically damaged EC2 instances and EBS storage volumes, disrupted dependent cloud services across the region, and brought down core trading services at Coinbase for hours. Cooling capacity was not restored to pre-incident levels until approximately 1:50 PM PDT on Friday, May 8 — more than 20 hours after the first alerts appeared, according to Network World.

What We Know

The thermal event and its timeline

At around 5:25 PM PDT on Thursday, May 7, AWS posted its first update on the health dashboard, confirming that “EC2 instances and EBS volumes hosted on impacted hardware are affected by the loss of power during the thermal event,” as The Register reported. The affected zone was use1-az4, a single availability zone within the US-EAST-1 region.

AWS stated that “servers automatically shut down when the temperatures exceeded the operating thresholds in order to protect the hardware,” according to Network World. By 6:47 PM PDT the company acknowledged that it was continuing to work to bring temperatures back to normal levels, and at 8:06 PM PDT it admitted that progress was “slower than originally anticipated,” Network World reported. Additional cooling capacity came online near 10:11 PM PDT, allowing some racks to begin recovering, The Register noted.

Full cooling capacity was restored at 1:50 PM PDT on May 8, ending an incident that lasted more than 20 hours. AWS shifted traffic away from the impacted zone as part of mitigation and warned customers of “longer than usual provisioning times,” The Register reported.

Services that showed early recovery improvements included IoT Core, Elastic Load Balancing, NAT Gateway, and Redshift, Network World reported.

Coinbase: backup systems fail during a multi-zone cascade

Coinbase confirmed that the disruption stemmed from “increased temperatures in the affected AWS service” and that “Coinbase systems are designed to be resilient to a single zone outage. In this case, we observed failures impacting multiple AWS zones, which caused an extended outage of core trading services,” as CoinDesk reported. Users were unable to trade, check balances, or process withdrawals during the outage, and markets briefly moved to “cancel only” mode before full restoration, CoinDesk reported.

Rob Witoff, Coinbase’s Head of Platform, described the outage as a multi-hour service disruption affecting trading, exchange access, and balance updates, per Yahoo Finance. Backup systems designed to isolate failures did not work as expected during the incident, Yahoo Finance reported.

Coinbase CEO Brian Armstrong stated that “we experienced an outage at Coinbase last night, which is never acceptable,” adding that “exchanges have unique architectures that optimize for latency and co-location of clients” and that “at a minimum, the duration of an outage should be able to be reduced considerably when an AZ move is needed,” according to Yahoo Finance. Armstrong said the company would revisit those architectural trade-offs following the incident.

What We Don’t Know

AWS has not published a formal post-incident review specifying exactly how many cooling units failed, what redundancy levels were in place in the affected data center hall, or what remediation steps will prevent a recurrence. The precise mechanism by which a single-zone thermal event cascaded into failures across multiple AWS zones — as Coinbase described — has not been explained publicly.

It is unclear whether the physical damage to EC2 instances and EBS volumes on affected racks resulted in permanent data loss for any customers, or whether all data was recoverable once hardware returned to operational temperature. AWS has not disclosed the total number of services or customer accounts impacted.

Analysis

The May 7 incident adds to a pattern at US-EAST-1, which has hosted major outages in December 2021 and October 2025, according to The Register. AWS executives have maintained that the region is not inherently more fragile than others but handles a larger operational scale, The Register reported.

The Coinbase disclosure is the most revealing engineering detail from this incident. The exchange’s own statement that its systems “are designed to be resilient to a single zone outage” but observed “failures impacting multiple AWS zones” raises questions about how a thermal event confined to use1-az4 cascaded beyond that zone. Armstrong’s acknowledgment that exchange architectures optimized for latency cannot be fully insulated from AZ failures without introducing latency trade-offs illustrates a structural tension that high-frequency financial platforms face on shared cloud infrastructure.

Bhuvie Chhabra, senior principal analyst at Gartner, highlighted an underlying question that this incident renewed: customers should evaluate “to what degree AZs are located in physically distinct facilities versus coexisting within the same physical data center,” Network World reported. Chhabra also noted that “reducing the concentration risk to zero is unattainable,” framing AZ isolation as a matter of degree rather than absolute protection, Network World reported.