AWS Cooling Issue in Northern Virginia Exposed Coinbase's Market-Layer Risk

## AWS cooling fault hit Coinbase's market layer, not custody  According to [Cointelegraph](https://cointelegraph.com/news/aws-data-center-overheating-impacting-coinbase-others), AWS said at 5:11 am UTC that it was seeing early signs of recovery after an overheating issue in its Northern Virginia data center disrupted Coinbase and other services. Coinbase said markets would move into "cancel only" mode before trading was re-enabled, and users on web and mobile were seeing degraded performance. FanDuel was also reported as affected. ## What the incident says about the stack The AWS note points to a cooling problem in `use1-az4` inside `US-EAST-1`. That detail matters because it shows the failure happened in the shared infrastructure layer, not in Coinbase's custody system or a crypto-specific code path. A regional utility problem was enough to slow order entry, delay cancellations, and force the venue to protect the book before new trades came back online. ### Cancel-only is a control state, not a recovery signal Coinbase's market-status documentation says cancel-only lets traders cancel open orders but blocks new ones. In other words, it is a containment step: the platform keeps old orders from mixing with a half-recovered system while it figures out which services are healthy. ## Why the distinction matters Coinbase also said customer funds were safe. That is the boundary readers should keep in view. Custody can remain intact while execution is impaired, and the two should not be collapsed into one. For a trading venue, the first symptom of trouble is often not a loss of assets but a break in market access: the book will not accept new orders, status banners lag the user experience, and clients are left waiting for the platform to reconcile state. ## The broader risk is concentration This was not a token-specific failure and it was not a chain outage. It was a cloud concentration event that propagated into the service layer people actually use. The same pattern shows up in other infrastructure incidents: when a single region or availability zone becomes a choke point, the platform can look "up" in a narrow sense while still failing the part of the stack that matters to end users.  ## What would reduce the next hit - multi-region failover that keeps trading available when one zone is impaired - status pages that mirror user-facing order states instead of trailing them - fewer single-zone dependencies for the market layer itself If those three pieces do not improve together, the next outage may look different on the surface and the same underneath: a shared utility problem reaching the order book before anyone can route around it. --- Author: [Alex Chen](https://x.com/AlexC0in) Source: [cointelegraph.com](https://cointelegraph.com/news/aws-data-center-overheating-impacting-coinbase-others)