Coinbase outage halts matching engine

Crypto exchange outages are different from normal app outages. If a food delivery app goes down, dinner is late. If an exchange goes down, people can’t trade, can’t move money, and sometimes can’t even see whether their balances are right. That is why Coinbase’s May 7 outage lands harder than a generic “service disruption” label suggests. The immediate trigger was outside Coinbase — a cooling failure inside an AWS data center in Northern Virginia — but the more important story is what that failure exposed inside Coinbase’s own architecture. ### What actually broke? AWS said a single facility in the US-EAST-1 region, specifically availability zone use1-az4, started suffering impaired EC2 instances and degraded EBS volumes after a thermal event and power loss on May 7. That matters because those are foundational building blocks — compute and storage. When they wobble, a lot of downstream software wobbles with them. Coinbase was one of the visible casualties. (status.coinbase.com) ### Why did Coinbase go down so hard? Coinbase’s own status history shows “Degraded Performance” beginning at 5:56 p.m. PDT on May 7, with trading later re-enabled and customer access restored by 2:28 a.m. PDT on May 8. In plain English, users were dealing with a long evening of broken or unreliable service. Reports tied the impact to core exchange functions — not just a cosmetic front-end issue — which is why this became a matching-engine story instead of a minor outage story. (health.aws.amazon.com) ### What is a matching engine, anyway? The matching engine is the part of an exchange that pairs buyers with sellers and turns orders into trades. It is the market’s gearbox. If that piece stalls, the exchange can still have a website, an app, and status updates — but the actual market function is crippled. That is why traders care less about whether a homepage loads and more about whether orders can enter, match, and settle normally. (status.coinbase.com) The catch is that this part of the stack is often optimized for speed above everything else. ### Why not just add more redundancy? Armstrong’s explanation is basically that Coinbase already designs most systems to survive the loss of one AWS availability zone, but not every exchange component met that bar on May 7. He pointed to a trade-off: centralized exchanges chase ultra-low latency and customer co-location, and full cross-zone insulation can make those systems slower or less attractive to serious traders. That is a real engineering choice — but after an outage like this, it also sounds like a risk acceptance statement. (stocktwits.com) ### Haven’t they seen this movie before? Yes — and that is what makes this more than bad luck. Coinbase published a retrospective after the October 20, 2025 AWS outage saying users saw 3 hours 17 minutes of degraded performance and that the company was reviewing regional deployment strategy and long-term fixes. Seven months later, another AWS-linked incident still knocked out important exchange functions for longer. Different trigger, same dependency theme. (stocktwits.com) ### Was this only a Coinbase problem? No. AWS’s own service health updates and broader reporting show the thermal event hit a single data center in a heavily used part of US-EAST-1 and rippled into many dependent services. But Coinbase stands out because financial platforms have a harsher standard. Users do not experience downtime as inconvenience. They experience it as lost optionality — the inability to act while markets move. (coinbase.com) ### Why does this matter right now? Timing made it worse. The outage hit just as Coinbase was under pressure from weaker results and a fresh round of layoffs. Stocktwits summarized first-quarter revenue at $1.41 billion, down 31% year over year, with a net loss of $394 million. When a company is already telling investors it is tightening up, a long outage turns into a credibility test about what exactly got hardened and what did not. (health.aws.amazon.com) ### Bottom line? The easy version of this story is “AWS went down and Coinbase got unlucky.” The harder, more useful version is that Coinbase still has exchange-critical systems where speed won the argument over fault tolerance. Armstrong says the company will revisit that trade-off. After two major AWS-linked disruptions in less than a year, that review no longer feels optional. (stocktwits.com)