Zircuit (@ZircuitL2): The AWS Outage Hit 6 L2s. Zircuit wasn't one of them.

The AWS Outage Hit 6 L2s. Zircuit wasn't one of them. 6 L2s went down when AWS did earlier this week. Zircuit was not one of them. Here’s why. The short answer is: we got lucky. But the real answer is more nuanced. L2s, many of which rely on centralized infrastructures like single sequencers and a few RPC providers, were hit particularly hard. Our infrastructure team, which has the battle scars, discipline, and foresight that come from years of running production systems on AWS, was particularly well equipped to anticipate these points of failure. We made some deliberate architectural choices that limited our exposure to us-east-1, historically the first and worst AWS region. Even with six AWS availability centers in Northern Virginia and Maryland, and another planned to go online in 2026, us-east-1 sucks. It is known. We intentionally deploy our testnet in us-east-2 (Ohio, USA), because us-east-1 hosts a huge portion of AWS’s global backbone - Route53, IAM, S3 control plane, and DynamoDB metadata. When this region stumbles, half the internet goes along with it. We anticipated an incident there roughly once a year, so we avoided it from day one. The us-east-2 region gives us geographic separation from Amazon’s busiest corridor while still keeping latency reasonable for North America. Ohio geographically sits in the Midwest and can reasonably serve both the East and West Coast fairly well. For our main sequencing infrastructure, we went even further. We avoided North America altogether. Our primary sequencer runs in eu-north-1 (Stockholm) with a warm failover site in ap-northeast-2 (Seoul). During the outage, we actually ran a failover test and restarted a replica in Seoul, verifying that we could promote the replica to the primary sequencer if Stockholm ever went dark. That was by design, not improvisation. Were we perfect? Nope. We still had some dependency risks, some of which degraded - but did not fully shut down - Sequencer Level Security during the event. But even with that exposure, transactions continued to flow. That’s the difference between redundancy and resilience: redundancy gives you options, resilience keeps you alive when the options get ugly. Our job as an infrastructure team is to assume failure and keep the system moving anyway. However, our architecture doesn't stop at AWS. Under Martin’s (@dr_zircuit) leadership, Zircuit also made the investment (in yet again, a different geographical region) of bare-metal infrastructure in Q1 of 2025. This was not a vanity project, and more than just an insurance policy. The math is clear: we can run our ZK Prover infrastructure (GPU-to-Proof) cheaper in-house on our own machines than on any of the major cloud providers. Insurance is a plus - a very relevant one in light of recent events. Because our stack is fully cloud-native and declarative, we can bootstrap and operate the sequencer on physical hardware if AWS ever suffers a global or jurisdictional outage. We’ve already validated the workflow for our ZK-prover infrastructure; this part isn’t theoretical. Extending the same approach to the sequencer is the next logical step. The goal isn’t to run everything on bare metal all the time, but rather when it makes sense. If the hyperscalers hiccup, we can redeploy and operate independently. That’s what sovereignty really means in practice. Looking ahead, it’s clear that the end game for resilient infrastructure is to achieve deterministic redeployment anywhere - the ability to stand up our entire system from code alone, on any common infrastructure. Whether it’s bare metal or cloud-hosted, the objective is to treat all of it as a flexible substrate that can be reconstituted on demand. When it comes to architectural design, conscientious design is oftentimes invisible to the end user. But when it truly matters, the lack of conscientious design is very much visible. We have always built with security and resilience in mind; that’s how Zircuit and Sequencer Level Security came to be in the first place. Zircuit’s continued operations during the AWS outage are a testament to our overarching architectural design ethos. For our users and developers, thank you for trusting us to be your L2 of choice. We’ll continue to keep you safe, so you can focus on what matters: building and innovating.