Stateful serverless applications: The new architecture

On October 30, 2026, Google Cloud announced stateful processing for BigQuery continuous queries in preview. This follows earlier 2026 updates where AWS Lambda extended timeouts to 90 minutes and introduced MicroVMs for isolated execution. The message from major cloud providers is clear. Building stateful serverless applications is no longer an anti-pattern.

We are officially moving beyond the traditional trigger-and-die model. Developers previously had to rely on external databases for every piece of persistent data, adding latency and complexity. Now, the platforms themselves are absorbing the burden of maintaining state across long-running operations.

According to amazon.com, AWS Fargate now supports native Amazon Elastic Block Store (EBS) volume integration. This provides persistent block storage for containerized applications without requiring you to manage the underlying servers. AWS Step Functions also coordinate state across stateless functions, with standard workflows capable of running for up to a full year.

Over at Microsoft, Azure Durable Functions handle state, checkpoints, and retries for long-running processes. Microsoft also designed Azure Container Apps for complex workflows, offering support for volume mounts and zone-redundant storage. When writing the configurations for these orchestration services, run your files through a JSON Validator to catch syntax errors before deployment.

Google Cloud Run is typically stateless, but developers can now configure volume mounts to connect to Cloud Storage buckets or NFS shares. The platform is also testing Google Cloud Run Instances in preview. These offer long-lived, dedicated containers that intentionally do not scale to zero.

They solve the need for continuous background processing, which is increasingly necessary for AI agents software development and persistent daemons. Additionally, Google Cloud Workflows orchestrate HTTP-based services while checkpointing state directly to Cloud Spanner for reliability.

You no longer have to provision permanent virtual machines just because a workload needs persistent storage. The operational overhead of managing servers is finally dropping for data-intensive applications. Teams can keep the scaling benefits of serverless while running workflows that take hours or days to complete.

Start by identifying long-running background jobs currently running on legacy infrastructure. Test them on Azure Container Apps or Google Cloud Run Instances to see how they perform. If you use infrastructure-as-code to manage these new deployments, a JSON to YAML converter helps keep your template files readable.