CoreWeave has made CoreWeave Database generally available, offering managed PostgreSQL inside customers’ CoreWeave Kubernetes Service (CKS) clusters. Announced during KubeCon + CloudNativeCon North America 2026 week, the service is designed to place database workloads alongside the AI training and inference applications that use their data.
The service runs on CloudNativePG (CNPG), an open-source Kubernetes operator for PostgreSQL that the Cloud Native Computing Foundation moved to incubation during the same week. CoreWeave combines CNPG’s database management capabilities with its own console and Kubernetes workflows, backups, monitoring and ongoing operations.
Database workloads within the cluster
CoreWeave says customers previously had to operate PostgreSQL themselves or host it in another cloud. Keeping a database elsewhere can add a network hop, an egress path and a separate environment to secure, according to the company. With CoreWeave Database, database instances run on CPU nodes in the customer’s CKS cluster and connect to workloads through an in-cluster Kubernetes service.
The placement is intended to serve applications that depend on persistent state as well as GPU-intensive computation. CoreWeave points to application state, metadata and embeddings as examples of data that AI agents and training pipelines may access during operation. The company says hosting the database in the same cluster lets it manage database operations for customers, rather than requiring teams to run PostgreSQL themselves.
Each database has a primary instance and read replicas distributed across hosts. The service provides automatic failover, while live data is stored on CoreWeave Distributed File Storage. The announcement does not specify database sizing limits or performance benchmarks.
Backups, access and vector search
Continuous backups are enabled by default. Customers can also use point-in-time recovery, schedule backups, initiate them on demand, and restore a full backup to a new cluster for validation. These options provide several recovery workflows, including testing a restore separately from the existing database.
CoreWeave says databases remain private to their cluster and are encrypted both at rest and in transit. The company manages TLS certificates and security patches. Customers manage access using PostgreSQL roles, Kubernetes network policies and audit logging.
The service supports pgvector, a PostgreSQL extension used to store embeddings and perform vector search. That allows teams to keep vector data and application data in one database, including for retrieval-augmented generation workflows. CoreWeave did not provide further details about pgvector configuration or service-specific limits.
CNPG manages primary and replica roles and failover. CoreWeave adds the surrounding managed-service functions, including backups, monitoring and operational workflows. The division of responsibilities means the service relies on an open-source Kubernetes operator for core PostgreSQL orchestration while CoreWeave supplies the service interface and operations.
PostgreSQL compatibility and availability
CoreWeave describes the service as fully PostgreSQL compatible and says existing databases can be migrated using native PostgreSQL tools. The announcement did not identify particular migration tools, supported PostgreSQL versions or a migration service, so customers evaluating a move will need to confirm compatibility and migration requirements for their deployments.
The company also cited contributions to the surrounding open-source ecosystem. These include S3 compatibility improvements to the Barman backup tool and connection configuration work in the pgexporter Prometheus exporter. CoreWeave Database is generally available now; the company did not disclose pricing.
For organizations already running workloads on CKS, the offering provides a managed option for placing PostgreSQL in the same Kubernetes environment as applications and AI pipelines. Its central proposition is operational and architectural: keep database-dependent state close to the compute using it, while relying on CoreWeave and CNPG for database management, failover and recovery capabilities.
