Neon Serverless PostgreSQL

Neon is a fully managed serverless PostgreSQL database platform that separates compute from storage, enabling instant branching, scale-to-zero compute, and usage-based billing. Founded in 2022 and headquartered in San Francisco, Neon targets developers building applications that require elastic database capacity without provisioning overhead. The platform is especially suited for AI engineering workflows, offering an AI Gateway that provides a single API for frontier and open-source models, powered by Databricks. Neon also includes managed authentication with users and sessions stored in Postgres, compute functions without timeouts, and S3-compatible object storage that branches with projects. It is designed for teams that want a Postgres-native experience with modern serverless capabilities, from prototypes to production workloads.

Neon's architecture decouples compute from storage, allowing databases to scale to zero when idle, eliminating costs for unused resources. The platform supports advanced autoscaling up to 56 Compute Units (1 CU = 1 vCPU + 4 GB RAM) on the Scale plan, and claims to prevent 54,210 performance degradations daily via autoscaling. Instant branching creates full database copies in seconds for development, testing, and CI/CD workflows, with unlimited branches on the Free plan and up to 50 simultaneous branches on Scale. Point-in-time recovery (PITR) is available up to 30 days on Scale, with storage pricing at $0.20/GB-month for PITR data. Compute costs dropped 15-25% across all tiers in 2025, and storage pricing fell from $1.75 to $0.30 per GB-month, reflecting Databricks' operational scale after the May 2025 acquisition.

Neon competes directly with Supabase, PlanetScale, AWS Aurora, and Google Cloud SQL. Its key differentiators are separation of compute and storage for scale-to-zero, instant branching, and a generous free tier (100 CU-hours/month and 500 GB data transfer). Compared to Supabase, Neon offers a narrower feature set—it lacks built-in authentication and storage services, requiring additional integrations for those capabilities. Against AWS Aurora, Neon provides a simpler, serverless experience with lower entry costs, but may be more expensive for large, steady-state workloads. PlanetScale offers a similar branching model but uses MySQL, making Neon the choice for Postgres-native teams. Google Cloud SQL provides managed Postgres but without scale-to-zero or branching.

Honest trade-offs: Neon requires additional services for authentication and storage, increasing integration complexity. Customer support issues have been reported by some users, particularly during peak usage. For larger workloads, costs can be higher than competitors like Supabase or self-hosted Postgres due to per-CU-hour pricing and storage overage charges. The Scale plan's compute cost of $0.222/CU-hour is 59% higher than the Launch plan's $0.14/CU-hour, yet both have the same $5/month minimum, making the Launch plan more cost-effective for moderate traffic. Additionally, the free plan's PITR is limited to 6 hours or 1 GB, which may be insufficient for production recovery needs.

Usage-based with three plans: Free ($0/month, 100 CU-hours, 500 GB data transfer), Launch ($5/month minimum, $0.14/CU-hour compute, $0.30/GB-month storage), Scale ($5/month minimum, $0.222/CU-hour compute, $0.30/GB-month storage). Compute costs reduced 15-25% in 2025.