Why developers leave Supabase: connection pooling has documented edge cases at >5K RPS, vector store via pgvector functional but Pinecone is faster, and BaaS bundling means vendor coupling some teams prefer to avoid. Teams hitting these limits or preferring best-of-breed alternatives evaluate carefully.
Neon is the focused-database alternative. Just Postgres — better connection pooling at scale, scale-to-zero, branching for previews. Pair with Clerk for auth and R2 for storage. Best for teams that prefer best-of-breed services over BaaS bundling.
Firebase remains the alternative for mobile-first teams that found Supabase's mobile SDKs trailing. iOS, Android, Flutter SDK maturity is still ahead. The Firestore data model trade-offs apply.
Convex for TypeScript-only teams that want reactive queries without Supabase's SQL learning curve. Type inference flows from schema to React. Trade-off: TypeScript-only, proprietary execution model.
Three options: (1) PgBouncer in front of your Supabase Postgres, (2) split read traffic to Neon read replicas, (3) migrate the high-RPS service entirely. Most reviewers reported PgBouncer was the lowest-cost fix; Neon split for >10K RPS workloads.
Yes — Supabase Auth is consumable as a standalone service. Some teams use Supabase Auth + Storage but move the primary database to Neon for connection-pool reasons. The integration cost is real but workable.
For compliance-heavy teams: yes. Functionality at parity with cloud. The ops investment is real (Docker Compose or Helm + database management). For teams hitting cloud limits the self-host path enables full control over pooling and scale.
