Serverless database comparison posts in 2026 are mostly written by people who have used one provider and read the marketing pages of the other four. This is the version after running production workloads on Supabase, Neon, and a handful of client builds across PlanetScale, Turso, and Convex over the last 18 months. Five providers, real production data, no affiliate links.
I run Supabase as the primary on this site, on HostList (the 91,000-page directory), on the WordPress Stack Advisor, and on most client work in the last year. The HIPAA cluster on this site is largely Supabase-built — the $700/month Vercel-plus-Supabase healthcare setup is the architecture I default to for any healthcare project that does not specifically need a different vendor. The honest take below covers where Supabase wins, where the alternatives genuinely beat it, and where the choice is closer than the marketing implies.
The five providers in 60 seconds
- Supabase — Postgres-as-a-service with built-in Auth, Storage, Realtime, Edge Functions, and pgvector. Pro $25/month, Team $599/month, HIPAA add-on $350/month on Team or Enterprise. The most batteries-included of the five.
- Neon — Postgres-as-a-service, branching-first (every preview deploy gets a database branch), serverless compute scale-to-zero. Free tier generous, Launch $19/month, Scale $69/month.
- PlanetScale — MySQL-as-a-service, vitess-based for sharding, branching workflows. Removed the free tier in 2024. Scaler $39/month, Pro $79+/month, Scaler Pro for production at higher tiers.
- Turso — distributed SQLite (libSQL), edge-first replication, free tier 9GB and 1B row reads, Scaler $29/month. The fastest at the edge for read-heavy workloads.
- Convex — TypeScript-first reactive database with built-in functions, real-time queries, no SQL. Free 1GB, Pro $25/seat, Team and Enterprise tiers. The most opinionated of the five.
Where each provider actually wins
Supabase: app + content + auth in one Postgres
Supabase is the right call when your app and content share a database and you want auth, storage, and real-time included without integrating five separate vendors. The Postgres underneath is real Postgres — pgvector for AI features, full SQL, RLS for multi-tenant security. The HIPAA add-on at $350/month on Team is the cleanest healthcare path for a Postgres-shaped product. The downside is that the Pro plan ($25/month) is genuinely limited for production — Team at $599/month is where serious workloads live, and the price step is real.
- Wins on: full-stack apps, healthcare with HIPAA, projects needing pgvector for AI, teams wanting one vendor for DB+auth+storage.
- Falls short on: heavy multi-region writes (read replicas help but Postgres single-master is the limit), pure database-only briefs where the bundled features are noise.
Neon: Postgres branching for preview deploys
Neon's killer feature is branching — every Vercel preview deploy gets its own ephemeral Postgres branch with the full data, no fixtures. For teams that ship multiple PRs per week with database-touching changes, this single feature pays for the migration. The serverless scale-to-zero is genuinely useful for staging environments and low-traffic apps. The downside: Neon is a database-only service, so you bring your own auth, storage, real-time. Right call when you specifically want Postgres without the bundle.
- Wins on: development workflow with branching, Postgres-only projects, teams already using Auth0/Clerk/Cognito for auth.
- Falls short on: lack of integrated auth/storage means you stitch services, no HIPAA on Launch tier (Scale-and-up only).
PlanetScale: MySQL at scale with branching
PlanetScale was the pioneer of database branching but removed their free tier in 2024, which materially hurt indie developer adoption. Right call now for teams already on MySQL at scale who want branching plus Vitess sharding without running it themselves. Strong DX, mature platform. Wrong call for projects that would otherwise pick Postgres — switching from Postgres to MySQL for the branching feature alone is rarely worth it now that Neon ships the same workflow on Postgres.
- Wins on: existing MySQL workloads at scale, teams that need Vitess sharding.
- Falls short on: indie/free-tier developers (no free tier since 2024), Postgres-curious teams who would benefit more from Neon.
Turso: edge-first SQLite for read-heavy workloads
Turso (libSQL, the SQLite fork) replicates your database to the edge globally and serves reads from whichever region the user hits. For read-heavy applications — content sites with database-driven content, e-commerce catalogs, directories — the latency win is dramatic. The 9GB free tier and 1 billion row reads is genuinely generous. The catch: SQLite has different transaction semantics than Postgres, write-heavy applications are not the optimal fit, and the ecosystem of ORMs and tools is smaller than for Postgres or MySQL.
- Wins on: edge-distributed read-heavy apps, content sites with database-driven content at scale, generous free tier.
- Falls short on: write-heavy workloads, complex multi-table transactions, projects already invested in Postgres ecosystem tools.
Convex: TypeScript-first reactive database
Convex is the most opinionated of the five — TypeScript schema definitions, server functions in TypeScript, real-time queries by default, no SQL. For teams that want to stay in TypeScript end-to-end and value the developer experience over flexibility, Convex is genuinely productive. The trade-off is the lock-in: there is no SQL escape hatch, the data model is Convex-specific, and migrating off Convex is a full rebuild rather than a database export.
- Wins on: TypeScript-first teams, real-time-heavy apps, prototypes-to-production where DX speed matters.
- Falls short on: anyone needing SQL flexibility, projects with strong data-portability requirements, complex analytical queries.
Decision tree — pick by the brief
Full-stack app with content + auth + real-time + AI features
Supabase. The bundled features (Auth, Storage, Realtime, pgvector) genuinely save vendor management overhead. The HIPAA-compliant Supabase + Vercel setup covers the production-grade version including the BAA story.
Postgres-only project with serious development workflow needs
Neon. Branching per preview deploy is the killer feature. Pair with Auth0, Clerk, or Supabase Auth (yes, you can use Supabase Auth without using their database) for the auth layer.
MySQL workload at scale with sharding requirements
PlanetScale. The Vitess sharding plus branching combo is unique. If you are starting fresh, evaluate Neon first; if you are already on MySQL at scale and want the platform without running it yourself, PlanetScale earns the price.
Read-heavy edge-distributed content app
Turso. SQLite at the edge with 1B free reads is genuinely a different shape. Right for content directories, e-commerce catalogs, programmatic SEO sites where read latency dominates the user experience.
TypeScript-only team, prototype-shaped product, real-time first
Convex. The DX speed for TypeScript teams is real. Accept the lock-in trade-off; revisit if the product is going to scale into something that requires SQL or data portability.
Cost economics — annual TCO for a typical workload
Anchored to a hypothetical SaaS app: 10K active users, 5GB database, 100K API requests per day, 5-engineer team, weekly deploys.
- Supabase Team: $599/month base. Plus HIPAA add-on $350/month if applicable. ~$11,400/year (or $7,200 without HIPAA).
- Neon Scale: $69/month + usage-based compute and storage, typically $40-100/month additional. ~$1,500-2,500/year.
- PlanetScale Scaler Pro: $79/month + usage. ~$1,500-2,500/year for similar workload.
- Turso Scaler: $29/month plus usage. ~$500-800/year.
- Convex Pro: $25/seat × 5 + usage = ~$2,000-4,000/year.
Supabase looks expensive on paper at this scale, but the comparison is unfair — Neon, PlanetScale, Turso, and Convex are database-only services. Adding equivalent auth (Clerk Pro $25/month + $0.02/MAU), storage (Cloudflare R2 ~$15/month), real-time (Pusher $49/month or self-host), and pgvector (managed via OpenAI embeddings + a vector DB) typically lands at $200-500/month additional. Once you bundle the integrations, Supabase Team often becomes cost-competitive.
FAQ
Is Supabase better than Neon?
For full-stack apps with auth, storage, and real-time requirements, yes — Supabase bundles features Neon does not. For Postgres-only projects with serious development workflow needs, Neon's branching is the differentiator. The two are optimised for different briefs; the choice is rarely about which is 'better' in absolute terms.
Why did PlanetScale remove the free tier?
Sustainability — running per-customer Postgres or MySQL at no cost is genuinely expensive at scale, and PlanetScale chose to focus on revenue-generating customers over the indie/learning audience. The decision hurt their developer-mindshare meaningfully; Neon and Turso captured most of the indie attention since. PlanetScale remains a strong choice for established teams on MySQL but is no longer the default for new projects.
Is Turso a real Postgres alternative?
No, Turso uses libSQL (a SQLite fork), which has different transaction semantics, no full multi-table relational features, and a smaller ecosystem. For read-heavy workloads at the edge it is genuinely faster than Postgres, but for general-purpose application databases Postgres remains the more flexible choice.
Can I use Convex for anything other than TypeScript apps?
In theory yes — Convex has SDKs for Python and other languages — but the productivity story is built around TypeScript-first development. Teams using Convex from non-TypeScript stacks tend to feel friction the architecture is not designed to absorb. For non-TypeScript teams, Supabase or Neon is usually the better fit.
Which serverless database has the best HIPAA story?
Supabase, with the HIPAA add-on at $350/month on the Team plan ($599/month). The combined platform layer with Vercel Pro BAA at $350/month puts you at $700/month for a defensible Next.js + Supabase HIPAA stack. Full setup detailed here. Neon and PlanetScale offer HIPAA on higher tiers; Turso and Convex do not have published HIPAA stories as of mid-2026.
Related reading
HIPAA-compliant Supabase + Vercel: the $700/month setup — production setup for healthcare apps using Supabase as the data layer.
Headless CMS Hub — when the CMS layer choice intersects with the database choice (Supabase as content vs separate CMS).
How I built a 25,000-page directory in Next.js — the production case study at scale, with Supabase as the data backbone.
WordPress Stack Advisor — the tool itself runs on Supabase + Vercel; production reference for the Supabase + Next.js pattern.
The database pick is rarely the bottleneck. The bottleneck is whether the team can ship in whichever database you pick. Pick by primitive-fit, not by feature checklist.
Book a 30-minute database call — describe the app shape, the team's stack expertise, the scale projection. Walk away with a Supabase-vs-Neon-vs-Turso decision that fits the brief.