Fazlur Rahman
PlanetScale discontinued the free Hobby plan in 2026 and now requires a paid subscription starting at $5/mo for Postgres non-HA single-node (PS-5 arm64) or $15/mo for MySQL/Vitess HA 3-node (PS-5). Postgres scales from PS-5 at $15/mo HA to PS-2560 at $4,529/mo HA. Metal (local NVMe) tiers run higher: M-160 with 2 vCPU/16GiB/100GiB costs $720/mo HA. Vitess production deployments include 3-instance primary+replica + 1,440 dev branch hours. Enterprise is custom-priced (PlanetScale official, 2026).
Last verified: Sep 16, 2026.
At a glance
- No free tier: Hobby discontinued; $5/mo minimum
- Postgres non-HA: from $5/mo (PS-5 arm64)
- Postgres HA 3-node: from $15/mo (PS-5) to $4,529/mo (PS-2560 arm)
- Metal HA: M-160 $720/mo (2 vCPU, 16 GiB, 100 GiB)
- Vitess production: 3 instances + 1440 dev branch hours included
- Enterprise: custom pricing; BYOC available
Why PlanetScale discontinued the free tier
PlanetScale discontinued the legacy Hobby free tier in 2026, requiring all databases to be on paid subscriptions starting at $5/mo. The pivot reflects PlanetScale's shift toward enterprise customers willing to pay for production-grade Vitess and Postgres database management. Free tiers are expensive to operate at scale, and PlanetScale determined that $5/mo is a reasonable minimum for serious workloads (PlanetScale official documentation, 2026).
The smallest HA Postgres cluster on PlanetScale is now $15/mo (PS-5 HA, 3 nodes of 1/16 vCPU, 512 MiB each). For serious workloads, the entry-level production cluster is PS-160 arm64 at $96/mo non-HA or $286/mo HA (PlanetScale official pricing, 2026).
Postgres pricing tiers (HA, 3-node, arm64)
PlanetScale Postgres HA pricing on AWS US East 1 with arm64 architecture. The arm64 architecture is generally 20-30% cheaper than x86_64 for equivalent resources. HA pricing is 3x single-node pricing because HA includes 1 primary + 2 replicas across availability zones (PlanetScale official pricing, 2026).
The progression: PS-5 HA at $15/mo (1/16 vCPU, 512 MiB) is suitable only for development. PS-10 at $30/mo (1/8 vCPU, 1 GiB) supports small production apps. PS-160 at $286/mo (2 vCPU, 16 GiB) is the realistic starting point for customer-facing production. PS-640 at $1,135/mo (8 vCPU, 64 GiB) supports mid-market workloads. PS-2560 at $4,529/mo (32 vCPU, 256 GiB) supports large-scale applications (PlanetScale official pricing, 2026).
Side-by-side: PlanetScale HA Postgres pricing
| SKU | vCPU | RAM | Storage | arm64/mo (HA) | x86-64/mo (HA) |
|---|---|---|---|---|---|
| PS-5 | 1/16 | 512 MiB | Included | $15 | $15 |
| PS-10 | 1/8 | 1 GiB | Included | $30 | $39 |
| PS-40 | 1/2 | 4 GiB | Included | $83 | $99 |
| PS-160 | 2 | 16 GiB | Included | $286 | $349 |
| PS-640 | 8 | 64 GiB | Included | $1,135 | $1,399 |
| PS-1280 | 16 | 128 GiB | Included | $2,265 | $2,799 |
| PS-2560 | 32 | 256 GiB | Included | $4,529 | $5,599 |
Metal tier for performance-critical workloads
PlanetScale Metal tiers use local NVMe storage instead of network-attached storage, providing higher IOPS and lower latency at higher cost. Metal is positioned for performance-critical workloads that need consistent low latency and high throughput. Metal storage scales from 10 GiB (M-10) up to 61 TB (M-7680). Pricing for M-160 with 2 vCPUs, 16 GiB RAM, and 100 GiB included storage is $720/mo HA (PlanetScale official pricing, 2026).
Metal also includes higher network transfer allocations than non-metal tiers. For example, M-160 includes 100 GiB egress (compared to 30-40 GiB for equivalent non-metal PS-160). For workloads with high egress — such as read replicas serving global traffic — Metal can be more cost-effective than equivalent non-metal tiers (PlanetScale documentation, 2026).
Vitess pricing model
PlanetScale Vitess (MySQL-compatible sharded database) pricing is based on cluster size, storage, branch hours, replicas, VTGates, and read-only regions. A single Vitess database cluster includes resources equivalent to 5 always-on instances within the base monthly plan cost. Production deployments include 3-instance primary+replica + 1,440 dev branch hours included.
Additional components billed separately: VTGates (Vitess proxy layer) for routing, additional replicas for high availability or higher QPS, read-only regions for global read scaling, and branch hours beyond 1,440 per month. The pricing model is more complex than Postgres because Vitess is designed for sharded workloads at much higher scale (PlanetScale official documentation, 2026).
Hidden costs and migration traps
Beyond cluster pricing, four cost lines determine PlanetScale total cost of ownership. First, storage overage: cluster SKU pricing includes default storage, but additional storage is billed separately. Second, branch hours: 1,440 hours of dev branch time are included monthly, but extended development periods with multiple always-on branches can add hundreds of dollars. Third, backups: backup storage is billed separately. Fourth, VTGates and read-only regions: for global production deployments, these add-ons can multiply the base cluster cost by 2-3x (PlanetScale official pricing, 2026).
For teams migrating from MySQL to Vitess, the migration overhead is significant. Vitess adds operational complexity that requires expertise to manage. For workloads that don't need sharding, traditional managed MySQL (AWS RDS, Cloud SQL, PlanetScale Postgres) may be simpler and cheaper (PlanetScale official, 2026).
How PlanetScale compares to alternatives
PlanetScale competes in three distinct markets. For MySQL with sharding and branching, PlanetScale Vitess is the most production-tested option, but AWS Aurora MySQL and Vitess-on-self-hosted are alternatives. For managed MySQL without sharding, AWS RDS, Cloud SQL, and PlanetScale Postgres are alternatives. For Postgres, Neon (serverless), Supabase (BaaS), AWS RDS, and self-hosted are alternatives.
PlanetScale's unique strengths are branching (database branching for preview environments, similar to Git branches), Vitess sharding for horizontal scaling, and a unified Postgres+MySQL platform. The recent discontinuation of the free tier signals PlanetScale's pivot to enterprise customers willing to pay for production-grade database management (PlanetScale official, 2026).
What enterprise buyers should do next
Three actions for organizations evaluating PlanetScale in 2026.
- Validate the no-free-tier impact. If you were using the Hobby tier for development, plan for the $5/mo minimum. For multiple development environments across a team, costs can add up.
- Compare Postgres to Neon for serverless workloads. PlanetScale Postgres is provisioned; Neon is serverless. For workloads with variable traffic, Neon may be cheaper. For steady always-on workloads, PlanetScale is more predictable.
- Test arm64 vs x86_64 pricing. PlanetScale arm64 is generally 20-30% cheaper than x86_64 for equivalent resources. If your application runs efficiently on arm64, choose arm64 for meaningful savings.
What to watch next
Three near-term datapoints. First, PlanetScale's continued enterprise pivot — expect more enterprise-focused features and pricing tier changes. Second, the broader MySQL and Postgres competitive landscape — Aurora, Cloud SQL, Neon, and Supabase are all positioning for PlanetScale customers. Third, Vitess sharding improvements — PlanetScale continues to invest in sharding features for horizontal scaling (PlanetScale official, 2026).
