PlanetScale

What PlanetScale actually is

PlanetScale is a managed database platform built by the people who built Vitess — the MySQL sharding system created at YouTube in 2010, donated to the CNCF in 2015, and now a graduated project. The company was founded in 2018 by Vitess co-creators, is headquartered in San Francisco, and employs a large share of the core Vitess maintainers. That pedigree is the product's foundation: when PlanetScale says it does non-blocking schema changes and horizontal sharding, it is not wrapping someone else's technology — it is operating the codebase its staff maintains.

The catalogue is wider than most people's mental model. There are three engines now: PlanetScale Postgres (GA since September 2025), Vitess for MySQL-compatible sharded workloads, and Neki for sharded Postgres. Each is available on a Metal variant with local NVMe storage and unlimited IOPS. The workflow layer is shared across engines: database branching with deploy requests (a git-like model for schema changes), online schema migrations, insights and query analytics, and a serverless driver for edge runtimes (a Fetch-API driver for Vitess, plus compatibility with the Neon serverless driver's HTTP mode for Postgres).

The pitch that matters in practice: you get the scaling model of a top-tier engineering org — sharding, replicas, safe migrations — without hiring a Vitess team. The alternative comparison is not "MySQL on a $5 VM"; it is running your own Vitess or orchestrating your own Postgres replication and migration tooling. Against that, PlanetScale's pricing is easy to justify. Against plain managed Postgres, it is a harder sell until you need what it does.

One correction to make immediately: our database row carried a "free-tier" tag. There is no free tier. The Hobby plan stopped accepting new databases on March 6, 2024 and was fully retired on April 8, 2024, and PlanetScale's plans documentation says outright that no free plan exists.

Pricing: what you really pay

All published prices are AWS us-east-1 figures and vary by region and cloud provider — PlanetScale says so explicitly and even ships a machine-readable pricing.md endpoint you can query per region. That is the most transparent pricing infrastructure we have seen from a database vendor.

Current entry points (us-east-1):

• Postgres single node PS-5: $5/month — 1/16 vCPU, 512 MiB RAM, 10 GB network-attached storage included. Development and low-traffic workloads. This is the current entry price of PlanetScale.
• Postgres HA PS-5: $15/month — one primary plus two replicas across three availability zones.
• Postgres Metal: from $50/month — local NVMe storage, unlimited IOPS, storage included in the price.
• Vitess PS-10: $39/month — the entry Vitess cluster (1/8 vCPU, 1 GiB, three production nodes — one primary, two replicas — plus roughly 1,440 development-branch hours per month). Vitess is x86-64 only and has no single-node option.

Vitess SKUs run from $39 (PS-10) to $7,199/month (PS-2800: 64 vCPU / 128 GiB), with intermediate steps documented ($99 for PS-40, $349 for PS-160, $999 for PS-400, and so on). On top of the cluster price you pay for storage beyond the included 10 GB (about $0.125/GB), backup storage, network egress (roughly $0.06/GB; single-node PS-5 includes 10 GB, HA includes 100 GB), extra development branch hours, additional replicas, VTGates for Vitess routing, and optional dedicated PgBouncer for Postgres. Enterprise adds custom pricing, SSO, upgraded support with 15-minute P1 response, and PlanetScale Managed — PlanetScale deployed into your own AWS or GCP account, billed as your cloud costs plus a percentage.

In our view the pricing is reasonable for what it is, with two caveats. First, the metered extras (storage, egress, backup storage, branch hours) mean the invoice is not the sticker price — that is standard for databases, but the numbers on the calculator page are where the real figure lives. Second, the folklore that "PlanetScale is expensive" mostly comes from quoting $39 Vitess clusters as if they were the whole catalogue; the $5 Postgres single node is genuinely cheap.

Performance & infrastructure

PlanetScale runs on AWS and GCP — 12 AWS regions (N. Virginia, Ohio, Oregon, Dublin, Frankfurt, London, Montreal, São Paulo, Tokyo, Mumbai, Singapore, Sydney) and 7 GCP regions (Iowa, Ashburn, Montréal, Seoul, South Carolina, Belgium, Netherlands), 19 in total, with the default for new databases at AWS us-east-2. Enterprise can deploy in any AWS or GCP region with at least three availability zones, or inside your own cloud account. Availability coverage is as good as it gets short of the hyperscalers themselves.

The reliability story is concrete where it counts. HA Postgres and Vitess run a primary plus two replicas across three availability zones with automatic failover, and the HA configurations carry a 99.99% SLA. Backups are included with the platform and additional backup storage is metered; per-database IP restrictions ship without plan gates. SOC 2 and compliance documentation exist for enterprise buyers, and the Managed option addresses data-residency requirements by putting the whole deployment in your account.

On raw performance, the Metal tier is the interesting one: local NVMe storage and unlimited IOPS remove the network-storage ceiling that constrains conventional managed Postgres, which is where a large share of database latency actually lives. We have not benchmarked PlanetScale ourselves (our /benchmarks program covers general-purpose VMs), so treat vendor performance claims as claims — but the architecture argument is sound.

The developer experience is the other half of the infrastructure. Branching and deploy requests make schema changes reviewable and reversible; online schema migrations mean no downtime DDL on Vitess; the CLI, API, and dashboard are mature; and a planetscale_prices relation you can query with SQL for cost planning is the kind of detail that signals engineering culture. There is also an insights layer for query analysis.

Where it falls short

No free tier and no published trial — the cheapest way to evaluate is a $5 single-node Postgres, which is fine, but hobbyists who remember the free Hobby plan should recalibrate. Single-node PS-5 is explicitly not highly available; production workloads want the $15 HA cluster, and that is before storage and egress.

Vitess constraints are real: no foreign keys in the classic Vitess model, and the sharding mental model is a genuine learning cost. Neki (sharded Postgres) narrows that gap, but the two sharding systems are different products with different trade-offs. Read scaling requires replicas — the Vitess base cluster bundles two, but extra replicas, read-only regions, and VTGates all add to the bill.

Complexity of the bill itself is the recurring complaint: cluster plus storage plus backup storage plus egress plus branch hours plus add-ons, with region-dependent rates. The transparency tools exist (pricing.md per region, SQL-queryable prices), which is more than most vendors offer, but you still have to use them.

Finally, strategic dependency: you are buying into a venture-funded independent's roadmap in a market where database companies get acquired regularly. The Managed/bring-your-own-cloud option is the hedge — it puts the deployment in your own account.

Who should pick this

We would pick PlanetScale for teams running MySQL at scale who want Vitess without operating Vitess, and increasingly for Postgres teams that want branching and safe migrations with an HA three-zone topology as a checkbox. It is our pick when schema-change risk is your actual problem: the branching and deploy-request workflow is the clearest answer to "how do we ship migrations safely" that the managed database market has produced.

We would not pick it for a hobby project (a $5 VM or a free-tier database elsewhere does more per dollar), for teams allergic to metered usage components, or for anyone who needs foreign keys as a hard requirement on MySQL.

Key takeaways

• Three engines: Postgres (GA Sept 2025), Vitess for MySQL, Neki for sharded Postgres — each with a Metal local-NVMe variant.
• No free tier: Hobby was retired in April 2024. Entry is $5/month for single-node Postgres; HA is $15/month; Vitess starts at $39/month.
• Transparent, region-aware pricing: published SKUs up to $7,199/month (Vitess PS-2800) and a SQL-queryable pricing endpoint.
• 19 AWS/GCP regions, three-zone HA with automatic failover, and a 99.99% SLA on HA configurations.
• Branching, deploy requests, and online schema migrations are the core value — migrations as a reviewable process.
• Watch metered extras (storage, egress, backups, branch hours) and Vitess constraints (no foreign keys; sharding complexity).

Sources

• https://planetscale.com/pricing
• https://planetscale.com/about
• https://planetscale.com/docs/concepts/regions
• https://planetscale.com/docs/vitess/pricing
• https://planetscale.com/blog/self-run-vs-managed-vitess-with-planetscale
• https://stackwrite.com/blog/planetscale-pricing-2026
• https://whichdevtool.com/tools/planetscale
• https://toolradar.com/tools/planetscale/pricing
• https://layerbase.com/blog/planetscale-free-tier-alternatives
• https://howmakersbuild.com/tools/planetscale/pricing
• https://platform.tracxn.com/a/d/company/58be358fe4b0013e260d30b7/planetscale