PlanetScale Pricing 2026: Postgres, Vitess and Metal

PlanetScale Postgres starts at $5/mo for a single node, $15/mo for a high-availability cluster (one primary and two replicas across three availability zones) and $50/mo on Metal with local NVMe storage. Vitess, the MySQL-compatible option, starts at $39/mo. There is no free plan. Those are AWS us-east-1 list prices, as of October 2026, from PlanetScale's pricing page and the pricing catalog it publishes, read on 18 October 2026.

My read: PlanetScale is now one of the cheapest ways to run a small production Postgres, and one of the more expensive ways to run a hobby project, because the cheapest tier still bills every month. The real decision is single node versus HA, which roughly triples the compute line at every size.

Toolradar data: in the October 2026 snapshot on PlanetScale's Toolradar profile, its entry price ranks #3 of 20 serverless tools with verified entry pricing, against a category median of $10 a month. It is also in the minority without a free plan: 83% of its serverless peers offer one. Toolradar is the software directory we run.

PlanetScale plans at a glance

Every price is per cluster per month on the Base plan, which needs no contract or sales call. Prices vary by cloud and region, so a database in Frankfurt or on GCP will not match the us-east-1 figures here.

How PlanetScale bills

The cluster price is compute only. PlanetScale's own catalog says the listed price excludes extra storage beyond the included volume, backups beyond the defaults, egress above the included amount, a dedicated PgBouncer and replicas beyond the cluster's own. Those are the lines that turn the cheapest HA database into a larger invoice.

The main levers are the instance size, the topology (single node, three nodes, number of shards), storage, and usage such as backup storage and network egress. You pay month to month with no commitment, and you can resize up or down. Metal is the exception on storage: its local NVMe is part of the cluster price, so the number you see is closer to the number you pay.

What Postgres costs at each size

The ladder below is PlanetScale Postgres on AWS us-east-1 with ARM instances, the cheaper architecture. Single node is one machine; HA is a primary plus two replicas.

HA costs about three times single node at every size, because you are paying for two more machines. For a side project or an internal tool, a single node is a fair trade. For anything customers depend on, the HA column is the honest price, and a 1 vCPU, 8 GiB production database is $148 a month, or $1,776 a year before storage.

ARM is the default worth taking. The same 2 vCPU, 16 GiB HA cluster costs $349 a month on x86 against $286 on ARM, about 22% more for the same vCPU and memory. The smallest PS-5 size costs the same on both.

Metal: storage inside the price

Metal runs Postgres on local NVMe instead of network-attached disks, and the storage is bundled. The entry Metal cluster is 1/8 vCPU and 1 GiB with 10 GiB of disk on ARM. A 1 vCPU, 8 GiB Metal cluster with 100 GiB of NVMe costs $320 a month on ARM, and a 2 vCPU, 16 GiB cluster with 100 GiB costs $570.

Metal costs about twice as much as HA at the same compute size, so it only pays when disk speed is the bottleneck. A read-heavy app with a working set that fits in memory gets little from local NVMe; a write-heavy app doing large scans gets a lot. Price the Metal cluster against an HA cluster of the next size up before choosing, because extra memory often solves the same latency problem for less.

Vitess and Neki: paying for sharding

Vitess is PlanetScale's original MySQL-compatible product. Its smallest cluster, 1/8 vCPU and 1 GiB on three nodes, costs $39 a month, and the ladder runs to 64 vCPU and 128 GiB at $7,199. Vitess includes 10 GB of storage, then charges $0.50 per instance per additional GB, and PlanetScale's catalog lists it mostly on x86, so the ARM discount rarely applies.

Neki is the newer sharded Postgres. Each shard is a three-node cluster priced per shard, from $30 a month for the smallest ARM shard, with routers, storage, backups and network usage billed on top. Sharding multiplies the compute line by the shard count, so a four-shard Neki setup costs four times the per-shard rate before routers. Neither product makes sense until a single Postgres or MySQL primary is the limit.

Add-ons and costs outside the cluster price

• No free plan. PlanetScale's documentation says plainly that the old Developer or Hobby plan is gone. A side project pays at least the single-node rate every month.
• Postgres development branches are billed monthly, each at the price of the smallest single node. A team with three long-lived branches pays three times that every month, on top of production.
• Vitess development branches are metered. Each database includes two times the hours in the month of branch time (1,440 hours in a 30-day month), then about $0.014 an hour.
• SSO is a $199/mo add-on. On a small HA database that is more than the database itself, so budget it the day your security review asks for it.
• Extra backups are $0.023 per GB-month for user-scheduled backups beyond the defaults.
• Storage beyond the included volume on network-attached Postgres is priced by region, cloud and storage type, so check the calculator for your region instead of assuming a flat rate.

A first production database, costed

Take a typical startup setup: one Postgres database at 1 vCPU and 8 GiB on ARM with HA, two development branches, and SSO added after the first enterprise customer.

Storage beyond the included volume, backups and egress come on top and depend on your data. The SSO line is the biggest single surprise: it more than doubles this bill. Teams that only need SSO for a compliance checkbox should ask about Enterprise pricing before they add it to a Base plan.

How PlanetScale compares on price

At the small end, PlanetScale is competitive. Its $10 single-node 1 GiB database undercuts Supabase's pricing page, where Pro starts at $25 a month with compute credits that cover one Micro instance, though Supabase adds auth, storage and a free tier. Neon's pricing page charges $0.106 per CU-hour on Launch, with 1 CU equal to about 4 GB of RAM, so a 0.25 CU (1 GB) compute left on all month costs about $19 before storage, and less if it scales to zero when idle.

PlanetScale wins when the database is always on and you want HA without managing replicas. Neon wins for bursty or idle workloads because it scales to zero; Supabase wins when you want the backend around the database. The full matchup is in Neon vs PlanetScale, the line-by-line Neon bill in Neon pricing, and the Supabase bill in Supabase pricing.

Which PlanetScale setup to buy

Start on a single node for development, staging or a low-traffic app, on the smallest sizes in the ladder above. Move to HA the day an outage would cost you more than the HA premium, which is about $98 a month at 1 vCPU and 8 GiB. Pick Metal only after profiling shows storage I/O as the bottleneck. Pick Vitess or Neki only when one primary cannot keep up, because sharding multiplies the bill.

If you need a free tier, PlanetScale is the wrong place to start: the best free database hosting list covers the options, and cloud free tiers compared covers the wider platforms. For the product itself, read the PlanetScale review, set it against the Neon review and the Supabase review, and for the broader field the best Postgres hosting ranking. Directory profiles are at PlanetScale on Toolradar and Neon on Toolradar.

How we checked these prices

Every PlanetScale price above was read in US dollars for AWS us-east-1 on PlanetScale's pricing page, its pricing catalog and its plans documentation on 18 October 2026. Supabase and Neon figures come from their own pricing pages the same day. The startup bill and the ARM saving are our own arithmetic from those published prices. We did not run databases on any of these platforms. Prices differ by region and change without notice, so check the calculator before you deploy.

FAQ

How much does PlanetScale cost in 2026?

PlanetScale Postgres starts at $5/mo for a single node, $15/mo for a three-node high-availability cluster and $50/mo on Metal, in AWS us-east-1. Vitess starts at $39/mo. Storage beyond the included volume, extra backups, egress and add-ons such as SSO are billed separately.

Does PlanetScale have a free tier?

No. PlanetScale retired its free Hobby plan, and its documentation confirms there is no free plan today. The cheapest option is the smallest single-node Postgres database, billed every month.

How much does high availability cost on PlanetScale?

About three times the single-node price at the same size, because HA runs a primary and two replicas across three availability zones. A 1 vCPU, 8 GiB Postgres database on ARM costs $148 a month with HA, against roughly a third of that as a single node.

Is PlanetScale cheaper than Neon?

For an always-on database, often yes at the small end: a 1 GiB single node is $10 a month on PlanetScale, while a 1 GB Neon compute running all month costs about $19 in compute on Launch. Neon is cheaper for databases that sit idle, because it scales to zero and bills by the CU-hour.

What does PlanetScale Metal cost?

Metal's smallest cluster is 1/8 vCPU and 1 GiB with 10 GiB of local NVMe on ARM, and storage is included in the price rather than billed on top. At 1 vCPU and 8 GiB with 100 GiB of NVMe it is $320 a month.

What extra costs should I expect on PlanetScale?

Postgres development branches are billed monthly per branch, SSO is $199 a month, extra user-scheduled backups are $0.023 per GB-month, and storage beyond the included volume is priced by region. Vitess adds $0.50 per instance per extra GB after 10 GB.

Cite this: Devshot, "PlanetScale Pricing 2026: Postgres, Vitess and Metal", October 2026.

Devshot sends developers a short daily email on the tools and prices behind their stack. Subscribe at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.