PlanetScale Metal for Postgres Pricing and Sizing

8 posts · Mar 2025 – Dec 2025

What this trail covers

PlanetScale Metal offers enhanced database performance and cost-effectiveness by utilizing local NVMe drives instead of network-attached storage. This provides significantly higher IOPS and lower, more consistent latency, making it ideal for high-I/O workloads and reducing the need for read replicas. Durability is maintained through semi-synchronous replication across three availability zones and daily, tested backups. Metal is available in AWS and GCP, with pricing that is often more competitive.

All PlanetScale trails

Dec 2025$50 PlanetScale Metal is GA for PostgresThis post announces the GA of smaller PlanetScale Metal sizes for Postgres, starting at $50/month, and introduces the decoupling of CPU/RAM from storage capacity. This allows customers to provision resources more precisely to match their workload needs, offering up to 300GB of storage per GiB of RAM. The new sizes are available in AWS regions on Intel and ARM CPUs, with GCP support in progress. This change addresses customer demand for lower starting prices and more flexible storage options.
Nov 2025$50 PlanetScale MetalIntroduces a new entry-level pricing tier for PlanetScale Metal ($50/month) and more granular storage sizing, allowing independent allocation of CPU, Memory, and storage. This makes the high-performance Metal offering more accessible to budget-conscious startups and expands the range of available node types (M-10 to M-160) with flexible disk size options.
Mar 2025IO devices and latencyThis post provides a historical overview of non-volatile storage devices (tape, HDD, SSD), explaining their functionality and performance characteristics through interactive visualizations. It highlights the advantages of SSDs over tape and HDDs due to the absence of mechanical components, leading to lower latency. The post also introduces PlanetScale Metal, which utilizes locally attached NVMe drives for cloud databases, promising faster queries, low latency, and unlimited IOPS, contrasting it with network-attached storage.
Mar 2025Announcing PlanetScale MetalAnnounces the general availability of PlanetScale Metal, a new class of nodes powered by locally-attached NVMe SSD drives. This fundamentally changes the performance/cost ratio for hosting relational databases on AWS and GCP, offering unlimited I/O on `M-` cluster types. Customers have seen up to a 65% drop in p99 query latency and 53% cost savings compared to Amazon Aurora. Metal has been in production for 3 months, serving over 5 trillion queries across 5 petabytes of storage.
Mar 2025PlanetScale Metal: There’s no replacement for displacementThis post introduces PlanetScale Metal, a new offering that replaces network-attached storage (like Amazon EBS and Google Persistent Disk) with fast, local NVMe drives for MySQL-compatible and Postgres databases. It details the performance and cost advantages of local NVMe drives over network-attached storage, citing specific IOPS and latency improvements. The post also explains how Metal maintains durability through semi-synchronous replication and tested backups, and provides a cost comparison showing Metal's superior IOPS per dollar.
Advertisement
Mar 2025Upgrading Query Insights to MetalThis post details the migration of the PlanetScale Query Insights database, which is write-heavy with ~10k UPDATE/INSERT statements per second, from provisioned IOPS EBS volumes to PlanetScale Metal. The migration of the busiest shard to Metal resulted in a substantial decrease in query latency across all percentiles (P50, P90, P95, P99), with the upgraded shard executing queries faster than others. After upgrading all shards, similar performance improvements were observed, leading to a lower average backlog in Kafka consumers and increased capacity for future message volume, all without application or architecture changes.