GCP persistent disk types cost: pd-standard vs balanced vs ssd

Quick answer

GCP persistent disks bill per provisioned GB-month by type: pd-standard ~$0.04, pd-balanced ~$0.10, pd-ssd ~$0.17, with Hyperdisk/Extreme adding provisioned IOPS. You pay for size, not bytes used. Default to pd-balanced for most workloads, pd-standard for cold bulk, pd-ssd only for latency-sensitive databases — and right-size capacity, because a half-empty disk bills for the whole thing.

Persistent disk cost is one of the simplest line items on GCP and one of the easiest to overpay, for two reasons: the type you pick can swing the per-GB rate 4x, and you're billed for provisioned capacity whether you use it or not. Get the type and the size right and disks are cheap; default everything to pd-ssd at generous sizes and they aren't.

The disk types

Type	~Cost/GB-mo	Best for
pd-standard (HDD)	~$0.04	Cold, throughput-oriented bulk storage
pd-balanced (SSD)	~$0.10	Default for VMs and boot disks
pd-ssd	~$0.17	Latency-sensitive databases
Hyperdisk / Extreme	+ provisioned IOPS	High, tunable IOPS/throughput

pd-ssd is ~70% more than pd-balanced. For most VMs, pd-balanced delivers SSD-class performance at the lower rate — the default that saves money without anyone noticing a difference.

You pay for provisioned size

A disk bills for its provisioned capacity, not the data on it. A 1 TB pd-balanced disk is ~$100/month whether it holds 1 TB or 100 GB. A fleet provisioned "with room to grow" can be paying for mostly empty space. Size to real usage with headroom; persistent disks can be resized up later without downtime, so you don't need to over-allocate upfront.

Don't forget snapshots and orphans

Snapshots bill separately per GB-month (incremental). Schedule them with expiration; un-pruned snapshot chains creep the same way EBS snapshots do on AWS.
Orphaned disks. Deleting a VM doesn't always delete its disks. Detached persistent disks keep billing at the full rate — reconcile periodically.

FAQ

What are the GCP persistent disk types and their cost?

Standard (pd-standard, HDD) is cheapest at ~$0.04/GB-month; Balanced (pd-balanced, SSD) ~$0.10/GB-month; SSD (pd-ssd) ~$0.17/GB-month; and Extreme/Hyperdisk add provisioned IOPS and throughput billed separately. You pay for provisioned capacity, not used bytes, so a 500 GB disk bills for 500 GB even if half-empty.

Which persistent disk type should I use?

pd-balanced is the right default for most VMs and boot disks — SSD performance at a moderate price. Use pd-standard for cold, throughput-oriented bulk storage; pd-ssd for latency-sensitive databases; and Hyperdisk/Extreme only when you need high provisioned IOPS that the standard types can't deliver.

Why is my persistent disk bill high?

Usually over-provisioned capacity or pd-ssd used where pd-balanced would do. Disks bill for provisioned size regardless of usage, so a fleet of 1 TB disks that are 70% empty wastes most of that spend. Defaulting every disk to pd-ssd also pays ~70% more than pd-balanced for performance many workloads don't need.

Do persistent disk snapshots cost extra?

Yes. Snapshots are billed separately per GB-month (incremental, so only changed blocks since the last snapshot) and are a common source of creeping cost when retention isn't managed. Set a snapshot schedule with expiration rather than keeping every snapshot forever.

How do I reduce persistent disk costs?

Right-size disks to actual usage, default to pd-balanced instead of pd-ssd, use pd-standard for cold bulk data, manage snapshot retention, and delete disks orphaned by deleted VMs (detached disks keep billing). Capacity right-sizing is the biggest lever.

How does C3X estimate persistent disk cost?

C3X prices a google_compute_disk from its type and provisioned size, so the cost difference between pd-standard, pd-balanced, and pd-ssd at a given capacity is visible before deployment.

What to do next

The disk type and size are explicit in your Terraform, so the cost is decidable up front. C3X prices a google_compute_disk from its type and provisioned size, so a pd-ssd default or an over-provisioned volume shows its cost in review — and you can see the saving from switching to pd-balanced immediately. The quickstart runs it in minutes.