Amazon EBS pricing starts at $0.015 per GB-month for sc1 cold HDD and reaches $0.125 per GB-month for io2 Block Express, before any performance or snapshot charge lands.
Most of an EBS bill grows in two places the rate card doesn't cover: snapshot chains no one prunes and performance charges tuned to the previous day’s workload.
Amazon EBS pricing at a glance
All rates are US East (N. Virginia) list pricing from Amazon EBS’s pricing page, verified as of September 2026. Per-GB rates vary by region, so São Paulo and Sydney run higher than Virginia for identical configurations.
Billing runs in per-second increments with a 60-second minimum on volume storage, provisioned IOPS, and provisioned throughput.
Amazon EBS pricing breakdown
gp3 general purpose SSD: $0.08 per GB-month
What's included: 3,000 IOPS and 125 MB/s of throughput at any volume size, from 1 GB to 64 TB. Additional performance scales to 80,000 IOPS and 2,000 MB/s and bills separately, per the EBS volume types reference.
Best for: Almost every workload that used to sit on gp2. Boot volumes, single-instance MySQL and PostgreSQL, dev and staging, application servers.
Pros:
- ✅ Capacity and performance decouple, so a small volume can hold high IOPS
- ✅ 20% cheaper per GB than gp2 at identical baseline performance
- ✅ The free 3,000 IOPS baseline covers a large share of production volumes outright
Cons:
- ❌ Throughput above 125 MB/s bills at $0.04 per MB/s-month, which adds up on write-heavy volumes
- ❌ Caps at 80,000 IOPS, so the heaviest databases still need io2
A 1,000 GB gp3 volume with no extra performance costs $80.00 per month. The same capacity on gp2 costs $100.00 and delivers the same 3,000 IOPS. Migrating that one volume saves $240 per year with no configuration change beyond the volume type.
gp2 general purpose SSD: $0.10 per GB-month
What's included: I/O bundled into the per-GB rate, with baseline performance tied to size at 3 IOPS per GB, a 100 IOPS floor, and a 16,000 IOPS cap. Volumes under 1 TB burst to 3,000 IOPS.
Best for: Existing volumes you have not migrated yet. There is no new workload where gp2 is the right choice.
Pros:
- ✅ Single line item with no performance charge to model
- ✅ Bursting covers spiky small volumes without provisioning
Cons:
- ❌ 25% more expensive per GB than gp3
- ❌ Buying IOPS means buying capacity you do not need, since performance scales with size
A workload needing 9,000 IOPS on gp2 requires a 3,000 GB volume, costing $300.00 per month. On gp3, a 500 GB volume with 9,000 provisioned IOPS costs $40.00 in storage plus $30.00 in IOPS, or $70.00 per month. Same performance, 77% less.
io2 Block Express and io1 provisioned IOPS SSD: $0.125 per GB-month
What's included: io2 Block Express delivers up to 256,000 IOPS, 4,000 MB/s, 64 TiB, and 99.999% durability. io1 caps at 64,000 IOPS and 1,000 MB/s with lower durability, at the same per-GB rate.
Best for: SAP HANA, Oracle, IBM DB2, Microsoft SQL Server, and Cassandra clusters where sub-millisecond latency carries revenue.
Pros:
- ✅ io2 IOPS pricing tiers downward, so scaling performance on one volume gets cheaper per IOPS
- ✅ 1,000 IOPS per provisioned GB on io2, against 50 on io1
- ✅ Multi-Attach with I/O fencing across up to sixteen Nitro instances
Cons:
- ❌ Storage costs 56% more per GB than gp3 before a single IOPS is provisioned
- ❌ io1 charges a flat $0.065 per IOPS-month with no tier relief, making it the worse buy at every level
io2 IOPS tiers apply per volume, not per account. One volume at 60,000 IOPS bills 32,000 at $0.065 and 28,000 at $0.046, or $3,368 per month.
Splitting that same 60,000 IOPS across six volumes of 10,000 keeps every volume inside the first tier, at $3,900 per month. The split costs $532 more each month for identical aggregate performance.
st1 throughput optimized HDD: $0.045 per GB-month
What's included: 125 GB to 16 TB, 40 MB/s per TB baseline throughput, bursting to 250 MB/s per TB, capped at 500 MB/s per volume. I/O is bundled.
Best for: Sequential large-block work. Kafka brokers, log ingestion, ETL staging, data warehouse table scans.
Pros:
- ✅ 44% cheaper per GB than gp3 with no performance line to model
- ✅ Credit-based bursting supports a full-volume scan at burst rate
Cons:
- ❌ 500 IOPS per volume makes it unusable for random-access workloads
- ❌ The 125 GB floor wastes money on anything smaller
sc1 cold HDD: $0.015 per GB-month
What's included: The lowest per-GB rate in EBS. 125 GB to 16 TB, 12 MB/s per TB baseline, bursting to 80 MB/s per TB, capped at 250 MB/s per volume.
Best for: Cold data that has to sit on a block device attached to an instance and gets scanned a few times a day.
Pros:
- ✅ 81% cheaper per GB than gp3
- ✅ Same burst-credit model as st1 for occasional full scans
Cons:
- ❌ 250 IOPS per volume rules out anything interactive
- ❌ Object storage is far cheaper for cold data that does not need block semantics
EBS snapshot pricing: $0.05 per GB-month standard, $0.0125 archived
What's included: Point-in-time copies stored in S3 behind the service. The first snapshot of a volume stores every written block. Every snapshot after that stores only changed blocks and references the rest.
Best for: Active recovery points inside a short rollback window, and long-hold copies once a chain retires.
Pros:
- ✅ Empty blocks are never stored, so a 1,000 GB volume holding 200 GB of data snapshots at 200 GB
- ✅ Archive drops the rate 75% for genuinely long retention
Cons:
- ❌ Change is tracked in 512 KiB blocks, so a 4 KiB write bills as 512 KiB
- ❌ Deleting a snapshot might not lower the bill, because data referenced by later snapshots is preserved and reallocated
Snapshot blocks are indexed in 512 KiB units, confirmed by the block index definition in the EBS direct APIs. A database writing scattered 4 KiB pages across a 600 GB volume can touch 20,000 distinct blocks in a day while changing under 100 MB of real data.
The snapshot bills roughly 10 GB. Over 30 daily snapshots that rounding alone adds about $15 per month per volume, and it scales linearly with volume count.
Archiving converts an incremental snapshot into a full one. Archive a mid-chain snapshot holding an 18 GB delta on a 600 GB volume, and you pay $7.50 per month for a full 600 GB archived copy, while the 18 GB stays billed at standard rates because a later snapshot references it. The 90-day minimum locks in $22.50 for a move that reduced nothing.
Archive earns its rate only when a chain is retiring. A 1,122 GB standard-tier chain billing $56.10 per month becomes a single 600 GB archived copy at $7.50 once you delete the intermediate snapshots and keep only the last.
Snapshot and volume features that bill on their own line
Five capabilities never appear in the per-GB rate and rarely appear in a cost model.
Fast Snapshot Restore starts billing as soon as it's enabled on a snapshot in an Availability Zone and keeps billing until you disable it, whether or not anyone restores. EBS direct APIs bill when third-party backup or DR platforms read snapshot blocks directly.
Time-based Copy applies to cross-region jobs with a specified completion window. Provisioned Rate bills per volume created from a snapshot, and Volume Clones bill per same-AZ copy.
Fast Snapshot Restore left enabled on four snapshots across three Availability Zones for a full month costs 4 × 3 × 730 × $0.75, or $6,570. The snapshots themselves might store 1 TB at $50.00. The pre-warming outruns the storage 131 to 1.
Provisioned Rate scales with full snapshot size rather than the recovery footprint you needed. Volume Clones, by contrast, cost $0.48 to copy a 600 GB written volume in seconds, useful for CI pipelines that need production-shaped data on demand.
Cross-region and cross-account copy charges
Cross-region copies bill for data transfer plus standard snapshot storage in the destination region. Regulated workloads with a documented cross-region recovery point objective are where these charges concentrate.
Copies inside one account and region move no data and carry no charge, provided encryption status holds. After the first copy into a region, subsequent copies bill only for changed blocks against the most recent completed copy.
The first copy into any new region always bills as a full copy. Delete the most recent destination copy and the next one reverts to full. Re-encrypting to a different KMS key forces a full copy too.
A 600 GB snapshot seeded into a second region costs a full 600 GB transfer, then $30.00 per month to store there. Run that through Time-based Copy at a one-hour completion duration and the copy job adds $9.60. Two DR regions double the storage line, and a retention policy that trims the destination too aggressively resets you to full copies.
EBS charges hiding under EC2-Other
EBS volume storage, provisioned IOPS, throughput, and snapshot storage all bill under the EC2-Other service line, not under EBS itself. An EBS spend investigation usually starts as an EC2-Other spend investigation.
Usage types under EC2-Other are granular enough to separate volume storage, IOPS, throughput, and snapshots. Cost allocation tags on snapshots attribute a chain back to its owner. Volume charges keep running on stopped instances and on volumes attached to nothing, one of the standard AWS cost reduction categories.
Snapshot lineage has no owner. A volume gets deleted, its chain keeps billing, and nothing in the console connects the two.
Which EBS volume type to choose
Choose gp3 if you:
- Run any workload currently on gp2, without exception
- Need under 80,000 IOPS and under 2,000 MB/s
- Want capacity and performance priced independently
Choose io2 Block Express if you:
- Run SAP HANA, Oracle, DB2, or SQL Server where latency carries contractual weight
- Need more than 80,000 IOPS on a single volume
- Can consolidate IOPS onto fewer, larger volumes to reach the cheaper tiers
Choose st1 if you:
- Move large sequential blocks through Kafka, log pipelines, or warehouse scans
- Can live inside 500 IOPS per volume
- Hold at least 125 GB per volume
Choose sc1 if you:
- Hold cold data that still requires block semantics and instance attachment
- Scan it a few times a day at most
- Have already ruled out object storage for the same data
When EBS is the right storage choice
EBS is priced fairly for durable low-latency block storage attached to an instance. Overspend usually traces to configuration decisions that outlive the workloads they were built for.
EBS is worth the cost if you:
- Need block-level access with single-digit millisecond latency from EC2
- Run databases, boot volumes, or anything requiring a filesystem
- Size performance against measured utilization, then revisit it quarterly
Look elsewhere if you:
- Are parking cold or archival data on a volume, where object storage costs a fraction. Our guide to cutting AWS S3 costs covers the lifecycle rules that make that migration hold
- Are treating snapshot chains as a retention strategy, which they were never designed to serve
- Need shared access across many instances, where a file system fits better
The distinction between a rollback point and a governed recovery copy determines how much of your EBS bill is defensible. We cover it in depth in this snapshot vs. backup breakdown.
EBS storage alternatives and pricing comparison
Same protected data, three destinations, very different economics.
The stored footprint drives the bill more than the destination does. Three factors stack faster than any per-GB rate difference between tiers: 512 KiB block granularity, a full copy on every new region, and retention rules nobody has reviewed since the last audit.
63% of cloud IT leaders say high storage costs force them to retain or protect less data than they should, per Eon's 2026 Cloud Data Infrastructure Report. That same group discovers unprotected workloads only after an incident, an audit, or a restore that does not complete.
EBS snapshots vs. Eon for backup retention
EBS snapshots handle short-window recovery: same-day rollback, AMI builds, restores where the chain is small enough to reason about by hand. Eon takes over past that window, once retention runs into months and snapshot lineage has outgrown manual review.
Eon's vault fingerprints and deduplicates blocks across snapshots and resources, and tracks change at the 4 KiB filesystem page. Both mechanisms remove the rounding that 512 KiB block indexing and per-volume chains build into the native path.
Coverage is the other half. Snapshot spend grows because nobody knows which chains still map to a live resource. Cloud Backup Posture Management (CBPM) addresses this by discovering and classifying resources on creation, then applying retention by data class without manual tagging.
Cost Explorer attributes the resulting spend account-to-service-to-resource, so an unowned chain surfaces with a specific owner instead of aggregating under EC2-Other.
NETGEAR cut backup storage costs by 35% after moving off its previous platform. SoFi runs Eon's logically air-gapped vault across five AWS regions with recovery under five minutes, using policy-driven retention in place of manual snapshot coordination.
Most teams keep both: short-window EBS snapshots for same-day rollback, Eon for everything past that window. That keeps the standard tier small, where snapshot spend concentrates.
The bottom line on EBS pricing
An EBS bill grows in two places. Rates are the smaller one, with two fixes: migrate lingering gp2 volumes to gp3, and audit provisioned IOPS and throughput against measured utilization. Neither needs architectural review.
The bigger one is drift: orphaned retention, chains that outlived their volumes, and Fast Snapshot Restore left enabled on snapshots that no longer need pre-warming. EBS pricing is no longer a rate problem past that point.
Our guide to AWS Backup pricing and a survey of AWS cost optimization tools will take you further on the mechanics.
Can you name every EBS snapshot chain in your accounts and the resource it protects? Book a demo and see how Eon maps snapshot coverage across your accounts and what your deduplicated footprint would cost against your current EBS bill.
Frequently asked questions
How much does Amazon EBS cost per GB?
Amazon EBS costs between $0.015 and $0.125 per GB-month in US East (N. Virginia). sc1 cold HDD sits at $0.015, st1 at $0.045, gp3 at $0.08, gp2 at $0.10, and both io1 and io2 at $0.125. Provisioned IOPS and throughput on gp3, io1, and io2 bill on top of those rates.
Is gp3 cheaper than gp2?
Yes. gp3 costs $0.08 per GB-month against $0.10 for gp2, a 20% reduction at identical 3,000 IOPS baseline performance. The saving widens on IOPS-heavy workloads, because gp2 ties performance to capacity while gp3 prices the two separately.
How is EBS snapshot pricing calculated?
EBS snapshot pricing bills on the changed 512 KiB blocks a snapshot stores, at $0.05 per GB-month standard or $0.0125 archived. The first snapshot of a volume stores every written block. Empty blocks are never stored, so snapshot size usually falls well below provisioned volume size.
Does deleting an EBS snapshot reduce my bill?
Not always. Only data referenced exclusively by that snapshot is removed. Blocks referenced by later snapshots in the chain are preserved, and their storage cost is reallocated to the later snapshot. Deleting mid-chain snapshots often changes the bill by nothing at all.
Do I pay for EBS volumes on stopped instances?
Yes. EBS volume storage bills on provisioned capacity for as long as the volume exists, regardless of whether the attached instance is running, stopped, or absent. Unattached volumes left behind by terminated instances bill at the full per-GB rate indefinitely.
Is the EBS Snapshots Archive tier always cheaper?
No. Archiving converts an incremental snapshot into a full copy and enforces a 90-day minimum retention, plus $0.03 per GB to retrieve. Archiving a snapshot that later snapshots reference raises total cost, because the full copy bills in archive while the referenced blocks keep billing in standard.
Is EBS included in the AWS Free Tier?
Partly. The AWS Free Tier covers 30 GB of storage, 2 million I/Os, and 1 GB of snapshot storage. Accounts created after July 15, 2025, instead receive up to $200 in credits applicable to EBS among other services, with a free plan running six months from account creation.
