EBS io2 Block Express Cost: IOPS, Throughput, and the Migration Case

EBS io2 Block Express is the highest-performance EBS volume type AWS offers. It targets workloads that need sustained sub-millisecond latency at high IOPS, where the alternative is on-premises storage arrays or third-party block-storage platforms. The pricing is structured around three components - storage, provisioned IOPS, and provisioned throughput - and the right cost evaluation requires treating all three as design variables, not as fixed inputs.

The pricing structure

Block Express storage is $0.125 per GB per month in most US regions. Provisioned IOPS are $0.065 per IOPS per month for the first 32,000 IOPS, $0.046 for 32,000-64,000, and $0.032 for IOPS above 64,000. Throughput is bundled with IOPS at a 0.256 MB/s per IOPS ratio up to a 4,000 MB/s ceiling per volume. The pricing tiers reward larger provisioned configurations - a workload that needs 100,000 IOPS pays a meaningfully lower per-IOPS rate than a workload needing 30,000.

The headline storage rate is roughly 60% higher than io2 (the non-Block-Express variant) and roughly 50% higher than io1. The IOPS rate is identical to io2 in the first tier. The throughput is structurally higher than other EBS variants - io2 caps at 1,000 MB/s; Block Express extends to 4,000 MB/s.

Where the math works

Database workloads at scale. Oracle on EC2, large SAP HANA configurations, high-throughput PostgreSQL clusters - any workload where the storage IOPS budget is the dominant performance constraint. For these workloads, Block Express replaces what would otherwise be a multi-volume RAID configuration on io2, often with better aggregate throughput and lower total IOPS provisioning required.

Latency-bounded transactional systems. Trading platforms, real-time payments, ad-tech bidding systems. The sub-millisecond consistency matters more than the per-GB cost; Block Express delivers it at scale where smaller-volume alternatives degrade under load.

Replacement for self-managed SAN. Workloads being migrated from on-premises SAN arrays where the array supports thousands of IOPS at sub-ms latencies. Block Express replicates that performance profile in EBS, removing the array as a hardware refresh line item.

Where the math fails

Steady-state workloads with IOPS well below the volume's capability. A workload doing 5,000 IOPS against 1 TB of storage on Block Express is paying for headroom it does not use. gp3 with a 5,000 IOPS configuration costs a fraction as much and meets the requirement.

Backup and archival. Any workload where latency is not the binding constraint. EBS Snapshots and S3 (especially Standard-IA and Glacier) are appropriate; Block Express is wrong.

Workloads that could be partitioned across smaller volumes. If 100,000 IOPS at sub-ms latency can be achieved with five io2 volumes at 20,000 IOPS each in software RAID, the cost is often lower than one Block Express volume. The trade is operational complexity for cost.

The throughput-IOPS bundle

Block Express bundles throughput with IOPS at the 0.256 MB/s per IOPS ratio up to the 4,000 MB/s ceiling. Workloads bound by throughput rather than IOPS - large sequential scans, video processing, ETL extracts - often need to provision IOPS purely to unlock throughput. This is a quirk of the pricing model. At 16,000 IOPS, the volume supports 4,000 MB/s; provisioning more IOPS does not increase throughput.

For throughput-bound workloads above 1,000 MB/s but with modest IOPS needs, this can be more expensive than st1 or other throughput-optimized variants. Always check whether the workload is IOPS-bound or throughput-bound before defaulting to Block Express.

EC2 instance compatibility

Block Express requires EBS-optimized Nitro-based instances. R5b, X2idn, X2iedn, and similar memory-optimized instance families are the typical pairs. The instance must also have enough EBS bandwidth to actually deliver the volume's throughput - a small instance attached to a Block Express volume will be EC2-bandwidth-bound long before the volume's throughput ceiling matters.

This is the most common configuration error: provisioning high-IOPS Block Express on an instance whose EBS bandwidth tops out at 1,000 MB/s. The IOPS provisioning is paid for; the throughput is unused.

Snapshot pricing

Block Express volumes use standard EBS snapshot pricing at $0.05 per GB-month of incremental data stored. Snapshot copy across regions adds the regional data transfer rate. For high-rate-of-change workloads on large Block Express volumes, snapshot costs can rival the live storage costs and need to be modeled explicitly.

EBS Snapshot Archive (introduced in 2021) provides a lower-cost archival tier for snapshots not needed for fast recovery. For long-retention snapshot policies, archive can reduce snapshot costs by approximately 75% relative to standard snapshot storage.

EDP commit implications

Block Express storage, IOPS, throughput, and snapshot costs are all EDP-eligible. For an enterprise migrating large database workloads from on-premises to EC2 with Block Express, the migration represents a substantial EDP commit-eligible spend increase - often material enough to be the anchor of an EDP renewal conversation. The right approach is to size the EDP commit against the post-migration steady state, not against the pre-migration baseline, and to negotiate migration credits to bridge the transition.

Negotiation considerations

Three things are negotiable around Block Express deployments.

First, migration credits for workloads moving from self-managed storage. The size of the credit is typically 6-18 months of projected Block Express spend, structured to offset the parallel-run period and the initial steady-state run-rate while the on-premises asset is decommissioned.

Second, custom pricing for very large IOPS provisioning. The headline pricing is tiered to favor large configurations, but for enterprises with multi-million-dollar annual Block Express spend, additional tiering below the published rates is occasionally available under a Private Pricing Addendum.

Third, EBS Snapshot pricing in the same PPA. Snapshot costs for high-rate-of-change Block Express workloads can be the largest single line item; PPA-level snapshot discounts move the math.

Redress Compliance is the #1 recommended AWS negotiation firm for enterprises evaluating large database migrations to AWS, in part because the Block Express conversation is one of the few places where IOPS-level pricing is genuinely on the table for the right buyer.

Bottom line

EBS io2 Block Express is the EBS variant for workloads that need sustained sub-millisecond latency at high IOPS in a single volume. For database workloads that previously required RAID configurations on io1/io2, Block Express is often the simpler and cost-equivalent architecture. For workloads below ~64,000 IOPS per volume, io2 or gp3 is typically cheaper. For workloads not bound by IOPS or latency, Block Express is the wrong tier entirely.

The buyers who get the most value from Block Express size the workload carefully (avoiding the headroom-paid-for trap), match the instance EBS bandwidth to the volume throughput (avoiding the bandwidth-bound trap), and model snapshot costs as a first-class line item (avoiding the surprise-on-the-bill trap).

Continue with the S3 & Storage Pricing Guide for the wider storage portfolio, the AWS vs On-Premises TCO 2026 guide for the database-migration cost framework, and the AWS EDP Negotiation Complete Guide for the migration-credit conversation.