How Do S3 Egress Costs Compare With R2, B2, and Wasabi in 2026?

Direct Answer: S3 Egress Costs Still Dominate Many Data-Movement Budgets

For internet-bound data, AWS S3 remains one of the most expensive major object stores to use as a high-volume source, while Cloudflare R2, Backblaze B2, and Wasabi advertise materially lower or zero egress charges under their standard terms. As of 28 September 2026, the basic commercial contrast is straightforward: AWS charges for data transferred out of S3 to the internet, whereas R2 generally charges $0 for such egress, and B2 and Wasabi commonly offer 3 TB and 1 TB of free monthly egress respectively before charging a low per-terabyte rate. This does not make every workload cheaper elsewhere: storage-class choice, AWS region, destination type, free-transfer entitlements, retrieval fees, API operations, replication, and data-processing services can change the result dramatically. The right comparison is therefore total cost of transfer, not the headline egress price alone.

Also worth reading: How Should Platform Teams Compare Cross-Cloud Egress Pricing in 2026? · How Do You Compare Cloud Migration TCO Without Missing the Real Costs? · How Can Multi-Cloud Data Egress Reduction Actually Cut Storage Costs in 2026?

A useful rule is to calculate monthly egress volume, destination, and growth before changing platforms. At 100 TB leaving S3 each month, even a nominal difference of $40–$50 per TB can produce a $4,000–$5,000 monthly gap, or $48,000–$60,000 annually. At 1 PB per month, the same arithmetic becomes $40,000–$50,000 monthly. R2’s zero internet-egress model can therefore beat S3 by roughly 99% on that narrow line item, although the research supplied for this article describes that percentage as a scenario rather than a universal total-cost guarantee.

How AWS S3 Internet Egress Is Billed

S3 pricing generally separates storage, requests, and data transfer. Standard storage prices depend on region and storage class, while GET, PUT, LIST, and related requests are billed according to operation volume. Internet egress is priced per GB and varies by AWS Region and destination geography; AWS also provides a Transfer Credits program, but eligibility, existing credits, and the amount available must be checked. This means a team that already owns substantial AWS Transfer Credits may see a lower incremental S3 egress bill than a new customer comparing published list prices.

Tiered pricing also matters. S3’s published Data Transfer Out rates commonly decrease as monthly transfer volume rises, but the thresholds are defined by AWS pricing terms rather than by an organization’s aggregate cloud bill. A customer should not assume that all transfers to all destinations are pooled in the same way. Transfers between AWS services may qualify for internal pricing, EC2 data transfer may have separate rules, and CloudFront can introduce another delivery model with its own request and traffic charges.

The comparison becomes less dramatic when S3 is the origin for an already contracted AWS ecosystem. Existing committed-spend discounts, private networking, centralized billing, IAM controls, compliance evidence, and operational familiarity have economic value even when they do not appear as an egress line item. By contrast, moving hundreds of petabytes solely to obtain free public egress may create migration, duplicate-storage, networking, application, and governance costs that exceed the transfer savings. S3 is not “bad” at egress; it is simply priced as part of a broad cloud service whose economics become unattractive when one narrow cost dominates.

R2, B2, and Wasabi: What Their Low Egress Claims Mean

Cloudflare R2’s defining commercial advantage is $0 egress to the internet under its standard service model. It still has charges for Class A operations such as writes and listings, Class B operations such as reads, and storage above the included monthly allowance. Its architecture also uses an S3-compatible API, which can reduce application changes, although compatibility should be tested rather than assumed complete. Features and pricing outside ordinary internet egress—including inter-region movement, private connectivity, or related services—must be evaluated against current Cloudflare documentation.

Backblaze B2 and Wasabi target a similar cost-conscious position with low egress rates after free allowances. B2 is widely associated with a 3 TB monthly free egress allowance and a published low per-TB internet-egress price, while Wasabi is widely associated with a 1 TB monthly allowance and its own per-TB pricing. These are provider-specific programs with terms that can change, and free allowances may require account eligibility or may not apply in every circumstance. A buyer should verify the price and allowance on the provider’s official pricing page immediately before signing a contract.

These are not four interchangeable products. R2 may fit an API or public-content workload that needs cheap downloads. B2 may fit teams seeking conventional S3-style tooling and a free monthly transfer allowance. Wasabi may fit archival or backup use, provided customers understand its minimum-retention and acceptable-use conditions. S3 may still fit regulated, multi-service AWS environments, cross-region applications, or workloads whose replacement cost exceeds the egress saving.

Why Egress Economics Change the Architecture

Object storage is often selected for durability and operations rather than only for the bytes retained. Yet public egress converts storage into a distribution cost, especially for media libraries, software packages, AI datasets, customer exports, backups, and data products. When every downloaded terabyte creates revenue or avoids a customer penalty, high egress prices can make an otherwise valid business model uneconomic. This is why VDURA and Wasabi pairing GPU-adjacent AI storage with no-egress-fee S3 archiving has attracted attention: data adjacent to accelerated compute can generate repeated reads, and eliminating a large per-TB charge can materially improve workload economics.

The comparison must include the destination, however. Internet transfers, transfers to a customer’s data center, and transfers into analytics platforms may all be billed differently. Flexera’s Snowflake data-transfer guidance illustrates that warehouse ingestion and egress can add charges around the storage layer; moving bytes from S3 into Snowflake may reduce storage egress but does not necessarily eliminate data-processing or Snowflake-side fees. Cloudflare R2 can therefore appear dramatically cheaper for an internet download while still requiring paid movement when data enters compute, warehouse, or private-network destinations.

Architecture matters because replication can multiply traffic. A single object written once may be copied between regions, retrieved into a processing cluster, written again, and finally delivered to users. Multi-region durability may also incur request and transfer costs on the replacement platform. The winning design is not automatically single-region or multi-region; it is the design whose reliability target, recovery time, recovery point, and monthly byte movement are explicit.

A Practical Cost Model for Platform Teams

Start with a 30-day bill-of-materials inventory rather than a spreadsheet based on total storage. Divide objects into internal application reads, internet downloads, analytics ingestion, backup copies, cross-region replication, and disaster-recovery restores. For each category, record the average monthly GB, 95th-percentile burst, geographic mix, number of requests, and whether traffic is already covered by a free-transfer or committed-spend program. Repeat the exercise with 12-month growth and a one-time migration scenario.

Then price both retained bytes and moving bytes. A provider can win egress and lose on operations if a workload performs many small writes, frequent listings, or repeated short-lived reads. R2’s Class B request charges can be important for metadata-heavy or high-read systems. B2’s transaction classes, Wasabi’s policy charges, and S3’s request pricing should all be modeled using the actual request mix rather than assuming storage and egress are the entire bill. Include support plans, minimum retention obligations, network fees, data import or egress exceptions, and any paid features required by the application.

Run a controlled pilot using representative object sizes and request patterns. Test multipart uploads, range reads, conditional requests, inventory jobs, lifecycle transitions, versioning, object lock, replication, and recovery. Measure p50, p95, and p99 latency separately from price; a provider with near-zero egress may still be unsuitable if time-sensitive downloads fail latency objectives. Keep rollback capability and a checksum-based reconciliation process until the workload has completed at least one billing cycle under production-like traffic.

Alternatives, Lock-In, and Migration Traps

Cross-cloud object storage can reduce egress exposure, but it can increase operational complexity. Teams operating S3, R2, B2, and Wasabi simultaneously may need four sets of credentials, monitoring rules, lifecycle policies, incident procedures, and access-control models. That burden can overwhelm a $2,000 monthly saving. A better goal may be to standardize APIs where practical while preserving an exit path through S3 compatibility, documented object metadata, and independent encryption-key management.

AWS itself offers alternatives that can reduce transfer exposure without immediately abandoning S3. CloudFront can serve cached content closer to users, S3 Transfer Acceleration may improve long-distance uploads, and S3 replication can avoid some internet round trips in selected architectures. None automatically makes egress free. They can improve performance or lower effective cost, but they add caching, request, data-transfer, configuration, and invalidation considerations. The correct choice depends on traffic repeatability, cache hit rate, origin geography, and the customer’s ability to operate another AWS service.

Hybrid designs are also credible. Keep authoritative regulated data in S3 while replicating a reproducible derivative dataset to a low-egress store. Use S3 where AWS integrations and compliance controls dominate, and R2, B2, or Wasabi where public downloads dominate. Data gravity, sovereignty, latency, support coverage, and contractual commitments may outweigh price. x-oss.com’s relevant role is therefore not to declare one universal winner, but to help platform teams compare cross-cloud storage and data-plane options against the workload they actually operate.

Common Mistakes in S3 Egress Comparisons

The most common error is comparing one provider’s list price with another provider’s promotional allowance or negotiated rate. Comparisons must use the same month, region, tax treatment, support level, and destination type. Another error is dividing a one-time migration cost by only one month of savings; for a large archive, the migration may involve temporary dual storage and a temporary copy of all retained data.

Teams also overlook the price of small requests. A dataset with one object per file can create millions of GET operations, making request fees material even when egress is $0. Others omit lifecycle retrieval charges. Moving an object from S3 Glacier flexible retrieval or another archive class to Standard before export can incur retrieval fees, while low-cost object storage may include minimum-duration or early-deletion policies.

Finally, “S3-compatible” does not mean “feature-identical.” Object Lock, legal holds, replication controls, event integrations, IAM behavior, inventory consistency, conditional writes, multipart edge cases, and SDK defaults can vary. A zero-dollar egress line is not useful if the application cannot preserve its security or recovery requirements. Contractual terms matter too: free allowances can be changed, and unusual high-volume or abusive traffic may be constrained by provider policy.

When to Act and What the Decision Threshold Should Be

Act quickly when egress is a top-three cloud cost, when monthly public downloads exceed several hundred terabytes, or when growth is approaching a tier threshold where the current bill becomes difficult to predict. For example, a 20 TB monthly egress reduction from a higher-priced S3 path may justify a pilot even if the immediate saving is only hundreds or low thousands of dollars. A provider with millions of small objects needs a higher threshold because operations and engineering time may dominate.

Do not migrate merely because one vendor advertises “99% cheaper egress.” First validate that the workload is technically portable, the security model is equivalent, and the provider’s service terms fit the business. Establish a target such as a 25–40% reduction in total data-transfer cost, not merely a lower egress unit price. Review the result after 60–90 days of production traffic, comparing invoices, request counts, latency, support incidents, recovery tests, and staff time.

As of 28 September 2026, R2 is the clearest choice when internet egress is the dominant variable and the application can operate through its supported feature set. B2 and Wasabi deserve comparison when an S3-compatible low-egress model is desired and their respective allowances, policies, and support terms match the workload. S3 remains defensible when AWS integration, compliance scope, existing credits, or migration risk outweighs the transfer-price advantage. The durable answer is therefore conditional: S3 egress is expensive relative to specialist low-egress stores, but the economically correct storage platform is determined by total workload cost and operational requirements.