Amazon S3 Tables
An AWS-managed feature providing native Apache Iceberg tables as a built-in S3 capability with automated Binpack / Sort / Auto compaction (512MB default target, 64MB minimum), snapshot lifecycle management, and orphan-file garbage collection. Exposes the Iceberg REST Catalog natively and accepts direct Amazon Kinesis writes into table buckets without a Lambda intermediary.
Summary
An AWS-managed feature providing native Apache Iceberg tables as a built-in S3 capability with automated Binpack / Sort / Auto compaction (512MB default target, 64MB minimum), snapshot lifecycle management, and orphan-file garbage collection. Exposes the Iceberg REST Catalog natively and accepts direct Amazon Kinesis writes into table buckets without a Lambda intermediary.
S3 Tables removes the operational burden of managing Iceberg table lifecycle on S3. Instead of running your own compaction jobs and snapshot expiration, AWS manages it — cutting the Metadata Overhead at Scale pain point and delivering up to 3× query performance and 80% storage reduction (paired with Intelligent-Tiering) versus self-managed Iceberg on standard S3.
- Not a query engine. S3 Tables manages Iceberg metadata and compaction; you still need Spark, Athena, Trino, or DuckDB to read the data.
- Compaction strategy matters. Binpack handles unsorted tables; Sort (including Z-order) requires a declared sort order but enables much more aggressive file skipping. Auto picks for you but only works well if you've told it what to sort on.
- Kinesis → S3 Tables direct ingest removes the Lambda hop but doesn't remove the need to think about streaming schema evolution — that still hits Iceberg metadata.
implementsIceberg Table Spec — native Iceberg table managementimplementsIceberg REST Catalog Spec — native REST catalog endpointaugmentsCompaction — binpack + sort + auto strategies run continuouslysolvesMetadata Overhead at Scale — automated compaction, snapshots, orphan GCscoped_toLakehouse — managed lakehouse tables on S3constrained_byVendor Lock-In — AWS-specific managed feature
Definition
AWS-managed Apache Iceberg tables as a native S3 feature, providing table-level storage optimization, automatic compaction, and snapshot management without external infrastructure. Exposes the **Iceberg REST Catalog Spec** natively so external engines (Spark, Trino, Athena, DuckDB) attach as standard Iceberg clients. Accepts direct Amazon Kinesis writes into table buckets, removing the Lambda compute hop that previously sat between streaming ingest and Iceberg commit.
Operating Iceberg tables on S3 requires managing compaction, snapshot expiry, and metadata cleanup. S3 Tables automates these operations as a built-in S3 feature, reducing operational overhead for lakehouse tables.
Managed lakehouse tables on S3, zero-ops Iceberg tables, automated compaction and snapshot management, direct Kinesis → Iceberg streaming ingest.
Recent developments
- Region expansion: AWS China and GovCloud. S3 Tables launched in the AWS China (Beijing, operated by Sinnet) and China (Ningxia, operated by NWCD) Regions on January 8, 2026, and reached the AWS GovCloud (US) Regions in February 2026 — extending the managed-Iceberg tier to sovereignty-constrained and public-sector estates. Per Announcing Amazon S3 Tables (Amazon Web Services China) and Amazon S3 Tables are now available in AWS GovCloud (aws-news.com).
- Object monitoring is its own billable line item: $0.025 per 1,000 objects per month. This fee is specific to table buckets and has no equivalent on general purpose buckets, so tables holding very large object counts carry a standing monitoring cost on top of storage, requests, and compaction. Per Amazon S3 Tables vs Self-Managed Apache Iceberg on S3 (AWS Builder Center), corroborated by Amazon S3 Pricing: 23x Cost Gap Between Storage Classes (Cloudburn).
- Compaction prices cut up to 90% (July 1 2025, automatic). AWS reduced S3 Tables compaction processing fees significantly: per-object price is now 50% lower; per-byte processing prices are 90% lower for binpack compaction and 80% lower for sort + z-order. Reduced prices took effect July 1 2025 and apply automatically in all S3 Tables regions. Per AWS What's New — S3 Tables 90% compaction cost reduction.
- Three compaction strategies: binpack / sort / z-order. Binpack is the default and combines files for basic optimization; sort organizes data by specific columns to reduce file scans; z-order optimizes for multi-column query patterns. Per AWS — S3 Tables features.
- 2026 cost reality check after the price cut. Onehouse's follow-up analysis names it "AWS S3 Tables: After the 10× Priceberg Plunge" — the new pricing closes most of the cost gap with self-managed Iceberg but the managed-vs-self-managed math still inverts for the largest tables in some access patterns. Per Onehouse — S3 Tables After the 10× Priceberg Plunge.
- The earlier 20× surprise that triggered the price cut. Onehouse's original analysis (early 2026) documented operators hitting up to 20× higher costs than expected on managed S3 Tables when compaction processing fees were factored in. The AWS price cut was the response. Per Onehouse — S3 Managed Tables, Unmanaged Costs.
- Self-managed-vs-managed deep dive for startups. AWS's own builder.aws.com community published a technical deep dive comparing Amazon S3 Tables vs Self-Managed Apache Iceberg on S3 — useful as a vendor-neutral procurement reference. Per AWS builder — S3 Tables vs Self-Managed Iceberg.
- Native cross-region and cross-account replication for Iceberg tables (re:Invent 2025). S3 Tables can now replicate Apache Iceberg tables between AWS Regions and accounts while preserving snapshot lineage, parent-child relationships, and chronological commit order — read replicas of managed tables land within minutes, without hand-rolled copy pipelines. Per Amazon S3 Tables Just Got Smarter: Intelligent-Tiering & Native Replication Explained (AWS Builders / dev.to), corroborated by AWS S3: Advanced Features and Performance Optimization (dasroot).
- Storage carries a ~15% premium over S3 Standard; service now spans 32 regions and 100,000 tables per bucket. S3 Tables storage in us-east-1 is $0.0265/GB-month for the first 50 TB vs $0.023 for S3 Standard, plus per-object monitoring ($0.025 per 1,000 objects/month) and the compaction processing fees covered above — the storage premium is the steady-state cost line to model after the compaction price cuts. Per the cloudchipr S3 pricing breakdown and cloudburn's S3 pricing analysis; scale figures per AWS re:Invent 2025 S3 Tables architecture session notes (AntStack).
Connections12
Outbound9
scoped_to3implements2augments1constrained_by1Inbound3
Resources6
Official S3 Tables user guide covering managed Apache Iceberg tables with automatic compaction and snapshot management.
S3 Tables product feature page with pricing, integration details, and query engine compatibility.
Launch announcement describing how S3 Tables delivers up to 3x faster query performance and 10x higher transactions per second than self-managed Iceberg tables.
Reference for the Binpack / Sort / Auto compaction strategies, 512MB default target, snapshot expiration, and orphan-file GC policies running on S3 Tables.
Measured compaction benchmarks and explanation of Z-order sort compaction for multi-column pruning.
AWS prescriptive guidance on Iceberg compaction strategies and when to favor Binpack vs Sort for a given workload.