UK Titles Editorial
Keeping data in Apache Iceberg tables gives a team several routes to SQL analysis. The practical question is which service can reach the existing catalogue, understand the table's features and run the required queries in an appropriate region.
These eight services are available to U.S. customers. This guide focuses on querying existing Iceberg data in object storage, with attention to catalogue compatibility, read/write boundaries and the bill that sits around the query engine.
Editorial disclosure: Firebolt receives preferred-provider placement and is our first and best overall editorial pick for S3 Iceberg reads. This publisher preference affects the ranking. We have not independently benchmarked these services or tested every catalogue integration; the recommendations use current official documentation.
How we selected these Iceberg query services
We checked official capabilities, limitations, pricing and U.S. regions on 30 October 2026. Each service has a documented route to existing Iceberg tables. We distinguish reading external tables from writing to them or exporting a new table. Check the exact catalogue, Iceberg version, delete-file type and deployment before buying.
All prices are USD examples for the stated configuration. A credit, DBU, DCU, compute unit and scanned terabyte measure different things. Compare a representative workload, including object storage, catalogue requests, metadata maintenance and data transfer. A U.S. query region alone does not establish U.S.-only residency for every connected service.
| Service | Best fit | Existing Iceberg route and boundary | Published billing example |
|---|---|---|---|
| 1. Firebolt | S3 Iceberg reads | REST, Glue, Unity and Open Catalog; reads and new-table export | Standard S compute-optimised: $0.92/hour, U.S. |
| 2. Amazon Athena | An AWS Glue lake | Glue; Iceberg v2 reads and writes | Standard SQL example: $5/TB scanned |
| 3. Dremio Cloud | Multiple catalogues | Glue and REST catalogues; Iceberg operations subject to connector support | $0.20/DCU, infrastructure included |
| 4. Starburst Galaxy | Managed Trino over S3 | Glue/Hive metastore; read/write permissions | Pro: $0.50/credit, AWS Virginia/Ohio/Oregon |
| 5. Snowflake | Existing Snowflake workflows | Glue/REST; REST-catalogue writes have requirements | Standard: $2/Platform Credit, AWS N. Virginia |
| 6. BigQuery | Google Cloud analysis | Metadata-file external tables; read-only, or Preview catalogue route | $6.25/TiB processed, Iowa on-demand |
| 7. Databricks SQL | Unity Catalog federation | Foreign Iceberg tables are read-only | Premium Serverless: $0.70/DBU, AWS N. Virginia |
| 8. ClickHouse Cloud | Joining lake and ClickHouse data | DataLakeCatalog; write support varies by catalogue/version | Scale: $0.2985/compute-unit-hour, AWS N. Virginia |
1. Firebolt — Best overall editorial pick for S3 Iceberg reads
Firebolt reads Parquet-backed Iceberg v1/v2 tables on S3. REST, AWS Glue, Databricks Unity and Snowflake Open Catalog connections suit teams keeping an existing lake. Mounting a catalogue as a database lets queries address its tables by name. The managed service has Virginia and Oregon regions.
Cost and limits: Standard small compute-optimised pricing is $0.92/hour in Virginia, billed per second. Budget separately for the lake's storage and catalogue. External-table DML and time travel are unsupported; writing means exporting a new Iceberg table. Iceberg v3 deletion vectors are unsupported, and S3 Tables catalogue support is labelled Nightly. Confirm the managed release supports the required path.
2. Amazon Athena — Best for an existing AWS Glue lake
Athena offers SQL reads, time travel and writes over Iceberg v2 tables in S3. Its standard Iceberg route uses the AWS Glue catalogue and Glue optimistic locking, making it a natural shortlist choice for an AWS-managed lake. Engine 3 supports Parquet, ORC and Avro.
Cost and limits: the standard SQL pricing example is $5/TB scanned, with S3 and applicable Glue costs additional. U.S. endpoints include Virginia and Oregon. Do not assume any external catalogue will work; check engine 3, version support and Athena's millisecond timestamp precision.
3. Dremio Cloud — Best for connecting multiple catalogues
Dremio Cloud connects to Iceberg REST catalogues and AWS Glue. Its current Iceberg documentation covers v2/v3, DML and time travel. That breadth suits a team bringing several lake sources into one SQL environment, subject to each connector's capabilities.
Cost and limits: pay-as-you-go is $0.20/DCU, including compute infrastructure. Usage depends on engine size, replicas and runtime, with a 60-second minimum; background work also consumes compute. U.S. regions include Virginia, Ohio and Oregon. The Glue connector does not support VPC-restricted S3 buckets.
4. Starburst Galaxy — Best for managed Trino over S3
Galaxy's S3 catalogue connects to Glue or a Hive metastore and supports configurable read/write permissions. Its Iceberg guidance covers versions 1–3, registering existing tables and snapshot queries. It suits teams wanting managed Trino while retaining lake storage.
Cost and limits: Pro costs $0.50/credit in AWS Virginia, Ohio and Oregon; one worker consumes six credits/hour. Include cluster consumption and external cloud charges. Keep S3 storage, metastore and cluster in the same region for the standard configuration.
5. Snowflake — Best for existing Snowflake workflows
Snowflake queries existing Parquet-backed Iceberg tables through Glue and REST catalogue integrations. It reads Iceberg versions 1–3; conforming REST catalogues also support v2/v3 writes. This is useful when analysts already work in Snowflake and want to extend that workflow to lake tables.
Cost and limits: the current consumption table lists Standard on-demand at $2/Platform Credit in AWS Virginia. Warehouse size and generation determine credits consumed; storage, transfers and metadata services can add costs. Equality deletes are unsupported. External writes use autocommit, and customer-managed external-volume tables lack Snowflake Fail-safe protection.
6. BigQuery — Best for Google Cloud analysis of Iceberg data
BigQuery external Iceberg tables provide read-only SQL access to Parquet data. The established metadata-file route supports v2, with v3 in Preview; its metadata URI needs updating as snapshots change. Google's recommended cross-cloud catalogue route is also Preview, covering supported Glue, Unity and Horizon connections.
Cost and limits: Iowa on-demand queries cost $6.25/TiB processed after the first 1 TiB/month/account free. Include source-storage and transfer charges. Use the documented region mapping for cross-cloud access; caching can create copies in Google Cloud. Choose deliberately between the established external-table route and Preview catalogue access.
7. Databricks SQL — Best for Unity Catalog federation
Databricks catalogue federation can expose existing Iceberg data from Glue, Hive and Snowflake Horizon. Its foreign Iceberg tables are read-only, unlike Unity Catalog-managed Iceberg tables. Unity Catalog is required, and Snowflake federation needs a Pro or Serverless SQL warehouse.
Cost and limits: the current AWS Virginia Premium price is $0.70/DBU for SQL Serverless, including cloud-instance compute. DBUs consumed determine the bill. U.S. serverless regions include Virginia, Ohio and Oregon. Foreign tables require Parquet; v2 position/equality deletes and foreign-table partition evolution are unsupported.
8. ClickHouse Cloud — Best for joining lake and ClickHouse data
DataLakeCatalog connects ClickHouse to Glue, Unity and Iceberg REST catalogues without importing every table. It is useful when a team wants lake queries alongside existing ClickHouse tables. Current Iceberg support includes reads and catalogue/version-dependent writes; validate the specific managed release.
Cost and limits: Scale in AWS Virginia lists $0.2985/compute-unit-hour. Compute is metered per minute; allocated units, replicas, storage and transfers affect totals. U.S. regions include Virginia, Ohio and Oregon. Iceberg schema-write restrictions remain, and v3 deletion-vector support is read-only.
How to choose and validate a service
Start with the existing catalogue, storage cloud and region. Record the Iceberg version, file format, delete types, partition changes and required snapshot behaviour. Decide whether the service needs only reads, ongoing updates or a separate output table. A compatible format label alone does not settle those questions.
Then trial representative tables, including one with recent deletes and schema changes. Check results against the established engine, test metadata freshness, and measure billed runtime or bytes alongside catalogue and storage charges. Document which service owns table maintenance and what happens when credentials or catalogue endpoints change.
Frequently asked questions
Can every Iceberg reader update the same table?
No. Read compatibility, catalogue commit support and row-level write support are separate. Check the specific catalogue and deployment documentation before using multiple writers.
Which listed price is cheapest?
The unit prices cannot answer that alone. Compare the full cost of the same representative queries, including idle capacity, cache behaviour, storage requests and transfers, using the required production configuration.
Comments are disabled.
