Drop-in query engines
for your Snowflake stack.
TRUSTED BY THE BEST DATA TEAMS
cut embedded analytics warehouse costs by 92%. 
runs queries 3x faster at 78% lower cost. 
spends 89% less on analytics and automations.
NOT EVERY QUERY NEEDS A WAREHOUSE
Warehouses like Snowflake excel at massive workloads, but 99% of BI queries are small, only scanning less than 100GB.
Multi-engine queries are here.
Greybeam provides compute infrastructure to automatically route each query to its optimal engine.
HOW IT WORKS
01
Bring your own storage.
Greybeam caches your Snowflake data as Apache Iceberg tables in a bucket you own. Provision it with Terraform, CloudFormation, or any IaC tool you already use. Already on Iceberg? We'll read directly from your existing tables.
02
Connect to Greybeam
If your tool connects to Snowflake, it already works with Greybeam. Just replace your connection string and your analytics tools will work exactly as they do today. No complex integrations, no migrations, no code changes.
03
Query with any engine.
Greybeam manages the query engines, translates your SQL, and automatically routes each query to the most efficient engine. You don't change a single query.
Built for serious data teams.
USE CASES
Cost optimization.
When 99% of your queries run on DuckDB, you're only paying your warehouse for the 1% that need it.
Faster queries.
Queries run fastest when they run on the right engine. Greybeam makes sure they do.
Data portability.
Greybeam manages your data as Apache Iceberg tables, kept in sync with your warehouse, stored in your cloud, readable by any query engine.
Frequently asked questions
No. Greybeam is a drop-in Snowflake proxy, simply swap your connection string. Your dashboards, reports, and queries continue to work exactly as they do today. Greybeam automatically translates and routes your SQL to the optimal engine.
Most customers see 75-98% reductions in Snowflake compute costs for read workloads. The exact savings depend on your query patterns, but since the vast majority of analytical queries are small enough for DuckDB, the impact is significant. Want a savings estimate? Email us at founders@greybeam.ai
Greybeam can manage the entire Iceberg process, it's an implementation detail. If you aren't using Iceberg yet, we set it up and maintain it for you. If you are, we can read directly from those tables.
Simple and transparent. Compute starts at $0.75/hr for our smallest instance (8 vCPU), plus a flat $100/month platform fee.
Greybeam automatically falls back to Snowflake. If any issue arises in the DuckDB path, the query is transparently rerouted.
Typically less than 100ms for queries that fall back to Snowflake, mostly from SQL parsing. By design, queries routed to DuckDB often complete faster than they would on Snowflake.
Yes. All cached data is stored in your own S3 bucket within the security of your AWS account. Data is accessed via private VPC endpoints.
Greybeam authenticates every request with Snowflake before processing a query. If a user doesn't have access to data in Snowflake, they won't have access via Greybeam either.
For Snowflake-native tables, Greybeam seamlessly replicates your data into Iceberg format using CDC streams, keeping it continuously in sync. If you're already on Iceberg, we read directly from your existing tables.
Routing is based on heuristics like table size, join complexity, and query patterns. We're actively building a learned model to optimize routing decisions further.
Greybeam works alongside existing Snowflake optimizers, many of our customers already run both to compound savings. Because Greybeam sits in front of Snowflake, only the queries that truly need a warehouse get sent there, where your existing optimizers can do their thing.