Data Plane · Cloud Data Warehouses

MotherDuck

Serverless cloud warehouse built on DuckDB, with dual execution splitting a query between laptop and cloud, and zero-copy sharing between accounts.

Overview

MotherDuck is DuckDB as a service, and the interesting part is that it does not assume the cloud is the only place a query should run. Dual execution splits the stages of a single query between your laptop and the cloud, or runs them entirely in one place, depending on where the data and the work sit.

That matters because the premise of DuckDB is that a great deal of analytical work fits comfortably on one machine. Most cloud warehouses require you to give that up: data goes to the warehouse, and every query costs warehouse compute regardless of size. MotherDuck keeps the local path and adds a cloud one, which is a genuinely different shape from Snowflake, BigQuery or Redshift rather than a cheaper version of them.

Zero-copy shares let another account query the same data without duplicating it, so collaboration does not mean copying a dataset per team.

The free tier is substantial enough to be a real evaluation rather than a demo: up to three internal active users and two service accounts, 10 GB of storage and 10 hours of compute a month, with role-based access control using preset roles and community support.

It is also explicit about agent use. The Lite tier lists shared context guides aimed at making AI agents perform better against your data, which places it among the tools treating agents as a first-class consumer rather than a bolt-on.

The constraint to weigh is the DuckDB heritage itself. Its strengths are single-node analytical work and fast local iteration; the largest, most concurrent enterprise workloads are what the conventional warehouses are built for.

Key features and capabilities

The same headings are used for every cloud data warehouses entry, so two tools can be read side by side.

Architecture and storage
  • Serverless cloud warehouse with DuckDB as the engine
  • Dual execution runs query stages on the laptop, in the cloud, or split across both
  • Zero-copy shares let another account query the same data without a copy
  • Compute is served by instances, with the entry tier limited to the smaller Pulse class
Performance and scaling
  • Inherits DuckDB's single-node analytical performance, which suits work that fits one machine
  • Local execution avoids a network round trip for data already on the laptop
  • Entry tier includes 10 hours of compute a month
SQL and workloads
  • DuckDB SQL, so queries developed locally run unchanged against the service
  • The same engine locally and in the cloud, rather than two dialects to reconcile
Open formats and sharing
  • Built on DuckDB, which is open source, so the engine is not proprietary
  • The managed service itself is commercial
Security and governance
  • Role-based access control, with preset roles on the entry tier
  • Service accounts distinguished from interactive users in the plan limits
AI and machine learning
  • Guides publish shared context intended to improve AI agent performance against your data
  • Agents treated as an expected consumer of the warehouse
Where it runs and what it costs
  • Fully managed cloud service; no self-hosted edition
  • Free Lite tier with 3 internal active users, 2 service accounts and 10 GB storage
  • Paid tiers lift the user, storage and compute limits

Pricing

SubscriptionFree Lite tier

Tiered, starting at a free Lite plan for up to three internal active users and two service accounts, with 10 GB storage, 10 hours of compute a month, Pulse instances only, preset-role access control and community support. Paid tiers raise those limits. The published free allowance is large enough for a real evaluation.

Vendor pricing page →

Demos and videos

About MotherDuck

MotherDuck builds a serverless cloud data warehouse on DuckDB, the open-source in-process analytical database. It publishes tiered pricing beginning with a genuinely free Lite plan, and documents the limits of that plan precisely — user counts, service accounts, storage and compute hours — which is more transparency than most of the warehouse category offers. Paid tiers scale users, storage and compute beyond those limits.

motherduck.com

Other cloud data warehouses tools

Amazon Redshift

Data Plane · Cloud Data Warehouses

Managed cloud data warehouse on AWS, with provisioned and serverless options.

  • Cloud service

ClickHouse

Data Plane · Cloud Data Warehouses

Open-source columnar database for real-time analytics, also offered as a managed cloud service.

  • Open core

Firebolt

Data Plane · Cloud Data Warehouses

Analytical database for low-latency, high-concurrency workloads, PostgreSQL-compatible and reading Iceberg. The engine is open source in preview; managed and bring-your-own-cloud options are sold.

  • Open core

Google BigQuery

Data Plane · Cloud Data Warehouses

Serverless data warehouse on Google Cloud.

  • Cloud service

Microsoft Fabric Data Warehouse

Data Plane · Cloud Data Warehouses

T-SQL data warehouse in Microsoft Fabric that stores its tables in Delta format in OneLake.

  • Cloud service

Snowflake

Data Plane · Cloud Data Warehouses

Fully managed cloud data platform with separate storage and compute, known for SQL warehousing and data sharing.

  • Commercial

Teradata Vantage

Data Plane · Cloud Data Warehouses

Enterprise analytics and data warehouse platform, available on major clouds and on-premises.

  • Commercial

Drafted with AI assistance and checked against the vendor’s own documentation.

MotherDuck: Cloud Data Warehouses · UDP