Motherduck

MotherDuck operates a serverless cloud data warehouse built on DuckDB, the open-source in-process OLAP engine created by Hannes Mühleisen and Mark Raasveldt. The company positions the product as a simpler alternative to distributed cloud warehouses for the majority of organizations whose data does not reach petabyte scale, arguing that a scale-up architecture based on DuckDB can be faster, cheaper and easier to use. Its core architectural concept is "hypertenancy": each user, dashboard, customer or AI agent session receives an isolated DuckDB compute instance, referred to as a "Duckling," which the company says starts in milliseconds (cold starts under 100 ms) and can be sized across tiers named Pulse, Standard, Jumbo, Mega and Giga, with up to 16 read-scaling replicas.

The platform combines managed DuckDB compute with shared storage options including Apache Parquet, Apache Iceberg, DuckLake, differential storage and cloud object storage such as S3. A dual/hybrid execution engine can run query stages locally, in the cloud, or across both, and zero-copy read-only "share" objects let teams query the same databases without copying data. Product components include Flights (scheduled Python data pipelines running on MotherDuck compute), Guides (a business-context layer for AI agents), Dives (interactive, embeddable visualizations generated by an agent or SQL), AI Functions for LLM-based classification, extraction and summarization inside SQL, and an MCP server that gives agents such as Claude, ChatGPT and Cursor sandboxed read and write access with scoped tokens. Access is available through the MotherDuck UI, a CLI, REST, JDBC/ODBC, a Postgres endpoint and clients for Python, Node.js, Java, Go, Rust, TypeScript, R and WebAssembly.

As of the September 2023 Series B announcement the company reported more than 2,000 users, 30 employees with plans to reach 45 by year end, and $100 million raised in total at a $400 million post-money valuation. Public messaging on the current site has shifted toward "agent-first" analytics, marketing the warehouse as infrastructure for AI agents alongside data teams and application developers.

Founding story

The concept originated when Jordan Tigani, who had helped turn Google's Dremel query engine into BigQuery, saw DuckDB and concluded that someone should build a serverless version of it, initially as a possible side project. Looker founder Lloyd Tabb named the company, convinced Tigani it was more than a side project, and made two introductions: to investor Tomasz Tunguz, who led the seed investment, and to DuckDB creators Hannes Mühleisen and Mark Raasveldt. Mühleisen and Raasveldt had founded DuckLabs to develop the core engine and were seeking partners to build a commercial cloud offering, so DuckLabs became a co-founder of the new company. Tino Tereshko, another Google BigQuery veteran, was the first employee to join, and the starting team reached 11 people drawn from 11 different companies.

Business model

MotherDuck sells a commercial, managed cloud service layered on the open-source DuckDB engine, offering consumption-based analytics compute and storage plus platform features. The site references a free tier and a 7-day free trial, paid plans with a platform fee, and a startup program in which qualifying companies pay an $8,000 annual commitment for $16,000 of value ($13,000 in consumption credits plus a waived $3,000/year platform fee).

Subscription/consumption pricing: an annual platform fee combined with consumption credits for compute and storage, alongside a free plan and time-limited free trial. Startup program pricing is described as a 50% discount on an annual contract that must be signed within three months of acceptance.

Traction

As of the September 2023 Series B, MotherDuck reported more than 2,000 users after an invitation-only launch in June 2023, 30 employees expanding toward 45, 16 integrations at launch plus 11 more announced in September 2023, and $100 million in total funding. Current marketing cites 50+ supported data tools and customer case studies, including a claimed 30x speed-up on heavy queries after migrating from a Postgres cluster and a data pipeline runtime reduced from 8 hours to 8 minutes.

Latest developments

Current website messaging centers on agent-first analytics: hypertenancy with isolated Ducklings for every user and agent, an MCP server for agents including Claude, ChatGPT and Cursor, agent "skills" and an agent-first CLI, Guides as a context layer, Dives for agent-built interactive and embeddable visualizations, Flights for scheduled Python pipelines, AI Functions in SQL, and support for DuckLake and Apache Iceberg storage. The site promotes a livestream on August 26 titled "Do AI Agents Need a Semantic Layer?" and a startup program offering 50% off.