Citus Data
Summer 2011AcquiredReal-time. Big data. PostgreSQL
Pressure-test this opportunity
Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.
Real-time. Big data. PostgreSQL
Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.
Citus Data made PostgreSQL scale across machines without asking developers to abandon PostgreSQL. Its software sharded distributed tables, replicated reference data, and routed or parallelized queries across a cluster. The decisive architectural move was turning Citus from a PostgreSQL fork into an extension, then releasing the core as open source.
The company sold enterprise software and a managed cloud service around that extension. It raised a $9.5 million Series A in 2015, opened engineering offices in Istanbul, San Francisco, and Amsterdam, and became part of Microsoft in January 2019. Microsoft integrated the technology into Azure's PostgreSQL offerings while the Citus project continued in public.
Citus did not fail as a product. The acquisition ended the independent company and moved its commercial platform into a hyperscale cloud provider. Its history shows how an infrastructure startup can reduce adoption risk by aligning with an established open-source system, then become strategically valuable to a platform that needs a credible managed version.
Umur Cubukcu, Sumedh Pathak, and Ozgun Erdogan met in graduate school at Stanford. Erdogan and Pathak later worked as engineers at Amazon, while Cubukcu joined Boston Consulting Group. Their shared frustration came from NoSQL architectures that gained scale by giving up familiar relational features.
The founders wanted horizontal scale without abandoning transactions, joins, foreign keys, SQL, and the PostgreSQL ecosystem. Rather than write an unrelated database, they built on PostgreSQL. They founded Citus Data in January 2011, moved to Istanbul to open the first office, and joined Y Combinator's Summer 2011 batch.
The early product was called CitusDB. MixRank became its first production user in July 2012. Headquarters moved to San Francisco in 2013, while the Istanbul engineering base remained part of the company. That split gave Citus a durable technical center and access to US customers and capital.
Citus transforms PostgreSQL into a distributed database. Large tables are split into shards and placed across worker nodes. Reference tables can be copied to every node. A distributed query engine directs work to the relevant shards and combines results, allowing the cluster to use more CPU, memory, storage, and input-output capacity than one server.
The product initially lived as a PostgreSQL fork. That gave the team deep control but imposed a recurring compatibility burden when PostgreSQL changed. In 2016 Citus “unforked” the codebase and delivered it as an extension. Existing PostgreSQL users could add distributed behavior while continuing to use familiar drivers, SQL, monitoring, and administration practices.
Citus Data offered three consumption paths: an open-source extension, enterprise software for customer-managed deployments, and Citus Cloud as a managed service. This structure converted community adoption into commercial opportunities without making a proprietary database the only way to try the technology.
Citus targeted engineering teams whose PostgreSQL databases were outgrowing a single machine. Multi-tenant software, real-time analytics, event streams, and time-series workloads fit its partitioning model. The buyer was often an infrastructure or database team that wanted scale without rewriting an application for a different query language.
The broad database market was enormous, but Citus served a specific transition: PostgreSQL users reaching limits that vertical scaling, indexing, read replicas, or data archiving could no longer solve. That threshold made the immediate market smaller and technically demanding. It also created high value because migrations of core databases are costly and risky.
Citus competed with proprietary distributed databases, data warehouses, NoSQL systems, PostgreSQL forks, and application-level sharding. Its strongest distinction was compatibility with PostgreSQL itself. The extension model reduced migration cost but could not erase distributed-systems tradeoffs; teams still needed an appropriate distribution key and workload shape.
Cloud providers were competitors and potential acquirers. They controlled infrastructure distribution and managed-database relationships. Microsoft's Azure needed credible PostgreSQL scale, making Citus strategically useful beyond what a smaller vendor could distribute alone.
Citus Data first sold proprietary database licenses and support. Open-sourcing the extension widened adoption, while enterprise capabilities and Citus Cloud provided paid paths. The managed service eventually became more important than enterprise licenses, according to the team's later account of its open-source transition.
This model joined developer-led adoption with high-value infrastructure contracts. Open source reduced evaluation friction and exposed technical quality. Paid hosting removed operational burden. The tension was deciding which capabilities remained commercial; after the Microsoft acquisition, the remaining enterprise features became open source in 2022.
The independent company ended because Microsoft acquired it, not because the product disappeared. Microsoft's announcement tied the purchase to Azure PostgreSQL performance and scale. Its acquisition history confirms January 24, 2019, and current Microsoft documentation still presents Citus as an open-source distributed PostgreSQL extension.
The acquisition fit both sides. Citus gained cloud distribution, infrastructure, and a long operating horizon. Microsoft gained a proven PostgreSQL team, an extension architecture, and community credibility at a time when cloud databases were strategic. The transaction price and standalone financial results were not disclosed, so the economic outcome for investors cannot be calculated.
The company made two choices that improved its strategic value. First, it built around a database users already trusted. Second, it reversed the fork and adopted the extension boundary, lowering the cost of following upstream PostgreSQL. That technical alignment made Citus easier to run as part of a managed platform.
There was a tradeoff. A cloud provider could distribute the managed service more effectively than a startup, making acquisition a logical destination. Citus preserved the project and expanded its reach, but the independent commercial layer became part of Azure.