Databricks adds Postgres database with $1B Neon acquisition
The vendor's latest purchase comes six months after it raised $10B in funding. By adding PostgreSQL database capabilities, it aims to better enable users to build AI applications.
Databricks on Wednesday reached an agreement to acquire Neon, a cloud-based database platform built on PostgreSQL, in a move aimed at better enabling developers and autonomous AI agents to build applications.
Databricks did not disclose full financial terms but confirmed to Informa TechTarget that the price was approximately $1 billion.
Based in Menlo Park, Calif., Neon is a 2021 PostgreSQL database startup that raised $129.6 million before its acquisition by Databricks.
While the acquisition adds open source database capabilities that potentially serve as the foundation for agentic AI development, it also shows that Databricks has grown beyond competing with data management vendors to being in league with hyperscalers such as AWS, Google Cloud and Microsoft, according to Sanjeev Mohan, founder and principal of analyst firm SanjMo.
Based in San Francisco, Databricks was one of the pioneers of the data lakehouse format for storing structured and unstructured data together. Over the past two-plus years, it expanded into AI development. Now, adding PostgreSQL database capabilities akin to those offered by AWS, Google and Microsoft, it more resembles those tech giants than data management specialists.
"This acquisition propels Databricks into a new stratosphere that is today occupied by hyperscalers," Mohan said. "For years, Databricks and Snowflake were in competition with each other. What Databricks has done is a coup. Now, they own a standalone, full-flight Postgres operational database that runs independently and is not an extension."
Addition through acquisition
The acquisition of Neon follows a series of other acquisitions by Databricks over the past two years.
With enterprise interest in AI development surging since OpenAI's November 2022 launch of ChatGPT represented a significant improvement in generative AI technology, Databricks has responded by creating an environment where users can develop AI tools.
The vendor's June 2023 acquisition of MosaicML provided the foundation for what is now Mosaic AI, while subsequent acquisitions of vendors including Arcion, Einblick and Lilac AI all added complementary AI development capabilities. In addition, Databricks' acquisition of BladeBridge added data migration capabilities that aid AI development, and the purchase of Tabular added support for Apache Iceberg tables that can also be used when building AI applications.
Like the acquisition of Neon, the purchases of MosaicML and Tabular were valued at $1 billion or more. Databricks, however, has raised significant venture capital funding, including $10 billion in December, which has enabled it to strategically buy other companies to add technology.
By acquiring Neon, Databricks adds yet another set of capabilities aimed at making it faster and easier to build AI applications.
PostgreSQL, along with MySQL, is one of the two most popular open source database formats. Part of what makes PostgreSQL databases so popular is their flexibility, according to Mohan. Beyond traditional relational database capabilities, they can be extended to include geospatial, time series, JSON and vector database capabilities.
Neon enables users to create PostgreSQL instances -- new running environments of the PostgreSQL database server -- in seconds. In addition, with compute and storage separate from one another, Neon databases automatically scale as workloads change. And because they are simple to set up and run, not only can developers create new instances, but agents can also be trained to do so, with more than 80% of Neon databases created by agents, according to Databricks.
Meanwhile, as agentic AI becomes the major trend in AI development, many vendors are scrambling to add planning, reasoning, multi-agent orchestration and memory to their platforms, according to David Menninger, an analyst at ISG Software Research.
Agentic AI is all the rage right now. The Neon architecture makes it an ideal candidate to provide the memory function to agentic systems. It can be spun up nearly instantaneously and persisted as long as necessary, and then spun down.
David MenningerAnalyst, ISG Software Research
"Agentic AI is all the rage right now," Menninger said. "The Neon architecture makes it an ideal candidate to provide the memory function to agentic systems. It can be spun up nearly instantaneously and persisted as long as necessary, and then spun down."
Regarding the $1 billion price Databricks paid for Neon, the cost of the acquisition shows how hot agentic AI development is right now, he continued.
As far as Databricks' strategy of acquiring capabilities rather than developing them internally is concerned, Menninger noted that the vendor's valuation makes it easier to buy than build.
"Databricks has been rounding out the portfolio with additional capabilities, such as MosaicML for AI," he said. "Others have been tuck-in acquisitions such as Tabular to strengthen its position with respect to open table formats."
Mohan, meanwhile, noted that while Databricks has made some strategic acquisitions, its purchase of Neon is perhaps the most significant so far.
"I've been following Databricks acquisitions for many years, and this is the one that excites me the most," he said. "This one is practical. I don't think the last two major acquisitions [of MosaicML and Tabular] add enough return on investment -- they are nice-to-have technologies. This one does."
In particular, Neon's PostgreSQL database enables agents to learn and adapt, which provides significant return on investment, Mohan added.
Next steps
Following the acquisition, which remains subject to due diligence and other customary closing conditions, Neon's staff -- including co-founders Nikita Shamgunov, Heikki Linnakangas and Stas Kelvich -- will join Databricks.
Meanwhile, with Databricks' annual user conference Data + AI Summit taking place in San Francisco next month, there is more the vendor could do to serve the needs of its customers, Mohan said.
As Databricks becomes an ecosystem unto itself with a swath of data management and AI development capabilities that is beginning to rival those of AWS, Google Cloud and Microsoft, Databricks should unify its user experience, according to Mohan.
"There needs to be a common plane for multiple personas to access Databricks, a user interface where any persona -- data scientist, data engineer, data analyst, AI engineer -- can view all these individual products," he said.
In addition, Databricks could do more to unify capabilities through its Unity Catalog at the metadata level, Mohan continued.
"Unity Catalog should now be extended to Neon and agents so users can discover assets that are not just data or AI models but also the end-to-end agentic piece," he said.
Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than 25 years of experience. He covers analytics and data management.
