Four months after its last funding round, Databricks Inc. today announced that it has closed an extremely late-stage investment worth more than $4 billion.

The Series L round was jointly led by Insight Partners, Fidelity Management & Research and J.P. Morgan. They were joined by more than a dozen other backers, including Robinhood Markets Inc.’s venture capital arm. Databricks is now valued at $134 billion, up from about $100 billion in August.

The steep valuation increase is a reflection of the company’s fast revenue growth. Databricks disclosed today that its annualized revenue topped $4.8 billion in the third quarter thanks partly to increasing demand from existing customers. The company’s net retention rate, a metric that tracks how users increase their spending over time, is over 140%.

There are also other drivers behind Databricks’ sales momentum. One of them is Lakebase, a managed PostgreSQL database it launched in June. The company disclosed today that the service has already been adopted by thousands of customers.

It will be a focus of the engineering investments Databricks plans to make using its latest funding round. According to the data management provider, the plan is to enhance the service and two other products called Agent Bricks and Databricks Apps. Companies can use the three offerings side-by-side to build applications powered by artificial intelligence agents.

Lakebase lends itself to storing features, the data points that AI models use to make decisions. Features are condensed versions of business records that are easier for neural networks to process than the original information. For example, an e-commerce company might distill statistics about website visitor activity into high-level data on what product listings draw the most interest.

The service supports so-called online feature serving. That’s a data management approach whereby a database not only provides AI models with features, but also regularly updates those features when new information becomes available. Lakebase can provide features to neural networks deployed on both Databricks’ platform and external infrastructure. 

Agent Bricks, another focus of the company’s engineering push, helps customers turn their AI models into agents. The tool generates synthetic training data that can be used to optimize agents for specific tasks. Additionally, it creates benchmark tests to measure the effectiveness of AI training runs and eases related tasks such as system prompt development. 

Companies can turn the AI agents they build with Lakebase and Agent Bricks into applications using Databricks Apps, the third product the company plans to enhance. The latter offering speeds up application development by automating several manual tasks. That includes the process of implementing authentication features, access controls and other cybersecurity guardrails.

“By anchoring transactional data in Lakebase, delivering intuitive experiences through Databricks Apps, and enabling advanced multi-agent systems with Agent Bricks, we’re giving customers a unified foundation to build trusted, high-performance data intelligent applications at scale,” said Databricks co-founder and Chief Executive Officer Ali Ghodsi.

The company also plans to invest in other areas. Databricks will use the funding to provide liquidity for employees, finance AI research and make acquisitions. The Lakehouse database service is based on technology it obtained through a May startup acquisition. 

Ghodsi told the Wall Street Journal today that the company hasn’t yet decided when it will go public. That suggests additional funding rounds could follow suit. Databricks has raised more than $14 billion from investors over the past two years.

Photo: Robert Hof/SiliconANGLE