Databricks is building ‘the operating system of the AI era’.
From Apache Spark to AI agents: Databricks‘s evolution follows that of the industry. At its Data + AI Summit 2025 in San Francisco, CEO Ali Ghodsi makes it clear that the company wants to be more than a data platform. It aims to become the backbone of how organizations use AI. With announcements like Lake Base, Agent Bricks, and a free education program for universities, Databricks wants to become accessible to everyone.
From Architecture to Ecosystem
The Lakehouse architecture remains the key aspect of Databricks’ strategy. Ghodsi points out that open standards aren’t just a marketing term: “99 percent of your data probably still sits behind closed platforms. That needs to change.” That’s why Databricks now supports full open collaboration between Delta Lake and Apache Iceberg via Unity Catalog. Databricks isn’t alone: support for Iceberg is also a flagship feature of major competitor Snowflake.
According to Glenn Lottering, Technical GM Benelux Nordics, regulated sectors such as banks, healthcare, and governments especially see the benefits of this open approach. “They know that good governance goes beyond access control. Visibility is at least as important,” Lottering explains.
Data Intelligence and Agent Bricks: AI with Context
The idea of data intelligence is that AI can provide specific answers based on your data and context. Databricks enables this through two tools: Genie and Assistant. Genie is an AI agent that executes text prompts within your datasets, and Assistant helps you code. Ghodsi: “You no longer need to be or have a programmer to work with data.”
The most notable announcement is Agent Bricks, a platform for “building high-quality domain-specific AI agents”. Companies can use the platform to benchmark their custom agents and modify them using various techniques.
Lottering knows that building agents is often not a technological problem, but a matter of clarity around their use. “We’re still early in the AI timeline, but the pace is increasing rapidly. Especially in sectors like healthcare, I see many use cases,” he says.
The Database Modernized
Lakehouse Database is a new cloud database architecture that allows agents to create databases. According to Ghodsi, at recently acquired Neon, 70 percent of databases were already created by agents. That’s a doubling in one year.
The openness that Databricks works with is important for customers, but not unique. Others offer the same openness. “Our customers want to build AI, but without lock-in. We offer that freedom. You need to stay multi-cloud and multi-model. That’s where things are heading.”
AI for Everyone
To make AI skills widely available, Databricks launched a free edition of its platform (no credit card or business login required) and is investing $100 million in education. Universities worldwide will get access to the complete learning package. This way, Databricks prepares students and young professionals for an AI-driven work environment. In India alone, they aim to train half a million people over time.
Databricks is ready to embrace and further develop AI. The technology is becoming increasingly powerful and accessible, becoming an integral part of our work processes. Those who stand still fall behind. Databricks welcomes with open arms those who want to keep learning.
read also