Databases have long sat quietly beneath enterprise software stacks — essential, stable and largely invisible. They power everything, yet remain the most conservative layer in enterprise technology, carefully provisioned, slowly changed, and often firmly controlled by human operators. But Databricks argues that the era is ending, and collapsing faster than most enterprises realize.

According to the cloud-based data and AI company’s newly released State of AI Agents report, AI agents now create 80% of databases and 97% of test and development environments on the platform. Just two years ago, agents barely registered in database activity, with human developers handling nearly all of that work. The shift signals that AI is no longer confined to copilots, dashboards, or analytics layers. Instead, AI agents are increasingly operating core infrastructure themselves, spinning up databases, branching environments, and managing data workflows at machine speed.

“For forty years, databases were designed under the assumption that a human administrator was in the loop. When AI agents become the primary operators, that “pet” model breaks immediately. Agents operate at a velocity and volume that humans simply can’t match,” Reynold Xin, chief architect and co-founder of Databricks, told me.

He said an AI agent working through a complex coding problem may need to spin up dozens of isolated database environments in parallel, test multiple hypotheses at once, evaluate the results, and then tear everything down — all within seconds. Traditional shared-nothing architectures, where compute and storage live on the same machine, were never designed to handle this volume.

“The provisioning times are too slow, and the storage costs of physically copying data for thousands of environments are prohibitive,” said Xin. That dynamic helps explain why Databricks has pushed aggressively into a new database category it calls “lakebase”. With Databricks Lakebase now generally available, the company is placing a deliberate bet that databases must evolve beyond static systems built for predictable, human-driven workflows.

The company is recasting databases as an elastic, programmable infrastructure — something AI agents can create, branch, and manage at machine speed.

“We built the Lakebase with two distinct focuses to enable agentic coding. By completely decoupling the transactional engine (compute) from the data (storage residing in the lake), we allow agents to spin up stateless compute heads instantly while accessing the same underlying data without moving it,” explained Xin. “With the explosion of custom apps, IT teams are rightfully afraid of the ‘tsunami’ of maintenance that comes with thousands of new databases. Lakebase turns database management into a serverless, policy-driven experience, ensuring operational burden on IT remains flat.”

Databricks’ architectural push is unfolding alongside strong financial momentum.

In December 2025, Databricks raised more than $4 billion in a Series L round, valuing the company at $134 billion. As of the third quarter of 2025, the company reported an annual revenue run rate of $4.8 billion, reflecting more than 55% year over year growth. AI products and data warehousing now anchor that expansion, with each surpassing a $1 billion annual run rate. Moreover, early adoption of Lakebase outpaced the company’s core data warehousing business, with revenue growing at twice the rate in its first six months.

AI Agents Are Taking Over the Data Layer

A recent report from MIT Technology Review shows that 67% of organizations already use AI-powered tools, and more than half of business leaders view agentic AI as a force multiplier for operational performance and decision-making.

What is changing, however, is not just how many companies use AI, but how AI behaves once it enters production. Usage data collected across the Databricks platform, spanning more than 20,000 organizations worldwide, indicates enterprises are moving rapidly beyond single-purpose chatbots toward coordinated multi-agent systems that can plan, reason and execute workflows autonomously.

Moreover, use of multi-agent systems grew 327% in just four months, according to the Databricks report, as companies increasingly rely on supervisor agents to orchestrate other agents and tools across specialized domains. Nowhere is that pressure more visible than in the database layer, as AI agents also create 97% of database branches, cutting the time required to clone or rewind environments from hours to seconds.

Two years ago, agents accounted for just 0.1% of such activity. Today, they dominate it.

“We realized is that what’s great for agents is also often great for humans, except that for humans it’s more of a nice-to-have, but for agents it’s a necessity,” said Xin. “Of course, humans would love databases to be provisioned in seconds instead of minutes or days, but humans are used to the slowness. If an AI agent’s logic fails, it can programmatically roll back the database to the exact millisecond before the failure, adjust its reasoning, and try again in a fresh branch.”

Vibe coding, where users describe requirements in natural language and let AI generate code, has led to a surge in so-called citizen AI developers. Since the public preview of Databricks Apps, more than 50,000 data and AI applications have been created, with usage growing 250% in six months.

“With the rise of ‘vibe coding’ and AI-assisted development, the barrier to building software has been lowered dramatically, and it will continue to be lowered,” Xin said. “We expect enterprises to start building orders of magnitude more custom applications simply because it is now economically viable to do so. Lakebase is specifically designed for that explosion.”

Why Databricks Built Lakebase

Databricks is reshaping its core infrastructure for an agent-driven world. After acquiring database startup Neon last May, the company launched Lakebase, a serverless database designed specifically for AI agents. Neon’s serverless Postgres technology now underpins Lakebase, providing a foundation built for millisecond-scale branching, elastic scaling and programmatic control.

The company first introduced the lakebase category in June 2025 to unify operational and analytical data — two worlds historically separated by ETL pipelines, duplicated datasets, and inconsistent security models. Now that Lakebase is generally available, that architectural vision has moved from concept into production.

“Humans don’t create and destroy hundreds or thousands of databases a day, but agents do. It gave us a glimpse into the future last year based on the data from a very specific segment of the market,” Nikita Shamgunov, VP at Databricks and former CEO & co-founder of Neon, told me. “With vibe coding and agentic workflows, the infrastructure must be as fluid as the code itself. Developers want to express intent and have the system provision exactly what is needed to fulfill that intent instantly.”

Lakebase’s early adopters include Warner Music Group, easyJet, Rivian and more.

Warner Music Group uses Lakebase to connect real-time operational applications with large-scale analytics, pushing insights from audience data, rights management and content performance directly into production workflows. Likewise, at easyJet, engineers replaced a decade-old desktop system and consolidated more than 100 repositories into two, cutting development cycles for commercial applications from nine months to four.

Lakebase inherits the same governance and security model as the Databricks Lakehouse, allowing AI agents to interact with operational data within existing access controls. This reduces data sprawl and ensures that autonomous actions are grounded in trusted, real-time information rather than stale or duplicated datasets.

“We see serverless as the critical enabler of velocity, where Lakebase supports scale-to-zero and near-instant resume so organizations can run thousands of agentic experiments without operational bottlenecks,” explained Shamgunov.

A Database for the Agentic AI Era

Recent market data ranks Databricks third in enterprise data warehousing adoption, with roughly 15% penetration among relevant organizations, as of Feb. 2026. The company trails Snowflake, which leads with 62% adoption, and Amazon Redshift, at 29%. However, Databricks shows comparatively stronger traction in the midmarket, where adoption reaches 16%, and has differentiated itself through generative AI capabilities and unified analytics rather than traditional SQL-centric warehousing.

Industry forecasts suggest a structural change in how software itself is built. Gartner estimates that by 2028, 90% of enterprise software engineers will use AI code assistants, up from less than 14% in early 2024. In this emerging model, databases will no longer be passive repositories. Databricks is betting that Lakebase can serve as a future foundation, and if AI agents continue to dominate database creation at current rates, the company’s timing may prove prescient.

“The database is the system of record for AI applications. It’s no longer just a place to store rows; it’s the persistent memory and coordination layer for multi-agent systems,” said Shamgunov. “With Lakebase, we are making the database part of the fabric that connects humans’ and agents’ logic to enterprise data. It evolves from a passive storage tool into an active participant that ensures all actions are grounded, governed, and perfectly synchronized with the rest of the business.”

Source link