Databricks Builds an Agentic CDP Inside the Lakehouse

Databricks just entered the marketing software business. Its new platform puts the data warehouse exactly where the CDP category used to sit.

What Databricks Actually Launched

The announcement ended weeks of speculation. Databricks unveiled CustomerLake at its Data + AI Summit on June 16, calling it an agentic CDP built natively on the lakehouse.

The product folds the familiar CDP toolkit into one governed place. It brings profile building, identity resolution, audience building, campaign automation, and activation into the same environment where a company already runs its data and AI models.

The framing is architectural, not cosmetic. Built on the lakehouse and governed by Unity Catalog, the platform avoids the duplicate, move, and re-govern steps that a separate CDP forces on a data team.

It is shipping in stages. StorageNewsletter reported that CustomerLake is in Private Preview, with early adopters including HP, Circle K, AB InBev, and Getnet by Santander.

The Two Agents and Infinity Campaigns

The heart of the product is a pair of agent types. Profile Agents turn raw, scattered records into unified profiles, while Campaign Agents build audiences, recommend next best actions, and push content across channels.

Together they power what Databricks calls infinity campaigns. CMSWire described these as continuous, agent-driven loops that replace the manual campaign workflow with always-on engagement against a defined goal.

The pitch is that a marketer sets an objective rather than a journey. Grow loyalty enrollment or reactivate lapsed buyers, and the agents work toward that target by reacting to real time signals.

Chief executive Ali Ghodsi put the shift in blunt terms. He argued that marketing stops being a series of campaigns and becomes a continuous loop, with agents that constantly analyze, decide, and act on every buyer.

Why This Rattles the CDP Market

The competitive read is straightforward. CX Today noted that the launch pushes Databricks beyond its traditional analytics role and squarely into the marketing technology market.

Analysts see a structural change, not a single product. CMSWire cited a Gartner prediction that by 2030, 80% of net new enterprise CDP deployments will be embedded in or composable with data platforms, advice that frames CustomerLake as an infrastructure decision rather than a feature.

The independent take was similar. David Raab of the CDP Institute told CMSWire that Databricks is moving up the value chain from data management into execution, which simplifies life for teams but challenges conventional CDP vendors.

The Skeptic's Footnotes Matter Too

A closer look adds useful caveats. MarTech Therapy reported that through Lakehouse Federation the platform can query trusted data where it lives, in Snowflake, BigQuery, or object storage, without copying it first.

That federation point is the genuinely strong part. The governance travels with the query instead of being rebuilt on the far side, which is a cleaner answer than many composable rivals offered.

There is a reality check on the autonomy claims. The same analysis noted that demos still showed a person choosing channels and campaign duration by hand, so the current design looks more like assisted decisioning than fully hands off automation.

Pricing is the quiet disruptor. The platform is consumption priced with no separate platform fee, a choice that reorders how buyers compare it against subscription based CDPs.

What Operators Should Take From This

The takeaway is not that every team should rip out its stack tomorrow. It is that the boundary between the data platform and the marketing tool is dissolving, and that changes the buying question.

Tools that own the engagement layer, like HubSpot, are not going away, but the data foundation underneath them is becoming the contested ground. Where audience data lives now shapes what agents can do with it.

The launch also fits a wider pattern, with the largest AI platforms expanding into adjacent software categories to capture more of the value their infrastructure already touches. CustomerLake is that strategy aimed at marketing.

For now it is a Private Preview with a strong thesis and some honest gaps. The direction, though, is hard to argue with, because the platform that holds the data is deciding it no longer wants to be a passive supplier to everyone else's application.