Why the Battle for the CDP Isn't Really About the CDP Anymore

This week got a lot more interesting, fast.

It started Monday. Hightouch published their "Agentic CDP" manifesto the day before the Databricks Data + AI Summit. The timing was deliberate. They knew what was coming. The move was a pre-emptive flag-plant: concede "data warehouses may even claim the title of CDP, fair enough," then claim "Hightouch will remain the place where marketing gets done." Clever framing. But read it carefully and the structural gap is still there. Their agentic layer plans and recommends. It still routes execution to Braze, Iterable, and the ESPs. The loop doesn't close.

Then Tuesday, Databricks confirmed exactly why the flag-planting happened. CustomerLake: a native Agentic CDP built directly into their lakehouse. Serious product, serious team. If your data lives in Databricks, pay attention.

That same day, Axios reported that Hightouch had reportedly offered Publicis $1.2B for LiveRamp's identity assets. More than 40% of their total valuation. The bid apparently went out the week before the Databricks announcement. Which means they saw it coming and knew the manifesto alone wasn't going to be enough.

Think about what that tells you. A company that raised $150M two months ago at a $2.75B valuation reportedly bid over a billion dollars for something it doesn't own, because the warehouse it built its entire business on just became its competitor. That's not a strategic acquisition. That's a company realizing its foundation shifted overnight.

Running through all of this, Marty Kihn from Salesforce has been making a clean argument: the winners are the lower stack (warehouses) and the upper stack (full-suite platforms). The middle gets squeezed. He makes a compelling case, and for large enterprises running CRM-mediated B2B engagement across sales, service, and commerce, he's probably right. What that argument leaves open is B2C: a customer mid-checkout on an online store doesn't need a CRM suite. They need a decision in milliseconds.

We announced our answer yesterday when we acquired Blueshift. The timing looks convenient. It isn't accidental.

BlueConic captures first-party behavioral context as it happens: what each customer just did, what you've already shown them, what worked last time. That context feeds real-time decisions that fire in milliseconds, mid-checkout, hitting a paywall, walking away from a cart. With Blueshift, we now execute those decisions natively across email, push, in-app, SMS, and web, without routing through a third-party platform. Every outcome feeds back into the next decision. The loop closes in one system.

Databricks owns the data foundation. Salesforce owns the enterprise suite. Everyone fighting for the warehouse-native layer is competing on a foundation that just got a new landlord.

We own the runtime layer. Whether the customer showing up is a human mid-checkout today or their AI agent transacting on their behalf tomorrow, someone needs to be the real-time interface between brand and customer — making the decision, delivering the experience, logging the outcome, without a handoff in the middle. That's the layer we own.

None of them close that loop natively. We do.