The CDP Is Dissolving Into the Stack Around It

For a decade, the customer data platform was sold as the center of the marketing stack, the neutral hub that unified every signal and fed the tools downstream. In 2026 that center is not holding. The CDP is being quietly absorbed into the platforms on either side of it: the data warehouse beneath it, and the engagement and conversation layers in front of it. Twilio’s SIGNAL 2026 launch, where a communications company shipped identity-resolved customer memory, is only the latest and clearest sign that the standalone CDP is becoming a feature, not a category.

How a conversation layer became a data layer

At SIGNAL on May 6, Twilio introduced Conversation Memory: identity-resolved customer profiles that connect data with conversation history, explicitly optimized for large language models to cut latency and token usage, alongside an Enterprise Knowledge API and a Conversation Intelligence layer that reads live interactions. Strip the branding and that is CDP work, unify identity, hold the profile, make it queryable in real time. Twilio is not the only vendor arriving at customer data from an adjacent starting point; it just happens to own Segment, the CDP it acquired in 2020, which makes the convergence literal.

The mechanism is the point. A CDP’s core job was always three things: ingest events from everywhere, resolve them to a single identity, and activate that profile in the tools that act. Once an engagement platform keeps its own identity-resolved memory, as Twilio’s now does, it has quietly taken over two of those three jobs for every conversation it powers. The reason to route that work through a separate CDP gets weaker with each release.

The bundling moment

This is not a Twilio quirk; it is the dominant 2026 pattern. CDP, messaging, and AI decisioning are converging into unified platforms because AI agents need the full loop, data, decision, action, running inside a single boundary. An agent that must call out to a separate CDP, wait, then act is slower and more brittle than one operating on memory it already holds. The architecture is bending toward consolidation because the agents demand it.

The merger math has pointed here for years. Standalone CDPs have been absorbed one by one: Twilio bought Segment, Rokt bought mParticle, Uniphore bought ActionIQ, and Fivetran bought Census in 2025. The category is not so much dying as being distributed into the platforms that surround it, a pattern we traced when CDP consolidation accelerated.

From central hub to context broker

Underneath the CDP, the same dissolution is happening from the data side. Warehouse-native and composable CDPs, Hightouch, Salesforce Data Cloud, Treasure AI, ActionIQ, and event tools such as Census and RudderStack, build identity and activation directly on the cloud warehouse (Snowflake, BigQuery, Databricks) rather than copying data into a separate store. Integrated, packaged platforms still command the bulk of category funding, but pure-play warehouse-native vendors are growing several times faster than the category average, according to 2026 industry analyses, the clearest tell of where new money thinks the architecture is going.

The packaged-versus-composable debate that once consumed martech conferences has quietly become a false binary; modern CDPs added warehouse connectivity, real-time streaming, and embedded AI, and most pragmatic teams now run a hybrid. The sharper reframe is functional: the CDP is shifting from being the central hub to being a context broker, a governed layer that resolves identity and enforces policy while the data itself sits in the warehouse and the action happens in the engagement platform. The category is even renaming itself around the idea. In February, Hightouch began billing itself as an agentic marketing platform; in April, Treasure Data became Treasure AI. The letters “CDP” are being sanded off the signage.

The skeptic’s case

The dissolution narrative deserves a counter-argument, and there are two worth holding at once. The first says the CDP was always transitional: separating customer data from enterprise data was a workaround for fragmentation that mature data platforms and direct agent access now make unnecessary, so its absorption is the system working as intended, not a loss. The second is the opposite risk, that as every engagement and conversation platform keeps its own memory, you end up with several partial customer profiles (one in Twilio, one in your email tool, one in the warehouse) and quietly recreate the fragmentation the CDP was invented to solve, now with vendor lock-in attached. Both can be true, which is exactly why the architecture choice can no longer be delegated to whichever platform you happened to buy.

What it means for the marketing leader

The decision is no longer which CDP to buy; it is where customer context should live and who is allowed to own it. Three questions follow. First, map where identity is resolved today and how many copies of the customer profile your stack is quietly maintaining, every platform that keeps its own memory is a candidate for drift. Second, treat portability as a hard requirement: if a conversation or engagement vendor holds an identity-resolved profile, know what it costs to take that context with you when you leave, because infrastructure chosen for support conversations is now a customer-data decision. Third, decide deliberately whether your governed source of truth is the warehouse or a packaged platform, and make everything else read from it rather than hoard its own copy.

The throughline connects to the broader shift we have tracked, from CDP consolidation to the arrival of the agentic marketing stack: as AI agents move to the center, the tidy boxes of the last decade are being redrawn around them. The CDP is not disappearing, it is dissolving into everything, which means the marketers who win will be the ones who decide, on purpose, where their customer context lives. Track the shift in Customer Data & CDP.