The number that matters more than it looks

Data 360 ingested 32 trillion records in a single quarter this year — up 119% year over year. That's not a marketing statistic; it's a signal about where enterprise data architecture is actually heading. Organizations aren't experimenting with a unified data layer anymore. They're committing to one, at scale, because the AI initiatives layered on top of it don't function without it.

Zero-copy is the architectural shift underneath the growth

The more consequential change this year is quieter than the ingestion numbers: zero-copy architecture has gone from an edge case to the default pattern. Data Cloud can now query Snowflake, Databricks, or BigQuery warehouses in place, without physically moving data into Salesforce first. For any organization that spent the last few years avoiding a Data Cloud commitment because it meant duplicating a warehouse they'd already built, that objection has largely disappeared.

What it doesn't remove is the governance work. Zero-copy access to a warehouse full of undocumented tables, unclear ownership, and inconsistent field-level definitions doesn't produce a trustworthy agentic foundation — it just makes the mess more visible, faster.

What this actually requires of your data model

Where this connects to the rest of the platform

This is also the piece that determines whether agentic work built on top of it holds up in production — see our take on multi-agent orchestration and why data hygiene is still the real blocker. And if part of your data estate includes reasoning workloads that don't fit neatly inside native Salesforce AI, the same governed-data logic extends to how Claude now operates directly inside Snowflake.