Snowflake And Salesforce Data Cloud: The Complete Guide To Zero Copy In Data 360
Snowflake stores and models your enterprise data; Salesforce Data 360, formerly Data Cloud, turns it into customer context that Salesforce teams and Agentforce agents act on. This guide covers how Zero Copy connects them in both directions, the 2026 V2 setup, costs and limits.
Key Takeaways
- Snowflake and Salesforce Data Cloud (now Data 360) are partners, not rivals. Snowflake is the enterprise data platform; Data 360 turns that data into unified customer profiles, segments and context that Salesforce users and Agentforce agents act on.
- Zero Copy runs both ways. Data federation brings Snowflake data into Data 360 without copying it; data sharing exposes Data 360 objects inside Snowflake without loading them there.
- The Data 360 → Snowflake setup changed in 2026. The new Snowflake V2 data share target uses OIDC and an Enrollment ID. Legacy targets can no longer be created and must be migrated by October 31, 2026.
- Federation has three modes: live query (the default, nothing stored), cached query (refreshes every 15 minutes to 7 days), and file federation over Apache Iceberg (no Snowflake compute).
- Costs land on both platforms: Data 360 consumes credits for federation and sharing, and Snowflake bills the warehouse compute that runs federated queries.
- The best architecture is usually hybrid: ingest the data that drives identity resolution and activation, and federate the rest.
Most growing and mid-market companies that run Salesforce eventually hit the same wall: the customer data that sales, service and marketing need lives in a warehouse, and the warehouse team doesn't want another nightly copy job. Snowflake and Salesforce Data Cloud (now Salesforce Data 360) solve that together. Snowflake stays the place where the business stores and models its data at scale; Data 360 brings the relevant slice into Salesforce as unified profiles, insights and segments, without duplicating it.
This guide explains what each platform does, how the Zero Copy integration works in both directions, how to set it up with the current V2 connector, which federation mode to pick, what it costs on each side, and the limits most write-ups skip.
Do You Need Both Snowflake And Data 360?
You need both when your customer data lives in Snowflake and your customer-facing teams work in Salesforce. Salesforce itself states that Data 360 is not a replacement for Snowflake: Snowflake handles the broader data lifecycle, while Data 360 serves front-office users and AI agents that act on unified customer profiles. Zero Copy connects the two without a pipeline.
The integration fits if transactions, product usage, loyalty or billing data already sit in Snowflake and your sellers, service agents, marketers or Agentforce agents need that context inside Salesforce. It also fits when your data team wants Salesforce engagement and unified profile data in Snowflake for modeling, without building and maintaining an export.
You may not need it yet if nearly all of your customer data already lives in Salesforce, you have no warehouse, or the only requirement is a monthly report. Data 360 connects Salesforce CRM sources natively, and a reporting need may be served by Tableau or CRM reports first. Add Snowflake to the picture when warehouse data has to drive action, not just dashboards.
What Is Salesforce Data 360 (Formerly Data Cloud)?
Salesforce Data 360 is the data layer of the Salesforce platform. It ingests, harmonizes and unifies batch and streaming data from Salesforce and external sources, resolves records into unified customer profiles, and makes them usable in Sales, Service and Marketing, in analytics, and as grounding for Agentforce. It was named Data Cloud until October 14, 2025.
Salesforce now positions Data 360 as the unified data layer of the Agentforce 360 Platform: the context every agent reads before it answers or acts. Under the hood, it works in a sequence of stages that are worth knowing before you design a Snowflake integration, because Snowflake data can enter or leave at several of them.
Connect and ingest
Built-in connectors bring in Salesforce CRM, Marketing Cloud and Commerce data, files, and external systems in batch or streaming mode. Ingested data lands in data lake objects (DLOs). Zero Copy federation adds external data without ingesting it.
Harmonize to a data model
DLOs map to data model objects (DMOs), a standard model that separates the physical source from the business model. A renamed column in a Snowflake table doesn't have to break the segments built on top of it.
Unify identities
Identity resolution rulesets match records across sources into unified profiles. Calculated insights add metrics such as lifetime value or engagement scores, stored as calculated insight objects.
Act on the data
Segments activate to marketing and advertising channels, data actions trigger flows, Tableau analyzes the results, and Agentforce agents use the unified data as grounding. Data shares send objects out to Snowflake.
What Is Snowflake?
Snowflake is a fully managed data platform that runs on AWS, Microsoft Azure and Google Cloud. It separates storage from compute: data sits in a central repository, and independent compute clusters called virtual warehouses run queries against it. Each warehouse can be started, stopped or resized on its own, and compute is billed per second in Snowflake credits.
For a Salesforce team, a handful of Snowflake capabilities matter most, because they shape how the integration behaves:
- Secure Data Sharing. Shared objects are read-only, no data is copied between accounts, and the consumer pays only for the compute used to query them. This is the model the Data 360 → Snowflake direction builds on. Snowflake documentation.
- Apache Iceberg tables. Tables whose data lives in customer-managed object storage. Data 360 file federation reads Iceberg tables directly.
- Snowpark. Libraries for Python, Java and Scala that run processing inside Snowflake without a separate cluster. Snowpark documentation.
- Cortex AI Functions. Large language models called from SQL or Python, running inside Snowflake's security perimeter. Cortex documentation.
- Snowflake Marketplace. Third-party datasets that can enrich the customer view before it reaches Salesforce.
Data 360 Vs Snowflake: Side By Side
The two platforms overlap on storage and query, but they are built for different users and different jobs. Snowflake is where data engineers and analysts store, transform and model data. Data 360 is where that data becomes customer context inside Salesforce. Use this table to decide which platform owns each job in your architecture.
| Dimension | Salesforce Data 360 | Snowflake |
|---|---|---|
| Primary job | Unify customer data into profiles that Salesforce apps and agents act on | Store, process and analyze enterprise data at scale |
| Main users | Sales, service and marketing teams, Salesforce admins, Agentforce agents | Data engineers, analysts, data scientists |
| Data model | Standard customer data model (DMOs) plus identity resolution | Any schema you design: tables, views, Iceberg and hybrid tables |
| Compute | Managed by Salesforce; runs on AWS | Virtual warehouses you size; runs on AWS, Azure or Google Cloud |
| AI | Grounds Agentforce; AI Models (formerly Einstein Studio) | Cortex AI Functions, Snowpark ML workloads |
| Activation | Native segments, activations and data actions into Salesforce and ad channels | Not a native activation tool; feeds downstream systems |
| Pricing | Consumption credits (Data Services credits or Flex Credits) | Credits for compute, plus storage and data transfer |
| Role in the integration | Federates Snowflake data in; shares Data 360 objects out | Serves federated queries; mounts Data 360 shares as databases |
Sources: Salesforce Help, salesforce.com and Snowflake documentation, checked October 2026.
How Does Zero Copy Work Between Snowflake And Data 360?
Zero Copy lets each platform use the other's data where it lives, without an ETL job. Salesforce describes two directions: data federation (data in), where Data 360 queries Snowflake tables at runtime, and data sharing (data out), where Data 360 objects appear in Snowflake as databases that are queried on demand. Neither direction duplicates the data.
Snowflake Tables And Views → Data 360
A Snowflake connection in Data 360 maps external tables to external data lake objects. They can then be mapped to DMOs, joined with CRM data, used in calculated insights and segments, and read by Agentforce. Queries run live by default, with optional caching or Iceberg file federation.
Data 360 Objects → Snowflake
A data share bundles DLOs, DMOs and calculated insight objects and links them to a Snowflake V2 data share target. In Snowflake, the shares are mounted as catalog-linked databases. In Snowflake's words, the data stays in Salesforce and Snowflake queries it on demand.
How The Integration Has Evolved
| Date | Milestone |
|---|---|
| Sept 2023 | "Bring Your Own Lake" data sharing from Data Cloud to Snowflake becomes generally available |
| Apr 2024 | Bidirectional sharing between Snowflake and Data Cloud generally available; Salesforce launches the Zero Copy Partner Network |
| Oct 2025 | Agentforce 360 generally available; Data Cloud rebranded to Data 360 |
| Aug 2026 | Snowflake V2 data share target (OIDC) rolls out; legacy targets can no longer be created |
| Oct 31, 2026 | Deadline to migrate existing legacy Snowflake data share targets to V2 |
Sources: Salesforce News (2023), Salesforce press release (2024), Salesforce release notes (2026).
How To Share Data 360 Data With Snowflake (V2 Connector)
To share Data 360 data with Snowflake today, a Snowflake admin creates a Zero-Copy Connector and copies its Enrollment ID, a Salesforce admin creates a Snowflake V2 data share target with that ID and links a data share to it, and the Snowflake team mounts the shares as catalog-linked databases. Authentication uses OIDC, not a Snowflake username and password.
Prerequisites
- Data 360 in Developer, Enterprise, Performance or Unlimited edition (not Government Cloud). Salesforce is enabling V2 in sandbox and production orgs on a rolling basis from August 2026.
- A Snowflake account on Standard, Enterprise or Business Critical edition, in any AWS, Azure or Google Cloud commercial region. Virtual private Snowflake (VPS), government regions and China aren't supported. Data 360 runs on AWS, and cross-cloud and cross-region connections work.
- The Data Cloud Architect permission set in Salesforce. In Snowflake, the CREATE ZEROCOPY CONNECTOR privilege on a schema, plus CREATE DATABASE on the account to create catalog-linked databases.
Steps
- 1
Create the connector in Snowflake. In Snowsight, open Ingestion › Zero-Copy and connect "Salesforce Data 360 Connect for Snowflake" (or create a Zero-Copy Connector in SQL with the Salesforce partner option). Copy the Enrollment ID it returns.
- 2
Create a data share in Data 360. Under Data Shares, choose New, give it a label, pick the data space, and select the DLOs, DMOs and calculated insight objects to share.
- 3
Create the Snowflake V2 data share target. Under Data Share Targets, choose New, select the "Snowflake V2" connection type, and paste the Enrollment ID. The Snowflake connector moves to Connected on its own.
- 4
Link the share to the target. Open the data share and use Link/Unlink Data Share Target to attach it to the V2 target.
- 5
Mount and query in Snowflake. Mount the shares as catalog-linked databases from Snowsight or SQL. Snowflake creates the views automatically, and you can query the shared data lake, data model and calculated insight objects like any other database.
Two details matter for compliance. Every data share automatically includes a consent object (IndividualGDPRState) so the receiving team can honor consent by Individual ID. And migration from a legacy target is not automatic: link the share to a V2 target, verify the data, swap the databases in Snowflake, then unlink the legacy target. Step-by-step guidance is in Snowflake's setup guide and its migration guide.
How To Bring Snowflake Data Into Data 360 With Zero Copy Federation
To federate Snowflake data, create a Snowflake connection in Data Cloud Setup, authenticate with a key pair or with Salesforce as the identity provider, choose the Snowflake warehouse that will run the queries, and then create a data stream that maps the Snowflake tables or views you need. The data stays in Snowflake.
- 1
Prepare access. You need the System Administrator profile or the Data Cloud Architect permission set. Review the Data 360 IP allowlist with your Snowflake admin, and for large volumes ask them to set the security token expiry to 24 hours.
- 2
Create the connection. In Data Cloud Setup, open External Integrations, select Snowflake and choose New.
- 3
Choose authentication. Use key-pair authentication with a private key, or "Salesforce IDP Auth", which signs in through a dedicated Snowflake service user configured for OIDC. Don't convert a human user into the service user.
- 4
Select the warehouse. Federated queries run on the Snowflake warehouse you choose here, so size it for the workload and attribute its cost.
- 5
Create the data stream. Under Data Streams, choose New, select the Snowflake source, the connection and the database, then pick the objects. Map the resulting external DLOs to DMOs so segments, insights and agents can use them.
If the Snowflake account must not be reachable over the public internet, Private Connect for Data 360 supports Snowflake federation, but plan it up front: an existing connection can't be switched to Private Connect later. Salesforce's developer documentation covers each field.
Live Query, Caching Or File Federation: Which Should You Use?
Start with live query, the default, which stores nothing in Data 360. Turn on caching when repeated queries are slow or expensive and data that is minutes to days old is acceptable. Use file federation when your Snowflake tables are Iceberg tables in Snowflake Open Catalog and you want to avoid Snowflake compute charges.
| Mode | How it works | Best for | Watch out for |
|---|---|---|---|
| Live query Default | Data 360 sends each query to Snowflake at runtime; results are processed in memory and discarded | Freshest data, low query volumes, data you can't copy | Snowflake warehouse compute on every query |
| Cached query Accelerated | A local cache in Data 360, refreshed every 15 minutes to 7 days | Frequently queried data, segmentation, lower latency | Incremental refresh handles upserts only; deleted rows stay until a full refresh |
| File federation Iceberg | Data 360 reads Iceberg files directly from storage with its own compute | Large tables on AWS- or Azure-hosted Snowflake, cost control | Iceberg views and some timestamp types aren't supported |
Ingest Or Federate?
Federation is not automatically the right answer for every object. Salesforce's architecture guidance recommends a hybrid: ingest the data that drives identity resolution, regulatory reporting and activation, where performance and reliability are critical, and federate the long tail of reference and historical data that is useful but not used constantly. Zero Copy also respects source-level policies such as row-level security and masking, so governance defined in Snowflake carries through.
What Can You Do With Snowflake Data In Salesforce?
The value shows up where a Salesforce user or agent makes a decision with data that used to be out of reach. Salesforce's own examples for Snowflake federation cover sales insights, service context and segments activated in Marketing Cloud, and the sharing direction gives data teams Salesforce data for modeling.
Sales: account context
Show product usage, contract and invoice history from Snowflake on the account, and build calculated insights such as expansion signals that sellers can see without opening a BI tool.
Service: agents with full history
Give service reps and Agentforce service agents order, shipment or loyalty data held in Snowflake, so they can answer and resolve without asking the customer to repeat it.
Marketing: richer segments
Combine federated purchase and behavioral data with CRM engagement to build segments in Data 360 and activate them to Marketing Cloud and advertising channels.
Data teams: Salesforce data in Snowflake
Share unified profiles, engagement data and calculated insights into Snowflake for churn and propensity models in Snowpark, or AI enrichment with Cortex, without maintaining an export.
What Does The Snowflake And Data 360 Integration Cost?
Zero Copy removes pipeline and duplicate-storage costs, but it isn't free: costs land on both platforms. Data 360 consumes credits for federation and for rows shared and accessed, and Snowflake bills the warehouse compute that runs federated queries, plus its usual storage and data transfer. Salesforce's own FAQ makes the same point.
| Activity | Data 360 side | Snowflake side |
|---|---|---|
| Live query federation | Data Federation credit usage | Compute on the warehouse chosen in the connection |
| Cached federation | Data Federation credits; Salesforce notes that caching affects Data 360 costs | Compute when the cache refreshes from Snowflake |
| File federation | Data 360 credit usage; Data 360 reads the files with its own compute | No warehouse compute; storage stays in your object store |
| Sharing to Snowflake | Data Share Rows Shared and Sharing Rows Accessed usage types | Compute for the queries your team runs on the shared data |
Data 360 credits come as Data Services credits or as Flex Credits, which are shared with Agentforce and other features. Usage is measured per million rows processed, multiplied by a rate set for each usage type on Salesforce's rate card. Snowflake compute is billed per second, with a 60-second minimum each time a warehouse starts. Model both sides before go-live, and confirm current rates with Salesforce and Snowflake. Trailhead explains Data 360 credit consumption; Snowflake documents its cost model.
Limits And Pitfalls To Plan For (The Part Most Guides Skip)
These are the constraints that surface after the demo, when the integration meets real data volumes and real compliance reviews. Each one is documented by Salesforce or Snowflake, and each is cheaper to design around than to discover.
How Are Security, Permissions And Governance Handled?
Data stays governed where it lives. Snowflake encrypts all data at rest and in transit, and Zero Copy federation respects Snowflake policies such as row-level security and masking. On the Salesforce side, permission sets, data spaces, attribute-based policies and the Einstein Trust Layer control who and what can use the data.
- Permission sets changed in September 2025. Data Cloud Admin became Data Cloud Architect, Activation Manager and Activation Specialist were added, and older sets such as Data Aware Specialist are now legacy. Review assignments before you build.
- Use dedicated service identities. Key-pair or OIDC service users on the Snowflake side keep federation separate from human accounts and simplify audits.
- Grant least privilege in Snowflake. Connector creation, connector usage and database creation are separate privileges; assign them to roles deliberately.
- Keep consent flowing. Use the consent object that travels with every data share, and match deletion behavior between ingested, cached and federated data.
A Decision Framework For Your Snowflake And Data 360 Architecture
Getting The Implementation Right
Connecting the two platforms is the quick part. The design work around it takes longer: deciding which objects to ingest, federate or share, mapping external data to the data model, setting identity resolution rules, sizing the Snowflake warehouse, and agreeing who owns the credits and compute. Most of the difficulty sits between the Salesforce team and the data team, not inside either platform.
Twopir Consulting works with growing, mid-market, and enterprise organizations that need help with complex CRM implementations, integrations, and business system challenges. We implement, rescue, optimize, and integrate complex CRM and revenue systems, including Salesforce Data 360 and Agentforce data and knowledge integration, alongside integration services on MuleSoft and other platforms for systems that Zero Copy doesn't cover.
Plan Your Snowflake And Data 360 Architecture
On a 30-minute discovery call, a Twopir Salesforce architect will review where your customer data lives today, what Salesforce needs to act on, and whether to ingest, federate or share each part, including any legacy Snowflake target that needs to move to V2.
Schedule a Discovery CallThe Bottom Line
Snowflake and Salesforce Data 360 do different jobs, and Zero Copy is what makes them work as one customer data architecture. Snowflake stays the platform where data is stored and modeled; Data 360 makes the relevant parts actionable in Salesforce and for Agentforce; sharing returns Salesforce context to the data team. Get the ingest-versus-federate decisions, the cost model and the V2 migration right, and you avoid the pipelines this integration exists to replace.
- 1
Data Cloud is now Data 360. The capabilities are the same, but some screens and permission sets still use the old name.
- 2
Federation brings Snowflake data into Data 360; sharing exposes Data 360 objects in Snowflake. Neither copies the data.
- 3
Use the Snowflake V2 data share target (OIDC and Enrollment ID), and migrate any legacy target by October 31, 2026.
- 4
Choose live query, caching or file federation per object, and ingest what drives identity resolution and activation.
- 5
Budget for both platforms: Data 360 credits and Snowflake compute.
Frequently Asked Questions
Yes. Salesforce rebranded Data Cloud to Data 360 on October 14, 2025, and states that the functionality is unchanged. Some screens and permission sets, such as Data Cloud Setup and the Data Cloud Architect permission set, still use the old name, so you will see both names in the product and documentation.
Snowflake is a cloud data platform for storing, processing and analyzing enterprise data, used mainly by data teams. Data 360 is the Salesforce data layer that unifies customer data into profiles, insights and segments that Salesforce apps and Agentforce agents act on. Zero Copy lets each use the other's data without copying it.
No. Salesforce states that Data 360 is not a replacement for Snowflake and does not compete with it directly. Snowflake remains the platform for the broader data lifecycle, while Data 360 serves front-office users and AI agents. Most organizations that run both connect them with Zero Copy federation and sharing.
Zero Copy means no ETL job and no duplicate copy to maintain. With live query federation, results are processed in memory and discarded. If you enable caching, Data 360 keeps a local cache refreshed every 15 minutes to 7 days, and query results still travel between the platforms when they are read.
Both platforms carry a cost. Live and cached query federation run on the Snowflake warehouse selected in the Data 360 connection, so Snowflake bills that compute, and Data 360 consumes credits for federation. File federation over Iceberg tables uses Data 360 compute instead and avoids Snowflake warehouse charges.
Salesforce no longer lets you create a legacy Snowflake data share target, which used username and password based OAuth. Existing data shares on a legacy target must be migrated to a Snowflake V2 data share target, which uses OIDC and an Enrollment ID, by October 31, 2026. The migration is not automatic.
No. A data share can't include an external data lake object, or a data model object mapped to one. Data that Data 360 federates from Snowflake already lives in Snowflake, so other Snowflake consumers should read it there rather than through a Data 360 share.
Snowflake supports the integration for accounts in all AWS, Azure and Google Cloud commercial regions, on Standard, Enterprise or Business Critical edition. Virtual private Snowflake, government regions and China are not supported. Data 360 runs on AWS, and cross-cloud and cross-region connections are supported.
Twopir Consulting
Twopir Consulting helps growing and mid-market companies solve complex CRM, integration, and business system challenges. We implement Salesforce Data 360, Agentforce, Sales, Service and Marketing Cloud, and integrations on MuleSoft, Workato and Celigo for clients across the US, UK, Australia, the UAE and Canada.