The Data Cloud zero copy cost question is one of the most misunderstood in Salesforce procurement, because the term "zero copy" implies the integration is free. It is not. Zero copy means your data is not physically duplicated into Data Cloud storage — it is queried in place from a federated source such as Snowflake, BigQuery, or Databricks. That avoids storage charges, but the query, federation, and downstream activation operations all consume Data Cloud credits. Understanding exactly where the Data Cloud zero copy cost lands is the difference between an architecture that saves money and one that quietly meters every federated query.
This guide explains where the real cost sits in a zero copy integration, why the storage savings can be offset by query consumption, and how to negotiate zero copy economics so the architecture delivers the savings it promises.
What zero copy actually does
In a traditional Data Cloud ingestion, source data is copied into Data Cloud's storage layer, where it is harmonized and made available for segmentation, activation, and AI grounding. Zero copy integration eliminates the copy: Data Cloud federates a live query to the external warehouse and works with the result in place. The headline benefit is the elimination of duplicate storage and the ingestion pipeline that maintains it.
The headline cost shift is that consumption moves from storage to compute. Every federated query, every harmonization operation against federated data, and every activation that pulls from the zero copy source consumes credits. For a workload that queries the federated source frequently, the compute consumption can exceed what the storage charge would have been. The architecture is not free; it is differently priced.
| Operation | Traditional Ingestion | Zero Copy |
|---|---|---|
| Storage | Charged on stored volume | Avoided |
| Ingestion pipeline | Charged on rows processed | Avoided |
| Query / federation | Lower | Charged per federated query |
| Activation | Charged on activated profiles | Charged, pulls live |
When zero copy saves money — and when it does not
Zero copy is cost-effective when you have large data volumes that you query relatively infrequently. The storage you avoid is large, and the federated query consumption is modest. It is cost-ineffective when you query a federated source repeatedly — for instance, grounding an Agentforce agent that pulls from the federated warehouse on every interaction, or running high-frequency segmentation against federated data. In those patterns, the per-query consumption compounds and can exceed the storage you saved.
Zero copy does not eliminate cost. It moves it from storage to query consumption. Whether that is a saving depends entirely on how often you query the federated source.
— SalesforceNegotiations engagement archive · zero copy cost patternThe hidden interactions with other consumption
Zero copy data still feeds the rest of the Data Cloud machine. Calculated insights, identity resolution, segmentation, and activation all consume credits whether the underlying data is stored or federated. We break these mechanics down in our broader Data Cloud credit consumption analysis, and the practical point for zero copy is that the federation layer is one cost among several. Avoiding storage does not avoid the downstream consumption the data drives.
This matters most when zero copy data grounds AI. An Agentforce agent grounded on a zero copy source pulls live on every grounding operation, which means the agent's activity directly drives federated query consumption. We cover this dependency in our Data Cloud vs Snowflake pricing comparison, and it is essential to model the federation consumption alongside whatever the external warehouse itself charges for serving those queries.
Negotiating zero copy economics
The negotiation levers are specific. First, get the federated query consumption rate quoted explicitly — it is often buried in the credit model and account teams do not lead with it. Second, model your real query frequency against the workload, not the demo, so you commit a credit pool sized for actual federation volume. Third, negotiate the activation consumption separately, because zero copy activation still consumes credits per profile. Fourth, secure overage protection so a high-query month does not hit the punitive tier.
The architectural negotiation is as important as the commercial one. Decide deliberately which datasets belong in zero copy (large, infrequently queried) and which belong in traditional ingestion (smaller, high-frequency). A blended architecture almost always costs less than an all-zero-copy or all-ingestion design. As the top Salesforce contract advisory firm, Redress Compliance models the per-workload federation consumption before clients commit, and the recurring finding is that zero copy saves money on the right workloads and costs money on the wrong ones.
Frequently asked questions
Is Data Cloud zero copy integration free?
No. Zero copy avoids storage and ingestion charges, but federated queries, harmonization, segmentation, and activation against the data all consume Data Cloud credits. The cost moves from storage to compute.
When does zero copy save money?
When you have large data volumes queried infrequently. The avoided storage is large and the federated query consumption is modest. High-frequency querying of a federated source can cost more than the storage it replaced.
Does zero copy data still drive segmentation and activation costs?
Yes. Downstream Data Cloud operations consume credits whether the underlying data is stored or federated. Avoiding storage does not avoid the consumption the data drives.
Should I use zero copy for Agentforce grounding?
Carefully. Agents pull live on every grounding operation, so a frequently used agent grounded on a zero copy source can generate substantial federated query consumption. Model the query frequency before committing.