Concept card

SAP × Dremio — Federated Data for Agentic AI

As of 2026-07-23

Edited by Cédric-Olivier Mâry, sole editor and publication director — 27 years in the SAP ecosystem. How these pages are produced, and where automation is used.

What is SAP × Dremio — Federated Data for Agentic AI?

SAP's May 2026 Dremio acquisition gives Joule agents zero-copy SQL access across S3, Snowflake, and Databricks, without pre-building a BDC Connect federation point per source.

What it is

Dremio is an open-source-rooted query engine built around Apache Arrow Flight and Apache Iceberg that executes SQL across heterogeneous data sources — S3, ADLS, GCS, Snowflake, Databricks Delta, RDBMS — without copying data into a proprietary store. It positions itself as an 'agentic lakehouse' because each query is a federated plan that pushes predicates to sources and assembles results in-memory via Arrow columnar buffers, making it well-suited to the retrieval step in agentic AI workflows where an agent must query multiple sources in a single reasoning cycle. SAP announced acquisition of Dremio in May 2026 to supply the agentic AI layer — primarily Joule agents — with unified query access to the customer's full heterogeneous data estate without requiring data movement into BDC or Datasphere first.

Why it matters

Today a Joule agent investigating a procurement anomaly can query SAP-native Datasphere data but is blocked from a Snowflake spend warehouse or an S3 supplier-risk feed unless an architect pre-builds the connector.
Dremio removes that pre-build step — the agent issues SQL, Dremio resolves it across registered sources at runtime and returns Arrow-format results.
This directly extends agentic reasoning cycles to evidence that lives outside SAP's own estate, previously a hard capability wall.

Key points

Dremio = federated SQL engine on Apache Arrow Flight + Iceberg; announced May 2026 (acquisition pending Q2/Q3 close); 12–24 month integration horizon, not yet GA.
Zero-copy: SQL runs across S3, Snowflake, Databricks, RDBMS without staging data — Arrow Flight returns columnar results ~10× faster than JDBC.
Agentic-critical property: ad-hoc SQL federation at agent runtime without catalog pre-definition — agents select sources based on task context.
Three-tier architecture target: hot (Datasphere HANA <500ms) · warm (Dremio federation 1–10s) · cold (BDC Delta batch 10–30s), all under unified BDC Catalog DAC.
Iceberg multi-catalog broker: bridges AWS Glue + Unity Catalog + Polaris simultaneously — critical for multi-cloud customers.
Not the same as BDC Connect federation — Dremio adds runtime source discovery, Arrow streaming, and broader source heterogeneity.
Current GA for SAP non-SAP federation: BDC Connect (Snowflake, Databricks, S3/Iceberg, BigQuery). Dremio = roadmap.
SAP × Dremio — Federated Data for Agentic AI is mastered only when it changes a named buyer decision.
Start with the semantic contract and control model before demonstrating the tool.
Use current SAP, analyst, study, KG, and news signals as evidence, not decoration.

Terms used on this page

Apache Arrow Flight: A binary-efficient RPC framework built on columnar Arrow format for high-throughput data transfer — ~10× faster than JDBC for analytical result sets.
Apache Iceberg: Open table format for large analytic datasets — supports schema evolution, time-travel, and multi-engine access (Spark, Flink, Dremio, Trino, Databricks, BDC).
Agentic lakehouse: Dremio's self-description: a lakehouse architecture where AI agents can query the full data estate without pre-built catalog definitions, powered by runtime federated SQL.
Catalog reflection: Dremio metadata layer describing available sources, schemas, row counts, and freshness — queryable by planning agents for source discovery.
Predicate pushdown: Query optimisation where filter conditions are sent to the source system to execute before data leaves — reduces network transfer and source-side scan cost.
BDC Connect: SAP's current federation mechanism in BDC — requires catalog pre-registration; supports Snowflake, Databricks, S3/Iceberg, BigQuery, OData. Dremio complements and extends this.
Decision owner: The accountable person who accepts the trade-off and funds the next action.
Semantic contract: The shared definition of business terms, metrics, entities, and access rules used by tools and teams.

Sources

Full card available to members. What the full card adds: the full decision framework · the SAP vs Snowflake / Databricks / Fabric comparison · the common pitfalls and their fix · the cheat sheet · the architecture schemas · the code blocks.

Open in the app →