Concept card

SAP Snowflake Solution Extension — The Third Best-in-Class Runtime

As of 2026-07-24T14:00:00Z

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 Snowflake Solution Extension — The Third Best-in-Class Runtime?

SAP Snowflake Solution Extension is full, ungated Snowflake (Cortex AI, Snowpark, 90-day time travel) sold under SAP's single contract — collapsing the 3-6 month vendor-onboarding cycle for customers who've already cleared SAP procurement.

What it is

SAP Snowflake Solution Extension is SAP's commercial and technical arrangement that brings the full Snowflake platform into SAP Business Data Cloud as one of three "best-in-class runtimes" sitting inside the Intelligent Compute layer, alongside SAP HANA Cloud and SAP Databricks. Understanding what it is requires separating two questions that are easy to conflate: what changes commercially, and what changes technically — because the answer to the first is substantial and the answer to the second is, deliberately, almost nothing.

Commercially, a customer buying SAP Snowflake Solution Extension transacts through SAP's own contract framework rather than negotiating a separate Snowflake agreement: one purchase order, one support escalation path, one security and compliance review, priced on SAP's commercial terms rather than Snowflake's own list price. For an enterprise where SAP has already cleared procurement, legal and security review as a strategic vendor, that collapses what would otherwise be a multi-month vendor-onboarding cycle for an entirely separate platform into an extension of an existing, already-approved relationship. Technically, almost nothing about Snowflake itself is gated or stripped down by the SAP wrapper: the customer gets the full Cortex AI surface — text-to-SQL, document AI, model fine-tuning and serving — full data-sharing capability across the Snowflake Marketplace and cross-cloud shares including data clean rooms, time travel for point-in-time recovery, Snowpark for running Python, Java or Scala compute inside the warehouse, and Dynamic Tables for streaming transformations. None of that is SAP-specific tooling standing in for the real thing; it is genuinely Snowflake.

Why it matters

The commercial mechanism is specific: single PO, single support escalation, single security review, priced on SAP's framework not Snowflake list price.
Nothing on the technical surface is gated — Cortex AI, data sharing, 90-day time travel (EE), Snowpark, and Dynamic Tables are all full Snowflake, which matters because the pitch is procurement speed, not feature-parity risk.
The fit rule is explicit: pitch to customers with an existing Snowflake estate who also want the SAP-native semantic layer (Datasphere + SAC) over it — not to greenfield customers, who should pick one Intelligent Compute runtime instead.

Key points

Third best-in-class Intelligent Compute runtime under BDC, alongside SAP HANA Cloud + SAP Databricks — each optimised for a different workload pattern.
'Solution Extension' = SAP-resold, Snowflake-engineered: single PO, single support, single security review under SAP's contract framework.
Full Snowflake feature surface — Cortex AI (text-to-SQL, document AI, LLM serving), data sharing, time travel up to 90 days (EE), Snowpark (Python/Java/Scala), Dynamic Tables.
HANA Cloud wins for sub-second in-memory S/4 analytics; Databricks wins for Spark+ML engineering; Snowflake wins for elastic cloud warehouse + cross-cloud data sharing.
Pitch only when customer has existing Snowflake estate they won't retire AND wants Datasphere + SAC semantic layer on top; do not pitch to greenfield customers.
Solution Extension collapses the 3-6 month vendor-onboarding cycle for customers where SAP is already the approved vendor.
Snowflake data sharing and clean rooms enable multi-party analytics (e.g. retailer + supplier) that neither HANA Cloud nor Databricks match natively.
Storage/compute separation in Snowflake enables elastic cost scaling — relevant for workloads with spiky query demand (month-end reporting, seasonal retail).
Dynamic Tables in Snowflake provide declarative streaming transformation — the Databricks equivalent is Structured Streaming, which requires more engineering overhead.
Cortex AI text-to-SQL operates over Snowflake's own schema context; grounding it in SAP semantics requires the Datasphere virtual layer — the architecture requires both products working together.

Terms used on this page

Solution Extension: A SAP commercial model where a third-party product (Databricks, Snowflake) is resold under SAP's contract framework but technically operated by the underlying vendor; customer gets single-vendor procurement and the full third-party feature surface.
Intelligent Compute: The compute layer inside BDC's Knowledge Core — houses HANA Cloud (in-memory OLAP), Databricks (Spark+ML), Snowflake (cloud warehouse + sharing), plus AI Database and Data Warehouse blocks; each runtime is purpose-optimised, not interchangeable.
Snowflake Cortex AI: Snowflake's native AI service layer — covers text-to-SQL (SQL generation from natural language), document AI (summarisation, extraction), LLM fine-tuning, and model serving; fully accessible inside SAP Snowflake Solution Extension.
Snowpark: Snowflake's compute framework for running Python, Java, and Scala code directly inside the warehouse, enabling data engineering and ML workflows without moving data to an external compute cluster.
Dynamic Tables: Snowflake's declarative streaming transformation feature — define the target table's SQL logic; Snowflake continuously refreshes it as source data changes. Lower engineering overhead than Databricks Structured Streaming for standard ELT patterns.
Data clean room: A governed Snowflake environment where two parties can run joint analytics on their combined datasets without either party seeing the other's raw data — enables retailer+supplier, pharma+insurer, or bank+partner analytics under privacy constraints.
Storage/compute separation: Snowflake's architectural pattern where storage (S3/Azure Blob/GCS) is billed independently from compute (virtual warehouses); compute can scale to zero when idle and burst to multiple warehouses in parallel — enabling elastic cost management for spiky workloads.
Best-in-class runtime: A purpose-built compute engine optimised for a specific data-processing pattern: HANA Cloud for in-memory analytics, Databricks for Spark + ML, Snowflake for cloud warehouse + Cortex AI + data sharing.

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 · the facts worth quoting.

Open in the app →