SAP to Acquire Dremio: Agentic AI Meets Iceberg‑Native Enterprise Lakehouses

SAP buys into the Iceberg-era lakehouse for agent workloads

On May 4, 2026, SAP SE and Dremio announced that SAP has agreed to acquire the open lakehouse vendor to deepen SAP Business Data Cloud and accelerate agent-ready analytics across SAP systems and broader enterprise data estates. Deal terms were not disclosed, and the companies expect the transaction to close in Q3 2026, subject to regulatory approvals.

This is not merely a tooling tuck-in. Across thousands of ERP and line-of-business rollouts, the limiting factor for credible AI assistants is rarely “better prompts.” It is whether models can access consistent, timely, governance-grade context without forcing brittle ETL choreography. SAP framed the acquisition precisely around that bottleneck: consolidating governed context on open table formats agents can reliably query.

What Dremio brings to SAP’s stack

Dremio promotes an Iceberg-centric execution model emphasizing federated access, semantic layers, and serverless elasticity. SAP’s announcement positions the acquisition as foundational for making SAP Business Data Cloud an Apache Iceberg–native lakehouse capable of spanning SAP-managed and third-party datasets without obligatory format conversion pipelines.

Architecturally, the promise is pragmatic: unify governed business meaning (entities, hierarchies, access rules, lineage) with an open interchange layer that multiple engines—not only SAP-built services—can target. CTO Philipp Herzig is quoted stressing that enterprises stall because data is fragmented, proprietary, or stripped of operational context—not because frontier models stalled.

Open catalog semantics and downstream agent risk

Roadmapped capabilities include an open catalog built on Apache Polaris and the Iceberg REST catalog API—positioned as a discovery and semantic backbone for downstream models and agents pulling from heterogeneous sources. Translation: fewer shadow copies when teams wire agents into finance, procurement, HR, supply chain, and customer systems.

Nevertheless, consolidating catalog and federation power inside a mega-vendor umbrella sharpens procurement and portability questions:

• Vendor neutrality: Open formats help; commercial packaging and SKU bundling still influence roadmap velocity and egress economics.
• Agent blast radius: Broader, faster query surfaces require tighter policy engines, deterministic tool contracts, and step-level lineage for automated actions—not just analytics.
• Skills mix: Success depends on stewards who understand Iceberg housekeeping, Polaris governance, SAP object models and LLM/agent orchestration—not one silo.

Parallel acquisition signals depth of SAP’s AI bet

Industry reporting notes SAP concurrently announced acquiring Prior Labs (tabular foundation model specialists) alongside the Dremio transaction, reinforcing that SAP intends to integrate model-facing IP—not only infrastructure—with its data cloud narrative. CIOs evaluating SAP’s pitch should insist on proofs that combine heterogeneous enterprise sources, realistic permissioning, measurable agent task completion—not slide-deck federation diagrams.

Snap takeaways for business and platform leaders

1. Standardize meaning before scaling agents: Invest in catalogs, lineage, and entitlement models that persist beyond any single Copilot SKU.
2. Prioritize reversible agent actions: Read-heavy analytics workflows differ materially from transactional automations touching core SAP objects.
3. Watch close milestones: Regulatory cycle through expected Q3 2026 closing will influence multi-year modernization bets.

If SAP executes, Business Data Cloud could become one of the more credible hyperscale-aligned answers to enterprise “agent + ERP context” fragmentation—provided customers demand open interoperability proofs, not slogan-level lakehouse rebrands.