Buying the lakehouse was the only realistic path. SAP launched Business Data Cloud in February 2025 with Databricks as its data engineering partner, and the platform has been quietly accreting capability through acquisition ever since. Reltio in March for master data. Now Dremio, announced today, for the open table foundation. The pattern is consistent. SAP is not building Business Data Cloud organically. SAP is assembling it.

Terms of the Dremio deal were not disclosed. The transaction is expected to close in the third quarter of 2026, subject to regulatory approval. SAP also announced a separate agreement to acquire Prior Labs, a Freiburg-based startup specializing in Tabular Foundation Models, with more than one billion euros pledged in development investment over four years. Two acquisitions on the same day, both targeting structured business data, both subject to regulatory approval.

A strategic error, now being corrected

The fast pace of AI is showing a troubling trend. The focus has been on the application layer of AI rather than a robust data and infrastructure layer. A lot of that foundation layer is open source. Apache Iceberg is an open-source table format designed for large-scale analytical datasets stored in data lakes, a kind of bridge between raw data files and analytical tools. SAP did not develop Apache Iceberg capability on its own. That was a mistake it is now rectifying with the Dremio acquisition. The deal will enable SAP to move faster than developing this capability themselves, and they should have done this earlier rather than waiting until 2026.

I shared a version of this argument with CIO.com's Evan Schuman earlier today. The longer version, the one that did not fit in the article, is that Iceberg has emerged as the open table format that enterprise customers expect their lakehouse to speak. Databricks recognized this in 2024 when it acquired Tabular, bringing the original Iceberg creators in-house. Snowflake responded with Apache Polaris as an open catalog. Every serious data platform vendor has had to declare an Iceberg position.

Until today, SAP did not have one. Business Data Cloud relied on partner integrations and zero-copy data sharing with Databricks and Snowflake to span SAP and non-SAP data. That works as a partnership story. It does not work as a platform story when the platform is supposed to be the foundation for agentic AI workloads that span every system in the enterprise.

The federation angle is what makes this acquisition different

Dremio is not just a lakehouse. It is a federated query engine that lets data stay where it is. That distinction matters more for SAP customers than it would for most other vendors' customers, because most large SAP estates are not clean centralized data environments. They are brownfield landscapes. SAP data, non-SAP data, legacy warehouses, departmental lakes, regional repositories, acquired systems, partner data, and industry-specific platforms.

Telling those customers that AI readiness begins with moving everything into one central platform is good for the vendor but expensive for the buyer. Dremio gives SAP a more pragmatic story. Keep more of your data where it is, access it faster, apply consistent catalog and semantic controls, and bring it into Business Data Cloud and AI workflows without forcing a major migration program upfront.

This matters most in manufacturing and other regulated verticals where customer chief information officers have refused for years to move data to the cloud. SAP has been pushing those customers toward S/4HANA Cloud and Business Data Cloud with limited success in the data layer specifically. Dremio gives SAP a face-saving way to reach that data without requiring it to migrate. The acquisition unlocks a category of SAP customer data that was previously out of reach.

Open commitment, scrutinized

SAP stated it will continue to invest in and prioritize Dremio's open-source contributions to Apache Iceberg, Apache Polaris, and Apache Arrow. That commitment deserves the same scrutiny every vendor's open-source pledge deserves at acquisition. Open formats and open architectures are different things. A customer can have data in Apache Iceberg format inside Business Data Cloud and still find that the operational tooling, agent workflows, semantic layer, and governance pipelines are not portable to another platform.

SAP will have to commit resources to help Dremio continue contributing to open-source Apache Iceberg, Apache Polaris, and Apache Arrow. We do not yet know the specifics. The press release language is one sentence. No funding commitment, no headcount commitment, no governance commitment, no timeline. Customers should ask for three specifics before regulatory close in the third quarter. First, named retention of the current Dremio committers and project management committee members. Second, what percentage of Dremio engineering effort will continue to flow to upstream Apache projects versus closed SAP-specific development. Third, a governance commitment that the universal catalog SAP builds for Business Data Cloud will keep its core APIs aligned with the Apache Iceberg REST Catalog standard, with no proprietary forks.

The Databricks-Tabular precedent is instructive. After that acquisition, Iceberg as a community standard continued to advance, including the V3 release. Iceberg as Databricks implements it inside Unity Catalog accumulated platform-specific features. Customers benefit from format compatibility but the catalog and the engine became more proprietary, not less. SAP is now positioned to follow the same pattern, and the same pattern would be a defensible business outcome that nonetheless erodes the openness premise customers heard at acquisition.

The Prior Labs signal matters more than the headlines suggest

The same-day Prior Labs acquisition is the clearer strategic statement. SAP is committing more than one billion euros over four years to develop Prior Labs into a research lab focused on Tabular Foundation Models. These are AI models built specifically for structured business data: tables, numbers, statistics, the kind of data that runs financial close, supply chain, and procurement. Large language models do not handle this data well. Prior Labs builds models that do.

SAP CTO Philipp Herzig said as much during the press conference Monday morning. LLMs do not deal well with numbers and they struggle with structured data. The practical difference shows up when systems try to predict the future rather than analyze the past, such as how well a product will sell over the next ten months, or which payments will be delayed. Tabular Foundation Models are built for those tasks.

SAP already shipped its own tabular foundation model, SAP-RPT-1, earlier this year. The Prior Labs acquisition extends that bet. Prior Labs' TabPFN-2.6 model currently leads the TabArena benchmark for tabular foundation models. The combination signals that SAP believes the productivity ceiling for enterprise AI is not raised by getting better at unstructured language tasks. It is raised by getting better at the structured data that actually drives business decisions.

What this means for the Databricks partnership

Business Data Cloud launched as a co-engineered offering with Databricks. The partnership remains in place. But the architectural overlap just expanded considerably. Databricks brought lakehouse, Iceberg via Tabular, and Unity Catalog. SAP now brings lakehouse, Iceberg via Dremio, and an Apache Polaris-based catalog. The two companies are now competitors at the lakehouse layer of the same platform they jointly market.

For information technology executives with active Snowflake and Databricks contracts this morning, nothing changes in the next two quarters. By the first half of 2027, expect SAP to steer net-new AI workloads toward Business Data Cloud regardless of what the partnership press releases say today. The chief information officers who plan for that trajectory now will negotiate from strength.

Compute and storage that data warehouse vendors provide is rapidly becoming a commodity. The defensible value in enterprise AI is migrating up the stack to the semantic layer, the catalog, the lineage graph, and the business context that lets an agent know what "active customer" means within an organization. SAP just bought the toolkit to own that layer for any company running SAP at the core.

Filling the gaps

SAP is now filling all the gaps through a mix of acquisitions and partnerships. The recent acquisition of Reltio filled the master data management gap as a feature acquisition. Dremio fills the lakehouse foundation gap. Prior Labs fills the structured-data model gap. The open question is whether SAP can integrate three acquisitions across three architectural layers without losing the openness premise that made each of the acquired companies worth buying in the first place.

Three acquisitions in roughly six weeks. Three different layers of the agentic AI stack. One consistent strategy. SAP is not adding capabilities to Business Data Cloud opportunistically. SAP is executing a deliberate sequence to make Business Data Cloud the credible third platform alongside Databricks and Snowflake, with a structural advantage in business application context.