SAP just admitted the real problem with enterprise AI was SAP
When SAP's chief technology officer says enterprise AI stalls not because the models are weak but because the data is locked in silos, you should pay attention. Especially when that company spent fifty years selling proprietary data formats as a feature.
SAP announced two acquisitions Monday that read as a single strategic confession. The German enterprise software giant agreed to buy Dremio, an open data lakehouse platform built around the Apache Iceberg open table format, and Prior Labs, a Freiburg-based company that builds Tabular Foundation Models. These are AI systems designed to make predictions on structured business data like payment delays, supplier risks, and customer churn, according to SAP News. Terms were undisclosed for both deals. The Dremio acquisition is expected to close in Q3 2026, pending regulatory approval; the Prior Labs deal timing has been described as Q2 or Q3 2026 in various reports. SAP committed more than one billion euros over four years to scale Prior Labs into a globally leading frontier AI lab for structured data, per RTTNews and t-online.
Independent analysts see the structural logic. "The data lakehouse is an established technology market," according to Forrester's Data Lakehouses Landscape, Q1 2026, with enterprises "increasingly prioritizing open data formats to avoid vendor lock-in and preserve long-term architectural flexibility." Constellation Research analyst Doug Henschen has noted that the lakehouse concept is "gaining advocates and vendor support" as organizations seek to consolidate workloads onto a single data platform. SAP is betting that the same architectural shift Forrester describes will accelerate demand for exactly the federation layer Dremio provides. Whether that bet pays off depends on execution — but the direction of the market supports the premise.
The headline acquisition is Dremio, and the headline problem is fragmentation. Most enterprise AI projects fail not because the models are inadequate but because the underlying data is scattered across proprietary systems, stripped of business context, and resistant to the kind of unified querying that AI agents require to operate at scale. SAP frames this as the core bottleneck in its own announcement. Dremio's lakehouse platform is serverless and elastic, built to federate queries across SAP and non-SAP data sources without the ETL pipelines that traditionally make this kind of integration a multi-month engineering project. After the acquisition, SAP Business Data Cloud will natively support Apache Iceberg as its foundation, meaning no data movement or format conversion when a customer wants to hook a Salesforce dataset onto an SAP HANA workload.
The competitive backdrop matters. Databricks and Snowflake both already support Apache Iceberg in their platforms. SAP is not leading the open lakehouse standard; it is arriving after competitors established the terrain. Dremio's value to SAP is not that it invented the approach but that it brings an existing enterprise customer base, including Shell, TD Bank, and Michelin, among those cited in the announcement, and a team that has been a primary contributor to Apache Iceberg, Apache Polaris, and Apache Arrow open source projects. GitHub activity across Dremio's open source repositories shows active commit maintenance through May 2026, with iceberg-auth-manager updated in November 2025 and dremio-oss refreshed in January 2026 — the cadence of a live project rather than a maintained-but-dormant acquisition target. Whether that gives SAP a genuine lead or merely catches it up to where the market already is depends on execution that hasn't happened yet.
There is a historical parallel worth considering. When Salesforce acquired Slack in 2021, it was a similar admission: an application-layer company acknowledging that its platform needed a data integration layer to stay competitive. Salesforce spent years afterward trying to make that acquisition pay off, with mixed results. The application layer has a way of reasserting itself. SAP is betting it can do what Salesforce struggled with — genuinely open the stack rather than just rebranding the integration layer. The evidence for that is thin.
Competitors are making analogous moves. Snowflake made a strategic investment in AtScale, a semantic layer services provider, in December 2025. Databricks has invested in AI agent companies and deepened its OpenAI partnership. Both are signaling that the semantic or data layer is where competitive advantage is shifting — which validates SAP's direction if not its specific execution. If SAP succeeds in turning Business Data Cloud into a genuinely open lakehouse, expect Oracle and Microsoft to feel pressure to respond, since neither has a direct equivalent. That is a signal worth watching.
The Prior Labs bet is cheaper and earlier-stage. Tabular Foundation Models are a real research area. The TabPFN family has been in development since 2022 under Frank Hutter's research group at the University of Freiburg, but the commercial track record is thin. TabPFN v2 handles datasets up to 10,000 samples. TabPFN-2.5, released recently, scales to 50,000 samples and 2,000 features, according to Prior Labs. One billion euros over four years is not large for a frontier AI laboratory build-out, which suggests SAP is acquiring early research capability rather than a proven production system. That is not necessarily wrong. The TabPFN approach of in-context prediction without traditional model training is genuinely different, but it carries significant execution risk.
What SAP is actually admitting, beneath the agentic AI framing, is that its proprietary data stack became a liability. The company that built its empire on integrated business software with tightly controlled data formats is now buying the antidote. The question for enterprise buyers is not abstract: Dremio's open lakehouse either becomes a genuine liberation of their data or a new lock-in vector wearing an open-source costume. The architecture supports the former. The incentives point the other way. Watch what SAP charges for once the acquisition closes.