On May 4, 2026, SAP announced it would acquire two companies in a single move: Dremio, an open data lakehouse platform, and Prior Labs, a tabular AI research company. The deals are separate, both pending regulatory approval, and together they represent SAP’s clearest statement yet about where enterprise AI actually breaks down.

It breaks down in the data, not the models.

What SAP Just Bought

Dremio is a high-performance, Apache Iceberg-native data lakehouse built for analytical workloads. It runs in a serverless, elastic architecture — scaling up when demand spikes, scaling back down when it subsides — and it connects to data wherever it lives, without forcing it into SAP’s proprietary formats. The platform is already used by companies that need to run real-time analytics across a mix of SAP and non-SAP systems.

Prior Labs is less well known but arguably the more interesting acquisition. The company, founded by former researchers at Google and DeepMind, specializes in tabular AI — models purpose-built to work with structured enterprise data like spreadsheets, ERP records, and operational databases. Most AI models are trained primarily on text. Prior Labs’ models are designed to understand the kind of row-and-column data that actually runs businesses.

Taken together, the acquisitions give SAP an open lakehouse for storing and querying enterprise data, and AI models specifically trained to reason over it.

Why This Matters More Than Another Platform Deal

Enterprise AI projects fail at a predictable point. It is not the model selection. It is not the prompt engineering. It is the moment when someone asks where the data is, and the answer is: spread across a data warehouse, three SaaS tools, two legacy ERP tables, and a set of spreadsheets no one has cleaned since 2019.

SAP’s CTO, Philipp Herzig, put it directly in the announcement: “Enterprise AI doesn’t stall because the models aren’t good enough; it stalls because the data isn’t ready for AI agents. Dremio eliminates that bottleneck. Combined with SAP Business Data Cloud, we can now take customers from raw, fragmented data to governed, AI-ready intelligence on a single open platform.”

That is a more honest framing than most enterprise AI marketing. The bottleneck is data fragmentation, not model capability. SAP is betting that whoever solves that problem at scale controls the enterprise AI stack.

The Apache Iceberg Bet

The Dremio acquisition is also a bet on Apache Iceberg as the open standard for enterprise data storage. Iceberg is a table format that allows different tools to read and write the same data without converting it into proprietary formats. When SAP integrates Dremio, its Business Data Cloud becomes Iceberg-native by default.

This matters for companies running mixed environments. If your operational data lives in SAP but your analytics run in Databricks or Snowflake, Iceberg lets those systems share data without constant ETL pipelines. SAP is positioning itself as the open hub for that data — not the data silo it has historically been.

For data teams, this reduces the friction that makes enterprise AI projects expensive before they start. You spend less time moving data between systems and more time building on top of it.

What the Prior Labs Piece Adds

The tabular AI angle is worth watching. Most general-purpose AI models — including the large language models powering most enterprise AI tools right now — are weak on structured, numeric, and relational data. Ask a language model to analyze a financial table with hundreds of rows and complex cross-references, and you will quickly hit the limits of what it was trained to do.

Prior Labs’ models are built for exactly that kind of data. They can reason over enterprise tables, identify patterns across structured records, and produce outputs that treat rows and columns as first-class inputs rather than text to be parsed.

Combined with Dremio’s ability to surface that data at scale, SAP is building a system where agentic AI can operate across real enterprise data — not just documents and emails, but the financial, operational, and supply chain records that actually drive decisions.

What This Means for Business

If you run SAP: Your data environment just became a more credible foundation for AI agents. The Dremio integration will reduce the integration work required to get AI working across your full data estate, including non-SAP systems you have been running in parallel.

If you don’t run SAP: The competitive pressure will move. Databricks, Snowflake, and Microsoft Fabric are all making similar moves. Every major data platform is racing to become the AI-ready layer for enterprise data. The stakes for getting your data foundation right are rising.

If you are a data team leader: The business case for cleaning up fragmented data just became stronger. The companies that can give their AI agents access to clean, unified, governed data will deploy agents that actually work. The ones that can’t will keep running proof of concepts that never reach production.

The data readiness problem is not new. What is new is that the cost of ignoring it is now visible in every enterprise AI project that stalls at the data layer. Gartner’s own research backs this up directly: organisations with successful AI initiatives invest up to four times more in their data and analytics foundations than those that do not.

The Harder Question SAP Still Has to Answer

Both deals are pending regulatory approval and expected to close in Q2 and Q3 of 2026. Integration timelines for enterprise software acquisitions are rarely short, and the history of large ERP vendors acquiring data companies is mixed.

SAP has made enterprise data acquisitions before — including Sybase in 2010 and BusinessObjects in 2007 — with varying degrees of integration success. The question is whether Dremio stays genuinely open and Iceberg-native inside SAP, or whether it gradually becomes another proprietary extension of the SAP platform it was supposed to bridge.

The Dremio founders’ statement suggested optimism: they expect the acquisition to “accelerate our agentic vision” and allow them to incorporate SAP technologies. Whether that vision survives the integration intact is the part worth watching.

What This Signals for the Market

More broadly, this move confirms a trend that Enterprise DNA has been watching across our network of data professionals and business leaders: the AI model race is largely settled for most enterprise use cases. The differentiation is now happening at the data layer, the integration layer, and the governance layer.

The companies and data teams that invest in data infrastructure today are not just preparing for better analytics. They are building the foundation that will determine how well their AI agents perform when the models are good enough and the infrastructure is not.

For the 220,000-plus data professionals in the Enterprise DNA community, this is a useful signal: the skills that matter most right now are not just model fine-tuning or prompt engineering. They are data modeling, governance, lakehouse architecture, and the ability to build data pipelines that AI agents can actually use.

The data layer is where the work is. SAP just spent a significant amount of money to confirm it.