SAP is spending roughly €1 billion to acquire two companies it believes can solve one of the most persistent headaches in corporate technology: getting AI to work on the messy, fragmented data that actually runs a business. The German software giant announced in May 2026 that it plans to buy Dremio, a data lakehouse platform, while continuing to fold in Prior Labs, an AI research outfit whose tabular foundation model is already being tested on SAP’s own enterprise datasets. Together, the deals represent SAP’s biggest bet yet that the future of business software belongs to AI agents that can pull insights from scattered systems and act on them.
What SAP is buying and why
Dremio, founded in 2015 and based in Santa Clara, California, built its reputation on an open-source query engine that lets companies run analytics across data lakes, warehouses, and cloud storage without moving or copying the underlying data. For SAP, the appeal is straightforward: most large enterprises run SAP for core operations like finance, procurement, and supply chain management, but huge volumes of relevant data sit outside SAP systems entirely, in Salesforce instances, cloud data lakes, IoT platforms, and legacy databases. Dremio’s architecture is designed to bridge exactly that gap.
According to SAP’s acquisition announcement, the goal is to “unify SAP and non-SAP data to power agentic AI.” In practical terms, that means giving SAP’s AI agents a single, queryable view of an enterprise’s full data estate, not just the slice that lives inside SAP’s own products. If a procurement agent needs to cross-reference supplier performance data in SAP S/4HANA with shipping telemetry stored in a cloud data lake, Dremio’s lakehouse layer would handle the connection.
Prior Labs, meanwhile, brings a different piece of the puzzle. The German AI research group developed what SAP now calls SAP-RPT-1, a large tabular foundation model trained on structured enterprise data. A preprint paper hosted on arXiv describes the model’s ability to predict business process outcomes and classify operational events using in-context learning, outperforming traditional machine learning baselines in controlled experiments on SAP datasets. Where large language models like GPT-4 excel at processing text, SAP-RPT-1 is purpose-built for the rows-and-columns world of ERP systems: purchase orders, invoice line items, production schedules, and HR records.
The strategic logic connecting the two acquisitions is that Dremio provides the data access layer while SAP-RPT-1 provides the AI reasoning layer. One gathers the information; the other understands it.
The €1 billion price tag
SAP has not disclosed official purchase prices for either deal. The combined figure of approximately €1 billion comes from reporting by outlets including Reuters and Bloomberg, which have cited people familiar with the transactions. The Dremio acquisition alone has been reported at roughly $1 billion, while Prior Labs was the subject of a separate, smaller deal reported at around €200 million. SAP’s investor relations materials have not published a breakdown, and no purchase price allocation or goodwill estimate has appeared in regulatory filings.
SAP’s most recent annual report, a Form 20-F filed with the SEC for fiscal year 2025, discusses acquisition-related risks in detail, including integration costs, management distraction, and the possibility that anticipated synergies may not materialize. The filing predates both deals but establishes the regulatory framework SAP operates under when pursuing transactions of this size. Investors should expect more specific financial disclosures once the Dremio deal closes, assuming it clears customary regulatory approvals.
The Dremio announcement itself uses the phrase “intends to acquire,” indicating the transaction has not yet closed. No timeline for regulatory review or completion has been made public.
Where SAP fits in the enterprise AI race
SAP is not the only major software company trying to merge data platforms with AI. Microsoft has been building out Fabric, its unified analytics platform, and embedding Copilot agents across its business applications. Databricks, which pioneered the lakehouse concept Dremio also uses, raised $10 billion in late 2024 at a $62 billion valuation and has been aggressively expanding its own AI capabilities. Snowflake launched Cortex, its AI and machine learning layer, to let customers run models directly on data stored in its cloud warehouse. Google has been integrating Gemini models with BigQuery. Oracle has been weaving generative AI into its cloud ERP suite.
What distinguishes SAP’s approach is its starting position. SAP’s ERP systems process roughly 87% of global commerce, according to the company’s own figures. That installed base gives SAP something its competitors lack: direct access to the transactional data that records how goods move through supply chains, how invoices get paid, and how workforces are managed. The challenge has always been that this data is locked inside rigid, highly structured systems that were not designed for the kind of flexible, cross-source analytics that modern AI demands.
Dremio’s open architecture and SAP-RPT-1’s tabular reasoning are meant to address that limitation directly. Rather than asking customers to export data into a separate analytics environment, SAP wants to bring the AI to the data, wherever it sits.
What SAP-RPT-1 actually does
The SAP-RPT-1 model, detailed in the arXiv preprint, is a foundation model designed specifically for tabular data, the kind of structured, column-and-row information that fills enterprise databases. Unlike general-purpose large language models, which are trained primarily on text scraped from the internet, SAP-RPT-1 was trained on structured enterprise datasets and evaluated on tasks like predicting whether a business process will complete successfully, classifying the type of event occurring in a workflow, and forecasting operational outcomes.
The preprint reports that SAP-RPT-1 using in-context learning, where the model is given a few examples and asked to generalize, outperformed traditional machine learning approaches such as gradient-boosted trees on several SAP-specific benchmarks. That is a notable claim because gradient-boosted models have long been considered the gold standard for tabular prediction tasks.
However, the paper is a preprint. It has not been peer-reviewed, and no independent team has published a replication study. The experiments were conducted on curated SAP datasets by SAP-affiliated researchers. Whether the same performance gains hold up in messy, real-world production environments with incomplete data, inconsistent formatting, and legacy system quirks remains an open question. No customer case studies or production benchmarks have been published.
Risks and open questions
Several significant uncertainties hang over both deals.
Integration complexity. Merging Dremio’s open-source, cloud-native architecture with SAP’s proprietary ecosystem is a substantial engineering challenge. SAP’s own SEC filings warn that acquisitions can strain internal controls and dilute margins if synergies are overestimated. Dremio’s existing customers, many of whom chose the platform precisely because it is vendor-neutral, may resist tighter coupling with SAP’s stack.
Talent retention. Acquisitions of research-driven companies like Prior Labs live or die on whether key researchers stay. No public details have emerged about retention agreements, the degree of autonomy Prior Labs’ team will maintain, or how quickly their work will be folded into shipping SAP products.
Competitive response. Databricks, Snowflake, and Microsoft are not standing still. Each has its own AI-plus-data strategy, deep pockets, and large customer bases. SAP’s advantage is its ERP installed base, but that advantage erodes if customers find it easier to pipe their SAP data into a competitor’s lakehouse than to wait for SAP’s integrated offering to mature.
Timeline to value. SAP has outlined a vision of AI agents operating across unified data estates, but no concrete product roadmap, general availability date, or integration milestone has been shared publicly. Enterprise customers planning their own AI strategies need specifics, not architectural diagrams.
Regulatory approval. The Dremio deal has not closed. Cross-border acquisitions of data infrastructure companies can attract scrutiny from regulators in the U.S., EU, and elsewhere, particularly given heightened attention to AI-related transactions.
What this means for SAP’s customers
For the roughly 300,000 companies that run SAP systems, the Dremio and Prior Labs acquisitions signal a clear strategic direction: SAP wants to become the platform where AI agents interact with enterprise data, not just the system of record where that data is stored. If the integration succeeds, customers could gain the ability to deploy AI agents that understand structured business data natively and can reach across SAP and non-SAP sources without requiring complex data pipelines or third-party middleware.
That is a compelling promise. It is also, for now, exactly that: a promise. The press release confirms SAP’s intent. The arXiv preprint shows the research is real. The SEC filing reminds us that acquisitions of this scale carry meaningful risk. Until closing confirmations, product launches, and independent benchmarks arrive, the story of SAP, Dremio, and Prior Labs is best understood as a credible, well-funded wager that the company can turn decades of enterprise data dominance into an AI advantage. The direction is clear. The execution is what will determine whether it pays off.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
