SAP Completes Prior Labs Acquisition, Sealing Europe's First Corporate Frontier AI Lab

SAP SE announced on July 17 that it has completed its acquisition of Prior Labs, the Freiburg-based startup that pioneered Tabular Foundation Models — a category of AI built specifically for the structured, row-and-column data that underlies virtually all enterprise software. The deal, first announced in May 2026, closed just 18 months after Prior Labs was founded, and comes with a commitment from SAP to invest more than €1 billion over four years to build the lab into a globally leading frontier AI research institution.

The transaction marks a significant moment in European AI: for the first time, a major European enterprise software company has put meaningful capital behind a frontier AI research operation that it intends to keep genuinely independent — not absorbed into a product team, but maintained as a research lab with the freedom to publish, hire, and scale.

Why Tabular AI Is the Untapped Frontier

To understand why SAP paid up for Prior Labs, it helps to understand what the startup actually built — and why it is not what most people picture when they think about frontier AI.

The dominant narrative of AI in 2026 is built around language models: systems trained on text that can write code, summarize documents, answer questions, and increasingly take actions in the world. The benchmarks that define AI progress — coding accuracy, reasoning chains, mathematical proofs — are almost all text-based.

But the overwhelming majority of the world’s economically meaningful data is not text. It is tabular: rows and columns, databases, spreadsheets, ERP records, financial ledgers, supply chain logs. The data that actually runs Nestlé’s production schedule, HSBC’s loan portfolio, or Volkswagen’s parts inventory is not a document; it is a structured table. And language models, which are extraordinarily powerful on text, have historically struggled on tabular data.

Prior Labs was founded to fix that. The company’s TabPFN (Tabular Prior-Data Fitted Networks) model series, published in Nature, established state-of-the-art performance on tabular benchmarks across hundreds of independent academic studies. Where traditional machine learning approaches to tabular data require extensive feature engineering, manual tuning, and large training datasets for each new use case, TabPFN approaches the problem as a foundation model: pretrained on a vast meta-distribution of synthetic tabular tasks, it can generalize to new datasets with minimal examples.

SAP CTO Philipp Herzig called structured data “the greatest untapped opportunity in enterprise AI” — a claim that sounds modest only until you recall that SAP’s software processes roughly 87% of global commerce.

An Acquisition That Keeps Its Promise of Independence

The term “independent” gets deployed liberally when large companies acquire startups, and the independence usually evaporates within a year or two. SAP has structured the Prior Labs deal differently, at least on paper.

Prior Labs will continue to operate as a standalone entity under its existing founders: Frank Hutter, a machine learning professor at the University of Freiburg and one of the most cited researchers in the field; Noah Hollmann, who led the TabPFN work; and Sauraj Gambhir. The lab will retain its own brand, its own research agenda, and its own hiring authority — with SAP providing the capital to scale what was previously a research group with academic constraints into an institution with the compute, talent, and time horizon to compete at frontier level.

SAP’s €1 billion commitment over four years works out to €250 million per year — modest by U.S. AI lab standards, where OpenAI and Anthropic have raised in the tens of billions, but substantial for a European research institution and consistent with the scale of labs like DeepMind before its Google acquisition and IDRIS before it became part of France’s national AI strategy.

The key test of independence will come when Prior Labs’ research produces results that don’t fit neatly into SAP’s product roadmap — or, conversely, when SAP’s commercial pressures push toward applied work that a frontier lab wouldn’t otherwise prioritize. For now, SAP’s public commitment is to fund long-horizon research, not just applied engineering.

What Europe Has Been Trying to Build

The acquisition matters beyond the immediate business case because it represents something European AI policy has been trying to manufacture for years: a credible path from academic AI excellence to commercial frontier research, without requiring researchers to emigrate to San Francisco.

Europe has excellent AI research. The universities of Oxford, Cambridge, ETH Zurich, and the Technical University of Munich have produced foundational AI researchers at a rate commensurate with their American peers. But the commercial structures to retain those researchers in Europe — well-funded labs, competitive equity packages, access to frontier compute — have been conspicuously absent. The result has been a consistent talent drain: European researchers who want to do frontier work have had to choose between staying in academia with constrained resources or joining American hyperscalers and well-funded startups.

Prior Labs was itself a product of this dynamic: Frank Hutter founded it in Germany specifically to test whether a European frontier lab could compete on equal terms. The fact that SAP acquired it 18 months in — rather than waiting for it to either fail or migrate to the US — suggests the model works, at least when a European strategic buyer sees the value before a US acquirer does.

For EU policymakers who have spent years and billions of euros trying to build European AI champions, the SAP-Prior Labs deal offers a template that government procurement, grants, and venture funds have struggled to replicate: a well-capitalized corporate backer that actually understands the domain and has a credible distribution channel for the technology.

The Strategic Calculus for SAP

SAP’s motivation is not purely philanthropic. The company has roughly 400 million users, processes trillions of euros in business transactions annually, and sits at the center of the enterprise software market. Its customers — the largest manufacturers, banks, retailers, and governments in the world — are sitting on decades of structured tabular data that they cannot currently use with the language model tools that are absorbing most of the AI attention and investment.

If Prior Labs can extend the TabPFN approach to the scale and complexity of real enterprise datasets — where tables have hundreds of columns, billions of rows, missing values, schema drift, and domain-specific semantics — SAP would have a defensible AI capability that no language model vendor can easily replicate. Tabular AI embedded in S/4HANA, the company’s core ERP system, could analyze a customer’s supply chain anomalies, flag credit risks, or forecast demand without the data ever leaving the customer’s environment.

That combination — frontier research credibility, European regulatory alignment, and deep enterprise distribution — is precisely what SAP is paying €1 billion to buy. Whether it pays off depends on whether Prior Labs can move from research excellence to production-grade AI at enterprise scale, which is a harder problem than it sounds. But as bets on the future of enterprise AI go, it is one of the more structurally coherent ones placed in Europe this year.