Eighteen months after an initial seed of €9 million, the Freiburg-based pioneer of TabPFN is being acquired by SAP and will raise more than €1 billion within four years. Terms are not disclosed; the strategic intent is unmistakable.
When Frank Hutter, Noah Hollmann, and Sauraj Gambhir founded Prior Labs in early 2024, the world of artificial intelligence was focused on almost everything except the kind of data their company was built around. Language models attracted capital, attention, and graduate students.
Spreadsheets, spreadsheets, and the structured records that actually drive businesses were not part of the AI conversation that was supposed to create a frontier lab. The founders announced on Monday that this hypothesis is about to be tested.
Previous Labs has signed a definitive agreement to be acquired by SAP, with the German software giant committing more than €1 billion over the next four years to transform the startup into a leading global frontier artificial intelligence lab.
It is by some distance the most ambitious venture AI research investment ever made in Europe by a European company.
The terms of the operation were not disclosed. SAP’s own announcement Prior describes Labs as a pioneer of Tabular Foundation Models and views the acquisition as a continuation of that work. SAP itself started with a model called SAP-RPT-1, which was developed by a larger company before most of the enterprise-software world recognized the category.
Prior Labs will continue to operate as an independent legal entity, retaining its brand, headquarters in Freiburg, offices in Berlin and New York, open source commitments, customer relationships and existing scientific advisory board including Yann LeCun and Bernhard Schölkopf.
The deal remains subject to regulatory approval and is expected to close in the second or third quarter of this year.
In a joint blog post, the founders chose the unusually direct phrase “next chapter” to describe what’s next.
What did the previous Labs actually build?
The work of the contract is contained in the technical notes. Previous Labs flagship model, TabPFN was published in the journal Nature in early 2025 and according to the founders, it has had more than 1,000 references and more than three million downloads since its release.
The newest generation, TabPFN-2.5, extends the architecture to a dataset of 50,000 samples and 2,000 features, and according to Prior Labs’ published benchmarks, it is currently at the top of TabArena, the standard benchmark for tabular machine learning.
Most importantly, TabPFN does something that most foundational models can’t. It operates in a single forward pass without task-specific training and matches or exceeds the accuracy of tuned tree-based models, including AutoGluon configurations running for hours.
It is the technical inflection point for the types of data sets prevalent in enterprise systems, customer records, financial transactions, manufacturing telemetry, clinical trial results.
A general-purpose model that does not require domain-specific retraining changes both the cost and deployment schedule of structured data AI.
It’s more than just the abstract concept of the AI lab that SAP bought. SAP’s enterprise customer base is at the heart of Prior Labs’ modeled categories: financial services, healthcare, manufacturing and industrial. Constellation Research’s analysis of the deal It describes it as part of a broader data platform strategy that SAP announced the same week, along with its parallel acquisition of Dremio.
The structured data layer is a part of enterprise AI that most companies still can’t do well; SAP has now purchased one of the most reliable attempts to fix this.
Previous Labs’ only previous financial event a It was announced in February 2025 before a €9 million splashIt is managed by Balderton Capital with participation from XTX Ventures, Hector Foundation, Atlantic Labs and Galion.exe.
Eighteen months later, the Balderton bet produced one of the cleanest exits the firm has had in its 19-year history. The exact economics won’t be revealed until SAP’s regulatory filings arrive for the announcement, but the multiple from the €9m round to the €1bn+ post-acquisition investment program is exceptional on any sober reading.
Why is this important for European AI?
Europe’s AI policy has spent the past two years trying to create the outcome that the Prior Labs deal reflects. TNW is following a broader effortThis included sovereign-cloud rewards, an AI Act compliance regime and an increasingly overt push by European technology leaders to deploy capital at scale within the EU rather than exporting it west.
The CEO of Mistral was particularly vocal To the question of whether Europe can build, rather than rent, the AI infrastructure that runs its own economies.
The Prior Labs deal is an obvious answer, at least on one reading. SAP, is currently the most valuable listed company in Europeuses its market capitalization to fund a frontier research lab on the continent, anchors a European founding team, is headquartered on a university campus in Germany, and focuses on the precise category of AI best suited to the operational businesses that SAP serves.
The structure, an autonomous, well-capitalized, open-source research lab within a strategic corporate parent, is closer to the Anthropic-Google or Mistral-Microsoft model than anything European enterprise software has previously attempted.
It’s also indicative of Europe’s AI talent. TNW previously reported The fact that Europe has more AI researchers than the US in absolute numbers, but turns fewer of them into commercial results.
Prior Labs is an academic team from Freiburg that, contrary to this pattern, developed a Nature paper in two years, an open source model with a strong community, a reliable commercial business and a billion euro deployment partner.
Whether this is the pattern or remains the exception is a question European politicians have been asking for several years. Monday’s announcement is the most positive empirical evidence yet.
There are several. The first is integration: large enterprise software companies have a mixed track record of maintaining the autonomy of acquired research labs.
SAP made it clear in both its press release and the founders’ own statement that it will maintain Prior Labs’ independence, brand, and open source commitments. Confirmation of this promise will come in the coming years. Many similar promises have been made before.
The second is competitive. Tabular foundation models are no longer a quiet research category. Microsoft, Google, AWS and several startups are moving into the space. The leadership of previous Labs, especially the architecture of TabPFN and TabPFN-3 (the founders inevitably mention), is real, but not infinite.
The speed with which SAP deploys the €1 billion investment, hires additional researchers and can ship production versions to SAP’s customer properties will determine whether Prior Labs remains a category leader or becomes a cautionary tale of being sold too early.
The third is geopolitical. Frontier AI labs in Europe are increasingly becoming strategic assets. The week Prior Labs was incorporated into SAP, Techzine reported on this In parallel, SAP’s acquisition of Dremio brings its analytics data infrastructure under the same corporate umbrella.
The combined effect is that a single European company now controls a significant portion of the AI stack of structured data. Whether European regulators will view consolidation as sovereignty or concentration is a question that will attract attention at some point.
Three indicators will indicate whether the announcement exceeds corporate ambition. The first is Prior Labs’ rate of staff growth: a budget of €1 billion over four years means hundreds of new researchers, and the job market for spreadsheet-AI specialists is unusually weak.
The second is whether TabPFN-3 is shipping (“ahead”) on the schedule the founders envisioned, and how it compares to private-sector competitors that will ship their responses at launch.
The third is whether SAP customers actually adopt the models available in their operating systems at the scale that the deal economy implies; It’s one thing to have a state-of-the-art spreadsheet model, it’s another thing to improve it over the existing stack of a Fortune 500 financial institution.
None of these signals will be public for a while. What became public on Monday is that Europe has now produced the academic-commercial AI output its politicians demand, and the strategic buyer is itself European.
The mission, in the founders’ own words, has not changed. It’s just accelerated and at a scale most European startups never reach.
The acquisition closes in the second or third quarter at the discretion of regulators. The interesting work begins after that.
