SAP Buys Prior Labs and Bets €1 Billion on a European AI Lab

SAP said on Monday that it has agreed to acquire Prior Labs, the Tübingen-based pioneer of tabular foundation models, and will invest more than €1 billion over the next four years to scale the startup into what the German software giant is calling a "globally leading frontier AI lab in Europe." Prior Labs will continue to operate as an independent entity, according to SAP's announcement.

The unusual phrasing matters. Most of the conversation about frontier AI has happened around chatbots and large language models, which excel at text but stumble on the numerical, tabular data that runs accounting, supply chain, and risk systems. Prior Labs has spent the last three years building a different kind of model. Its TabPFN series, originally published in Nature, treats spreadsheet-shaped data as a first-class input and has set the state of the art on tabular benchmarks across hundreds of independent academic studies.

Why structured data is the next race

The deal positions SAP at an angle to the LLM-focused investments dominating headlines. Co-founder Frank Hutter, who runs the lab with Noah Hollmann and Sauraj Gambhir, has argued for years that the data inside enterprise systems looks nothing like web text. Payment delays, customer churn, supplier failure, and demand forecasts are tabular problems, and language models trained on Wikipedia and Reddit have only a rough grasp of them.

Tabular foundation models are designed to predict outcomes from rows and columns directly. That puts SAP, whose business is built on the structured data sitting inside customer ERP and finance systems, in a position to offer prediction tools that don't require customers to convert their data into something a chatbot can read. TechTarget noted that the same announcement included a separate acquisition of Dremio, the open-source data lakehouse platform, broadening SAP's data infrastructure stack at the same time.

A European frontier lab, on purpose

The four-year, billion-euro figure is large by European AI standards, where private capital has lagged the United States and China. It also signals a sovereignty motive that has become more visible across the continent: European customers want enterprise AI tools whose training data, model weights, and operating jurisdiction stay close to home. The Cohere-Aleph Alpha merger announced in late April carried similar sovereignty language, with both the Canadian and German governments backing the deal as an alternative to American hyperscalers.

The transaction is expected to close in the second or third quarter of 2026, subject to regulatory approval. Terms were not disclosed. For Prior Labs, the appeal is staying independent while gaining access to SAP's customer base and data flows. For SAP, the prize is a research team that can keep producing models without the company needing to compete on chatbot performance against OpenAI and Anthropic. Whether tabular foundation models become a separate enterprise category or get folded into the next generation of multimodal frontier systems will shape how that bet pays out.