SAP commits €1B to Prior Labs in bet on structured-data AI

SAP has acquired Berlin-based Prior Labs and committed more than €1 billion to expand the startup’s research, infrastructure and hiring. The deal gives SAP control of one of Europe’s most closely watched efforts to build foundation models specifically for structured enterprise data.

What happened

Prior Labs was founded around TabPFN, a model designed for tabular datasets rather than documents or web text. That matters because much of the information used inside businesses lives in spreadsheets, databases and ERP systems: demand forecasts, supplier records, churn signals, payment histories and operational metrics.

The acquisition price was not disclosed. SAP said Prior Labs will retain its brand, leadership, research programme and existing customer relationships, while receiving more than €1 billion of long-term support. The startup has said organisations including Hitachi and TD have already used its technology.

Why it matters

Large language models are useful for unstructured text, but they are not automatically the best tool for predicting outcomes from rows and columns. A model trained specifically for structured data could help SAP customers forecast demand, identify supplier risk or predict late payments without building a separate machine-learning pipeline for every task.

Owning the technology also reduces SAP’s dependence on external model providers and gives it a differentiated layer that can be embedded across its business-software portfolio.

The bigger picture

The transaction shows how incumbent software companies are using acquisitions to secure specialised AI research teams. SAP is not only buying a product; it is financing a research programme that could become part of the intelligence layer across its enterprise applications.

The main execution risk is integration. Prior Labs must preserve research speed while proving its models work reliably across messy, highly specific corporate datasets—not only benchmark tasks.