SAP Buys an 18-Month-Old Lab to Own Structured Data AI
SAP’s acquisition of Prior Labs, an 18‑month‑old AI research company specializing in tabular foundation models for structured enterprise data, marks a strategic shift toward owning foundation models that natively understand the rows, columns and fields where business decisions are made. Rather than treating this deal as another headline in the enterprise AI acquisition cycle, it should be read as SAP’s declaration that the biggest value in AI is not in chatbots or content generation, but in prediction and decision-making on transactional data. SAP has completed the acquisition and committed more than €1 billion in investment to fund computing infrastructure, hiring and long-term frontier AI research, giving Prior Labs the runway to scale its TabPFN technology and expand its role as a pre‑eminent AI lab.
Why Tabular Foundation Models Beat General-Purpose LLMs in the Enterprise
Prior Labs pioneered tabular foundation models (TFMs), AI systems built specifically for structured enterprise data rather than free-form text. In most businesses, the decisions that matter—who will pay late, which supplier will fail, where demand will spike—live in tables of transactions, customer records, inventory levels and financial results. While large language models are impressive at interpreting and generating text, they are an awkward fit for this kind of numerical and categorical data. Traditional machine learning forces companies to train a separate model for each dataset and prediction problem, a slow and resource-heavy cycle of data prep and experimentation. TabPFN breaks that pattern with a single pre-trained foundation model that can be applied directly to many structured datasets to solve prediction tasks such as payment delays, churn, supplier risk and demand forecasting. In short, TFMs are purpose-built for the real backbone of enterprise analytics.
From Startup to Independent Lab: SAP’s €1 Billion Research Engine
The most telling aspect of the SAP Prior Labs deal is not the size of the investment but the structure of the relationship. SAP plans to support Prior Labs with more than €1 billion in capital for compute, hiring and long-term frontier AI research, while allowing the company to keep its brand, leadership, research agenda and customer relationships. That is a deliberate choice to preserve a research culture rather than absorb a product team. According to Prior Labs, its TabPFN models have already surpassed four million downloads and have been used across hundreds of independent research projects, ranging from pancreatic cancer diagnosis to wildfire prediction and next-generation battery materials. SAP is essentially building an internal but independent research engine focused on enterprise and scientific structured data, betting that multi-year programs in areas such as causal reasoning, relational data and agentic systems will pay off directly in its core software business.
Concrete Impact: Turning ERP Data into Predictions and Actions
The practical impact of tabular foundation models becomes clear when you look at everyday ERP and business intelligence workflows. Prior Labs’ approach is designed to reduce the time and expertise needed to build models for forecasting demand, predicting customer churn, identifying supplier risk and estimating payment delays, by allowing a single pre-trained model to tackle many datasets. Its technology is already in use, helping prevent train failures with Hitachi and improving financial forecasting with TD. With access to SAP’s vast ecosystems in finance, supply chains, manufacturing, HR, procurement and customer relationships, TFMs could be embedded directly into operational systems to anticipate late customer payments, flag supply chain disruptions, forecast product demand or identify equipment most likely to need maintenance. This is not AI as a separate analytics project; it is AI as a built-in prediction layer across the transactional systems organizations rely on every day.
What This Signals for the Next AI Battleground
SAP’s backing allows Prior Labs to pursue multi-year research initiatives that would have been difficult for an early-stage startup to finance on its own, with agendas spanning enterprise AI, scientific discovery, causal reasoning, relational data and agentic systems. Access to SAP’s enterprise data environments gives these efforts a deployment surface most AI labs can only dream of, even as the companies have not yet announced a specific product timeline. Strategically, this deal strengthens SAP’s position in AI that maps directly to its core software and lays the groundwork for future integration of TFMs into ERP and business intelligence platforms. The message to the market is clear: the next acquisition battleground is not generic language models, but structured data AI that can turn existing business tables into predictions and actions. Vendors that own their tabular foundation models will be able to differentiate not with user interfaces, but with the accuracy, speed and domain fit of their decision engines.
