SAP has announced a definitive agreement to acquire Prior Labs, the German startup specializing in foundational models for tabular data. The deal, still pending regulatory approval, aims to strengthen a line of Artificial Intelligence that could have a much greater business impact than it may initially appear: models capable of understanding tables, numbers, statistics, and structured data — precisely the kind of information that underpins most business processes.
The German company has not disclosed the purchase price but has announced an investment of over €1 billion over the next four years to turn Prior Labs into a leading European AI laboratory with global reach. According to SAP, Prior Labs will continue to operate as an independent entity, with its own brand, research team, and commitment to open source.
Why SAP is looking beyond large language models
This acquisition sends a clear message: SAP wants to differentiate itself where it has the most enterprise data. Large language models have demonstrated tremendous capacity for working with text, generating responses, summarizing documents, and assisting in conversational tasks. However, they are not always equally reliable when it comes to reasoning about spreadsheets, financial histories, supplier tables, collection forecasts, default risks, inventories, or customer churn patterns.
This is where Tabular Foundation Models, or TFMs, come into play. Unlike general-purpose LLMs, a TFM is specifically designed to work with structured data. Its goal is not to produce better writing but to improve prediction accuracy. SAP mentions use cases such as payment delays, supplier risks, opportunities for upselling, or customer attrition probability. In other words, everyday challenges in finance, procurement, sales, supply chain, and operations.
The move aligns with SAP’s prior launch of SAP-RPT-1, a model oriented toward predictions on enterprise data. With Prior Labs, the company aims to accelerate this line and incorporate one of the most recognized teams in tabular AI into its ecosystem. Its belief is that the next significant competitive advantage in enterprise AI will not only be in conversational assistants but also in the ability to anticipate happenings within the business using proprietary data.
For SAP, this makes strategic sense. Its clients already manage critical information across ERP, finance, procurement, HR, supply chain, and customer experience systems. If SAP can securely, govern, and connect TFM models to these data sources, it could embed predictive AI as a native feature within its applications, rather than adding an external layer afterward.
Prior Labs, TabPFN, and the appeal of structured data
Based in Freiburg with offices in Berlin and New York, Prior Labs was founded by Frank Hutter, Noah Hollmann, and Sauraj Gambhir. The company gained recognition for TabPFN, a family of foundational models for tabular data that has seen strong adoption within the technical community. SAP states that TabPFN exceeds three million downloads and will continue supporting its open source strategy.
The scientific interest in TabPFN is also notable. An article published in Nature introduced TabPFN as a foundational model for small to medium tabular datasets, capable of supervised learning on datasets with up to 10,000 samples and 500 features, outperforming traditional methods in this context.
This approach diverges from traditional machine learning pipelines that typically involve data preparation, feature selection, training, hyperparameter tuning, validation, and deployment — often requiring extensive, case-specific training. Prior Labs promises to reduce some of this complexity through models that can adapt to new datasets within context, without the need for prolonged retraining each time.
SAP cites TabPFN-2.6 as a leading model in TabArena and claims it can match the accuracy of a four-hour AutoML pipeline instantly, with less complexity. While this statement is within the context of corporate announcements and must be validated case by case in real-world environments, it points to an important direction: providing advanced predictive capabilities to business users who are not data scientists.
With a conversational interface on top, a user could ask about scenarios, select datasets, compare hypotheses, or run simulations without manually building pipelines. The goal is to bring predictive analytics closer to finance, sales, operations, or procurement profiles — assuming data is well-governed and predictions are explainable.
Europe gains a frontier enterprise AI laboratory
SAP frames this move also as a European initiative. Creating a cutting-edge AI lab focused on enterprise structured data can strengthen a position that Europe needs: not only competing in general chatbots but also developing applied AI for industrial, administrative, and corporate processes where its companies hold expertise, customers, and high-value data.
The scientific advisory board for Prior Labs includes figures such as Yann LeCun, Turing Award winner and CEO of Advanced Machine Intelligence, and Bernhard Schölkopf, director of the Max Planck Institute for Intelligent Systems and president of ELLIS. Their presence reinforces the research-oriented profile of the project and situates it within the international conversation on specialized foundational models.
SAP intends to integrate this research with its platform. After the transaction closes, the company plans to incorporate Prior Labs’ capabilities into SAP AI Core, SAP Business Data Cloud, and the agent layer of Joule. This is crucial because the commercial value isn’t solely in improving models but in embedding them within existing business workflows.
For example, a company could use historical order, invoice, customer behavior, and inventory data to predict delays, risks, or purchasing needs. An AI agent could then recommend actions, generate alerts, or trigger workflows within the ERP. The TFM would handle the predictive aspect; Joule and SAP applications would provide the interface and execution.
This move also underscores the idea that enterprise AI will increasingly be multimodal. LLMs will remain useful for language, reasoning, and conversation, but companies will need specialized models for tables, time series, images, documents, code, or specific processes. SAP aims to be at that intersection—where models not only excel at language but also predict and act upon real business data.
A still-young market with great potential
The category of foundational models for tabular data is still in its early stages. While promising, there are challenges: enterprise data is often incomplete, poorly labeled, dispersed across systems, and subject to privacy regulations. Additionally, in regulated sectors, it’s not enough to just achieve accurate predictions — explanations, documentation, and bias mitigation are essential.
SAP speaks of moving beyond correlation toward causality — a significant ambition. Predicting that a customer might churn is useful; understanding which factors cause that risk and what actions can reduce it is far more valuable. In finance, procurement, or supply chain management, this difference can transform an alert into a truly actionable decision.
While the acquisition of Prior Labs won’t solve all these challenges overnight, it lays an important foundation. The deal is expected to close in the second or third quarter of 2026, followed by technical and commercial integration. SAP will need to demonstrate that these models can be effectively deployed at scale, with security, compliance, and user-friendly operation.
This is still a significant move. While much of the market remains focused on conversational models, SAP’s bet on less glamorous but closer-to-core enterprise data — structured tables — could have deeper long-term impact. If AI learns to better understand these datasets, the effects could go far beyond faster email composition.
The acquisition highlights where the next phase of enterprise AI might head: more specialized models, embedded in proprietary data, integrated into workflows, and delivering actionable predictions. For SAP, it’s not just a technology purchase — it’s a strategic effort to defend its leadership in enterprise software amid an AI landscape where control over data, context, and action is paramount.
Frequently Asked Questions
What has SAP announced?
SAP has signed a definitive agreement to acquire Prior Labs, a German startup specializing in foundational models for tabular data. The deal is pending regulatory approval.
What are foundational models for tabular data?
They are AI models designed to work with structured tabular data such as sales, payments, inventories, risks, customers, or suppliers. Their primary goal is to make predictions on enterprise data.
How much will SAP invest in Prior Labs?
SAP plans to invest over €1 billion over the next four years to develop Prior Labs as a European frontier AI lab.
What about TabPFN and open source?
SAP states that Prior Labs will operate independently and will continue supporting its commitment to open source through TabPFN.
via: news.sap