On May 4, 2026, SAP announced a definitive agreement to acquire Prior Labs, a German startup founded just eighteen months ago, and committed to investing more than €1 billion (about $1.16 billion) over the next four years to turn it into a frontier AI lab for the structured data behind global business operations. Financial terms were not disclosed; the deal is pending regulatory approval and expected to close in Q2 or Q3 2026.
The bet is deliberately not on large language models. Prior Labs builds tabular foundation models (TFMs) — AI that predicts from tables and databases, the format where enterprise information actually lives — and its open-source TabPFN tool has been downloaded more than 3 million times. TechCrunch’s same-day reporting surfaced the other half of the story: SAP’s API policy bans OpenClaw and other non-endorsed agents, admitting only its own Joule and Nvidia’s NemoClaw.
The Deal: A Cash Exit Plus €1 Billion Over Four Years
The official release does not disclose the price, but European tech outlet Pathfounders, citing sources, described an “almost all cash” deal with “well over half a billion dollars” paid up front to the founders. Against Prior Labs’ funding history — a roughly $9.3 million pre-seed in February 2025 led by Balderton Capital — that is an extraordinary return on an eighteen-month clock. Balderton partner James Wise called it “one of Germany’s biggest ever venture outcomes.”
SAP says Prior Labs will keep operating as an independent entity to preserve research velocity, with productization running through SAP AI Core, SAP Business Data Cloud, and the Joule agentic layer. SAP also says it is “fully committed” to supporting the open-source strategy and developer ecosystem behind TabPFN.
Eighteen Months, Freiburg, and TabPFN
Prior Labs was founded by Frank Hutter (CEO), Noah Hollmann, and Sauraj Gambhir. It is headquartered in Freiburg with offices in Berlin and New York, and its staff is drawn from Google, Apple, Amazon, Microsoft, G-Research, Jane Street, Goldman Sachs, and CERN. Its scientific advisory board will include Turing Award winner Yann LeCun and Bernhard Schoelkopf, director of the Max Planck Institute for Intelligent Systems.
The technology has real credentials. The TabPFN series was published in Nature and has set state-of-the-art results across hundreds of independent academic studies; the latest TabPFN-2.6 ranks first on TabArena, the leading TFM benchmark, matching the accuracy of a four-hour AutoML pipeline instantly in a single model. Its defining property is in-context learning: predictions with no model training at all. The team is scaling toward millions of rows, real-time inference, and new data modalities.
Why Structured Data, Not LLMs
SAP CTO Philipp Herzig put it directly: “Early on, SAP recognized that the greatest untapped opportunity in enterprise AI wasn’t large language models.” SAP built its own relational pretrained transformer, SAP-RPT-1, to prove that conviction, and is now doubling down by acquisition. Planned capabilities include a conversational interface for natural-language questions and what-if scenarios, GDPR-compliant predictions for use cases like payment delays, supplier risk, upsell, and churn, and agentic systems that combine tables, language, and images while reasoning toward causation rather than correlation.
The category is heating up. Competitor Neuralk-AI has already raised (though less than Prior Labs), and Fundamental emerged from stealth in February 2026 with a $255 million Series A. SAP’s edge is the data itself — the world’s business tables overwhelmingly live inside SAP systems.
Agent Policy: OpenClaw Out, NemoClaw In
The other half of the story concerns agents. SAP’s API policy “prohibits” AI agents except “SAP-endorsed architectures” — a stance first spotted by The Information. The authorized options are SAP’s own Joule Agents, still in beta, and Nvidia’s NemoClaw, Nvidia’s enterprise-ready, security-focused OpenClaw deployment, admitted because Joule supports Nvidia’s Agent Toolkit announced in March.
The contrast with Salesforce is stark: its new Headless 360 architecture lets enterprises choose their own agents, including OpenClaw. Two enterprise-software giants have given opposite answers to “who may operate APIs on the user’s behalf” — a fault line every developer building enterprise integrations should watch.
What It Signals for European AI
Hutter said joining SAP “gives us the resources, data environment and customer reach” to take the category to its full potential, with the stated goal of building a “globally-leading frontier AI lab for structured data — in Europe, in the open.” In a European AI story usually told as model research or acquisitions by American giants, one of Germany’s biggest venture outcomes plus €1 billion of long-term investment is a real endorsement of applied frontier work staying in Europe. For SAP the stakes are equally clear: with the stock under pressure in 2026’s “SaaSpocalypse,” CFO Dominik Asam told CNBC in January that the question is how quickly SAP can embrace these technologies to preserve its scale advantage. The stock traded slightly upward on the news.
Sources
- SAP bets $1.16B on 18-month-old German AI lab and says yes to NemoClaw — TechCrunch
- SAP to Acquire Prior Labs, Establish Frontier AI Lab in Europe — SAP News
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.