SAP is acquiring 18-month-old German AI startup Prior Labs and investing €1 billion ($1.16 billion) over four years to turn it into a frontier AI lab for structured data. Simultaneously, SAP has banned OpenClaw and all unauthorized AI agents from accessing its products — allowing only its own Joule Agents and Nvidia's NemoClaw. The moves signal that Europe's most valuable tech company is playing both offense and defense in the AI era.
The Prior Labs Acquisition
Prior Labs was founded just 18 months ago by Frank Hutter, Noah Hollmann, and Sauraj Gambhir. The startup builds tabular foundation models — AI systems designed to make predictions from data that lives in tables and databases. This is fundamentally different from the language models that power ChatGPT and Claude.
Enterprise data does not sit in paragraphs. It sits in spreadsheets, SQL databases, and structured records. SAP's products for accounting, HR, procurement, and expense management all run on structured data. A model built specifically for tables is potentially more useful for SAP's customers than a general-purpose chatbot.
The deal was reportedly almost entirely cash, with founders receiving well over half a billion dollars upfront. Balderton Capital, which led Prior Labs' $9.3 million pre-seed, called it one of Germany's biggest ever venture outcomes. The startup's open-source TabPFN models have been downloaded over three million times.
SAP will operate Prior Labs as an independent unit. The company promised to maintain open-source versions of the models. The goal is to build a globally leading frontier AI lab for structured data — based in Europe.
Blocking Unauthorized AI Agents
The second announcement is equally significant. SAP has updated its API policy to explicitly prohibit AI agents from accessing its products unless they use SAP-endorsed architectures. OpenClaw and any other unauthorized agent frameworks are banned.
The only approved agents are SAP's own Joule Agents — still in beta — and Nvidia's NemoClaw, an enterprise-focused alternative to OpenClaw built on Nvidia's Agent Toolkit. SAP customers can create their own AI agents through Joule. They cannot use third-party agents to access SAP systems.
The move is defensive. As the agentic AI wave accelerates, enterprises are deploying autonomous agents that can access, modify, and execute actions across business systems. For SAP, uncontrolled agent access creates security risks, compliance concerns, and the potential for competitors to build AI layers on top of its platform.
The approach contrasts sharply with Salesforce. CEO Marc Benioff has taken the opposite strategy, allowing enterprises to choose their own agents — including OpenClaw — through Salesforce's new Headless 360 architecture. The two companies are making fundamentally different bets about whether controlling or opening the agent ecosystem is the winning strategy.
Why Structured Data Is the Next AI Frontier
SAP CTO Philipp Herzig framed the acquisition as a recognition that the biggest opportunity in enterprise AI is not language models. It is AI built for the structured data that runs the world's businesses.
The thesis is compelling. Enterprise AI revenue is surging. But most of that revenue comes from language-based tools — coding assistants, chatbots, document processing. The vast majority of enterprise data — financial records, supply chain logistics, HR systems, manufacturing data — lives in tables that language models handle poorly.
If Prior Labs can build AI that understands structured data as well as GPT understands text, the implications for enterprise software are enormous. Every SAP customer — and SAP serves 99 of the 100 largest companies in the world — could benefit from AI that works natively with their data rather than trying to convert tables into text.
The European AI Push
The deal adds to a growing wave of European AI activity. The Cohere-Aleph Alpha merger is building a sovereign Canadian-German AI platform at $20 billion. Ineffable Intelligence raised $1.1 billion in London. And now SAP is investing over a billion dollars in a German AI lab focused on the data infrastructure that enterprises actually run on.
The combined effect is a European AI ecosystem that is finding its niche. Rather than competing head-on with American frontier labs on language models, European companies are building AI for sovereignty, structured data, and regulated industries — areas where local expertise and data privacy requirements create natural advantages.
The Bigger Picture
SAP's dual announcement — acquiring Prior Labs and blocking unauthorized agents — reflects the two sides of the enterprise AI coin. On offense, SAP is investing heavily in AI capabilities tailored to its customers' data. On defense, it is locking down its ecosystem to prevent competitors from building AI layers on top of its platform.
Whether the walled-garden approach works in the agentic era — or whether it drives customers toward more open competitors like Salesforce — will be one of the defining questions in enterprise AI over the next two years.
