In the context where generative AI has not yet truly deeply penetrated enterprise business processes, European enterprise software giant SAP is choosing to take the initiative through large investments. The company recently announced its intention to acquire Prior Labs, an 18-month-old German AI startup, and plans to invest 1 billion euros (approximately $1.16 billion) in this business over the next four years to establish a "cutting-edge AI laboratory" for structured data in Europe.
The deal amount has not been officially disclosed, but sources say it is a substantial exit "almost entirely in cash," with more than $500 million in cash going directly to the three co-founders Frank Hutter, Noah Hollmann, and Sauraj Gambhir.
Prior Labs was established a year and a half ago and focuses on the so-called "Tabular Foundation Models" (TFM), which are AI models that predict common tables and database data in enterprises. Compared with large language models, this type of model is considered to be more consistent with enterprise scenarios, and is also more consistent with SAP's core strengths in finance, HR, procurement and expense management software - these products are highly dependent on the underlying database. The TabPFN model series launched by Prior Labs is very popular in the developer community, and its open source models have been downloaded more than 3 million times, making it a new force in the field of structured data AI. SAP promised in the acquisition announcement that the laboratory will operate as a relatively independent research entity and continue to maintain the open source version. At the same time, with the help of SAP AI Core, Business Data Cloud and Joule, the agent layer, the research results will be quickly productized and embedded into SAP's existing product portfolio.
For the Freiburg-based startup, SAP’s entry is equivalent to an “acceleration pass.” Founder Frank Hutter said on social platforms that he hopes to use the "huge impetus" provided by SAP to build Prior Labs into an important center for cutting-edge AI research on structured data in Europe and even the world, and to maintain a high-intensity research rhythm in an open environment. The acquisition exit comes less than two years after the company closed a pre-seed round of approximately $9.3 million (€9 million) led by Balderton Capital. Balderton partner James Wise described the deal on social platforms as "one of the largest exits in German venture capital history," while SAP shares edged higher after the news.
This acquisition also continues SAP’s overall layout idea in the AI field. As early as 2023, the company has invested in several large model companies, including OpenAI competitor Anthropic, and now Aleph Alpha and Cohere, which are planning to merge. SAP has previously self-developed a relational pre-training Transformer model called SAP-RPT-1. Its CTO Philipp Herzig has publicly stated that for enterprise AI, the real opportunity that has not been fully exploited lies not in the large language model itself, but in the AI capabilities that can understand and process the structured data that supports the operation of global enterprises. By absorbing Prior Labs’ team and technology, SAP is trying to use “shortcuts” to significantly shorten the development cycle on this track.
At the same time, SAP’s strategy on the new trend of “Agentic AI” is obviously biased towards defense and strong control. While preparing to build its own AI laboratory, Germany's most valuable company has adopted a blocking attitude towards unauthorized AI agents. "The Information" first disclosed that SAP has blocked all agent technologies that have not been explicitly approved, such as OpenClaw, from the system, and clearly written terms in the latest API policy: unless it belongs to the "SAP Approved Architecture", AI agents are prohibited from accessing SAP products through its API. Naturally, those currently whitelisted include SAP’s own Joule Agents, a product still in beta that allows customers to build their own enterprise-level agents. In March, Nvidia announced that SAP Joule would support its NVIDIA Agent Toolkit—the foundation for its enterprise-focused, security-focused OpenClaw competitor NemoClaw. This means that in the SAP ecosystem, NemoClaw has officially received a "pass", while other unrecognized agent technologies are shut out.
In the shock of the "post-SaaS era", this combination of SAP reflects the anxiety and self-rescue of traditional enterprise software giants. Affected by the so-called "SaaSpocalypse" market sentiment, SAP's stock price has been under significant pressure this year, and AI is regarded as a key variable in reshaping its scale advantage and profit structure. SAP CFO Dominik Asam admitted in an interview in January this year that the key for the company is how quickly it can incorporate these new technologies into its own R&D portfolio, thereby maintaining and amplifying its relative economies of scale advantages. In sharp contrast, Salesforce, which is also in the midst of turmoil, allows customers to freely choose agents through its Headless 360 architecture, and third-party solutions including OpenClaw can access it. In terms of agent strategy, SAP chose a high degree of control and "inward-looking openness", while Salesforce emphasized ecological diversity and external compatibility. How these two paths will affect their respective enterprise customer bases and AI transformation progress in the next few years will become a major focus of the enterprise software market.