Digital Sovereignty

Your Business Data Just Got a European AI Lab

SAP has completed its acquisition of Germany's Prior Labs for more than 1 billion euros and pledged another 1 billion to build a European frontier AI lab for structured business data. Why the enterprise AI prize is your tables, not a chatbot.

Digital SovereigntyBy Servola Tech Desk2026-07-193 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

Your Business Data Just Got a European AI Lab

Key takeaways

SAP has completed its acquisition of Germany's Prior Labs for more than 1 billion euros, just 18 months after the startup was founded, and pledged another 1 billion euros over four years to grow it.
Prior Labs builds tabular foundation models, AI trained on the rows and columns of business data rather than on chat, and its TabPFN model was published in Nature and downloaded more than 3 million times.
SAP is keeping the lab independent and open: same brand and leadership, published research, and models that stay freely available, which is unusual for an enterprise acquirer.
For a European owner this is a sovereign option for AI over core business numbers, an alternative to sending your structured data through a US chat-model vendor.
Before your next AI purchase, ask whether the tool actually models your data or just wraps a chatbot around it, because that is the line this deal is drawn on.

What SAP actually bought

SAP paid more than a billion euros for a lab most people have never heard of. On completing the deal in July 2026, the German software group finalised its purchase of Prior Labs, a startup founded barely 18 months earlier, and committed a further 1 billion euros over four years to scale it. SAP called the goal a globally leading frontier AI lab based in Europe.

Prior Labs is not a chatbot company. It pioneered tabular foundation models, or TFMs, a type of AI built for the structured data that sits in spreadsheets, ledgers and databases. Its flagship model, TabPFN, was published in the journal Nature and has been downloaded more than 3 million times.

Why tables, not chat, are the enterprise prize

The AI that pays off inside a company rarely looks like a chat window. A business runs on rows and columns: orders, inventory, payments, customer records, sensor logs. A model that can predict, fill gaps and spot anomalies directly on that structured data does work a general chatbot cannot, without a human retyping numbers into a prompt.

That is the bet SAP is making. It already holds the structured data of a large share of the world's big companies inside its systems, and a tabular foundation model turns that data into forecasts and checks. The chat interface everyone photographs is the demo; the tables are where the money is.

The unusual part is that it stays open

Most enterprise acquirers close the doors and fold the team in. SAP said it will not. Prior Labs is to keep its own brand, leadership and research agenda, continue publishing its work, and keep its models openly available. For a company buying a frontier lab, leaving the science open is the exception, not the rule.

There is self-interest in it. An open, published model attracts the researchers and the outside scrutiny that keep a lab honest and current, and it lets customers and academics build on TabPFN rather than wait for SAP. But the effect for an owner is real: the core technology does not vanish behind one vendor's paywall on day one.

The sovereignty angle owners should not miss

Europe just gained a frontier AI lab it controls, aimed at the data businesses actually run on. Most foundation-model attention and ownership sits with US labs, and using them means routing prompts and often data through American providers. A European-owned lab focused on structured data is a different supply line for a sensitive part of the stack.

For an owner in the EU or UK, that matters where your core numbers are involved: financials, health records, supply-chain data you may not want passing through a US chat model under uncertain terms. This deal does not solve that alone, but it puts a credible European option on the table for the layer that touches your most regulated data.

What to ask before your next AI purchase

Use this deal as a test you can apply to any vendor. When a supplier pitches you AI, ask a plain question: does it model my data, or does it wrap a chatbot around it? A tabular approach reads your actual rows and columns and produces a forecast or a flag; a chat wrapper mostly rephrases what you type. The two cost and behave very differently.

You do not need to buy Prior Labs or SAP to act on this. You need to know that structured-data AI exists as a category, so you are not sold a chat demo as if it were an analytics engine. The owners who win the next budget cycle are the ones who can tell the difference before they sign.

Frequently asked questions

What is a tabular foundation model?

It is an AI model trained to work directly on structured data, the rows and columns found in spreadsheets, databases and business systems, rather than on free text or images. Prior Labs' TabPFN is the best-known example. Given a table, it can predict missing values, forecast outcomes and flag anomalies without a data scientist building a custom model for each task.

Did SAP shut Prior Labs down or absorb it?

No. SAP said Prior Labs will keep operating as an independent entity with its own brand, leadership and research agenda, will continue to publish its research, and will keep its models openly available. SAP committed more than 1 billion euros over four years to grow it into a European frontier AI lab rather than fold it into the parent.

Why is this framed as a sovereignty story rather than just a deal?

Because the layer being bought touches the data companies are most careful with, their structured business records, and most foundation-model capacity is owned by US labs. A European company owning and keeping open a frontier lab for that data gives EU and UK owners a domestic option for AI over their most sensitive numbers, which is a supply-chain and control question, not only a financial one.

The headlines will call this another billion-euro AI deal, and it is. But the quieter fact is where the money went. Not to a smarter chatbot, to the boring, structured data that every business already owns and few have made pay. That is the layer worth watching, and now Europe owns a piece of it.

SAPPrior LabsTabPFNTabular foundation modelsEnterprise AIEuropean AIStructured dataM&A

Databricks' 188 Billion Is A Signed Term Sheet

Databricks says it signed a term sheet at a 188 billion dollar valuation led by Coatue. The round size everyone quotes is press-only. Read the roadmap, not the number.

3 min read

Capital

NetApp bought a roadmap it will not promise

NetApp acquired DataPelago on 16 July 2026. The same release says NetApp has no obligation to deliver any capability resulting from the deal. What that means for your next renewal.

7 min read