SAP closed its acquisition of Prior Labs on July 17 and pledged more than a billion euros to AI that predicts straight from structured business data. The chatbot is not the only AI worth a budget line.
The bet SAP just made
Most enterprise AI coverage this year has been about chat. Type a question, get a paragraph back. On July 17 SAP spent real money on a different idea: the most valuable AI inside a business does not talk at all.
SAP announced it completed its acquisition of Prior Labs, a German research lab founded barely 18 months ago, and committed more than 1 billion euros over four years to grow it into a frontier AI lab for structured data (per SAP's July 17 announcement). Deal terms were not disclosed. The strategy was.
What Prior Labs actually builds
Prior Labs is the pioneer of tabular foundation models, or TFMs. Its TabPFN models, published in Nature, set the pace on tabular benchmarks. Plain version: instead of training a fresh model for every dataset, one pre-trained model reads a table of business data and predicts the next value. Late payments. Customer churn. Supplier risk. Demand next quarter. It works on the rows and columns you already keep.
That matters because your firm does not run on prose. It runs on tables. Ledgers, claims, matters, invoices, patient records, positions. A chatbot summarizes those tables. A tabular model predicts from them. For a lot of professional-services work, the second job is the one that moves money.
Why SAP, and why now
SAP owns the systems where the world's structured business data already lives. It had shipped its own tabular model, SAP-RPT-1, so this deal buys a head start. Two weeks earlier, on July 6, SAP closed its purchase of Dremio, a data lakehouse (per SAP). Read the two deals together and the plan is plain: own the data layer, then own the models that predict on it.
| Question to ask | LLM chatbot | Tabular foundation model |
| Core input | Text and documents | Rows and columns of structured data |
| Core output | A written answer | A predicted value or score |
| Strong at | Drafting, summarizing, explaining | Forecasting, ranking, risk scoring |
| Governance question | What did it read? | What did it predict, and can you defend the score? |
The take
The chatbot got the headlines and most of the budget. The prediction engine that sits on your structured data got quietly expensive this month. A vendor roadmap that only mentions "an assistant" is describing half the category.
Here is the concrete takeaway. When a vendor pitches you AI, ask one question first: does it read my tables or only my documents? The answer decides which risks apply. A document assistant needs source control and human review. A prediction model needs data lineage, bias testing, and a way to explain any score to a regulator or a client who asks. That governance work is different, and you cannot start it if you never learned which kind of AI you bought.
SAP just paid over a billion euros to make sure the answer is both. Plan your budget for it.
AI, Properly is the Ayvede field report.
Advise. Vet. Deliver.
Ask a question in writing