| Noah Hollmann was just months into his associate role at Boston Consulting Group when he decided to leave the top tier job and go to medical school. He enrolled at Charité Universitätsmedizin Berlin in 2019, where he started experimenting with how AI could improve medical research. The problem? Existing AI tools were no match for the complicated genetic data that he was studying.
To overcome the problem, he partnered with University of Freiburg professor Frank Hutter and Samuel Müller to develop a new form of an AI model: one that could ingest data stored in complicated spreadsheets, instead of written documents that large language modelsworked best with.
Their experiment grew into AI startup Prior Labs, which now offers open source models that have been downloaded more than 4 million times by clients that want AI to analyze similar medical data, or even understand their finances, make inventory predictions or avoid issues with their supply chain.
“It works with everything that has a table underlying it, and that’s almost everything in the world,” Hollmann says. “If you have a database where you record whatever goes on in your business, [this] foundation model helps you make sense of the data.”
This month, the startup was officially acquired by SAP. The cofounders were reportedly paid more than half a billion dollars in cash upfront, and SAP has committed to pay more than $1.14 billion into Prior Labs over the next four years to help it scale.
This acquisition is just the second influx of money into the startup, which began as a research lab in 2024. In February 2025, they raised a $9.6 million pre-seed round from Balderton Capital and XTX Ventures. And they added experts like Yann LeCun (who developed neural networks) and Bernhard Schölkopf (a machine learning pioneer) to their board of advisors. From there, two critical changes made the billion-dollar acquisition possible, Hollmann says: recruitment and commercialization.
“We were able to attract the best talent in the world,” he says, adding that recruits include researchers from Apple, Google and Goldman Sachs. “And we showed that we are able to execute.”
Early clients include Oxford Cancer Analytics, and Hitachi Rail—a Japanese railway giant that makes and monitors trains and rail signal systems, and uses Prior’s TabPFN to predict railway defects.
Now Prior Labs is betting that SAPs 400,000-plus client roster will account for even more use cases at the enterprise level.
“One of the exciting bets that we are now working on is predicting why something happened, or what you can do to change a system,” Hollmann says. “Those are the holy grail of understanding data and something that’s completely unsolved at the moment.“
See you next week,
Alex & Zoya |
