SAP SE today announced plans to acquire two startups that help organizations analyze tabular data, or information stored in rows and columns.

Santa Clara, California-based Dremio Inc. raised more than $300 million prior to the deal. Prior Labs GmbH, the other startup that SAP is buying, is backed by about $9.3 million. The software giant didn’t disclose the two transactions’ financial terms.

Dremio has built a data management platform that can store and analyze large volumes of business information. A built-in artificial intelligence agent enables users to run queries without writing SQL code. The platform is built on two open-source projects, Apache Iceberg and Apache Polaris, that played a key role in SAP’s decision to acquire Dremio.

Iceberg is a data format for storing large spreadsheet tables. It eases tasks such as changing the schema of a dataset and splitting it into smaller pieces, which helps speed up queries. Additionally, there’s a built-in version control mechanism that makes it possible to quickly roll back erroneous changes.

Polaris, the other open-source tool that powers Dremio, eases the task of managing Iceberg tables’ metadata. Iceberg metadata tracks details such when a table was created and how many times it has been modified since. Additionally, it contains cybersecurity information such as which employees have permission to view a record.

SAP will use Dremio’s technology to enhance its Business Data Cloud offering. It’s a cloud service that enables organizations to combine datasets from different sources and make them available to AI models. According to SAP, Dremio will help improve the Business Data Cloud’s Iceberg support and metadata management capabilities.

“Enterprise AI doesn’t stall because the models aren’t good enough; it stalls because the data isn’t ready for AI agents,” said SAP Chief Technology Officer Philipp Herzig. ”Dremio eliminates that bottleneck.”

While Dremio’s technology eases the task of managing tabular data, Prior Labs focuses on helping users analyze it. The Berlin-based startup has developed an AI model called TabPFN-2.5 that is specifically optimized to process information stored in rows and columns. It can automate tasks such as scanning inventory tracking spreadsheets for erroneous entries. 

According to Prior Labs, TabPFN-2.5 can process up to 100,000 spreadsheet rows per task. A so-called distillation engine can generate lightweight, dataset-specific versions of the algorithm for user projects. Those versions are faster than the original and require less hardware to run.

SAP plans to integrate Prior Labs’ technology into multiple products including Business Data Cloud. The startup’s team, meanwhile, will continue to operate independently under the software giant’s wing. SAP will invest more than $1.17 billion in Prior Labs over the next four years to support its AI research.

Photo: SAP