SAP, also acquiring Dremio and Pryon Labs... targeting the enterprise AI "data preparation" bottleneck

To strengthen its table data analytics capabilities, SAP acquired American Dremio and German Prelab. Amid intensifying competition in generative AI, the move directly targets the “data preparation” issue that determines whether enterprise AI succeeds or fails.

The Dremio SAP acquired is a data management startup headquartered in Santa Clara, California, USA. Before the acquisition, it had raised more than $300 million, or approximately 443.1 billion Korean won. The Berlin-based Prelab, which was acquired along with it, is a company that received about $9.3 million, or roughly 13.7 billion Korean won, in investment. SAP did not disclose the specific amounts for these two acquisitions.

Dremio provides a platform for storing and analyzing large-scale enterprise data. The platform comes with an AI agent that can run queries without having to write SQL code directly. SAP’s particular focus is the open-source projects “Apache Iceberg” and “Apache Polaris,” which form the technical foundation of Dremio.

Iceberg is a format for storing large-scale tabular data. It allows data structures to be changed easily or large-volume tables to be split into multiple units. This helps improve query speed, and it also enables fast recovery from incorrect changes through version management features. Polaris is a tool for managing Iceberg table metadata. It systematically manages information such as table creation time, modification history, and access permissions, offering advantages in data governance and security.

SAP plans to integrate Dremio’s technology into its Business Data Cloud. The Business Data Cloud is a service that integrates data from multiple sources and supports AI model usage. SAP believes that by acquiring Dremio, it can further enhance its Iceberg support capabilities and metadata management functions.

SAP Chief Technology Officer (CTO) Philip·Hertzsch said, “The reason ‘enterprise AI’ is stuck is not that model performance is insufficient, but that the data is not prepared for AI agents,” adding that “Dremio will eliminate this bottleneck.”

Prelab’s technology will also be integrated into the Business Data Cloud

If Dremio has an advantage in data management, Prelab focuses on tabular data analysis. The company developed an AI model called “TabPFN-2.5,” optimized specifically for processing information made up of rows and columns. For example, it can automatically carry out tasks such as finding incorrectly entered items in inventory-tracking spreadsheets.

According to Prelab, TabPFN-2.5 can handle up to 100,000 rows of data per task. In addition, through a “distillation engine,” lightweight models optimized for specific datasets can be generated. This version is said to run faster than the original model and require fewer hardware resources.

SAP also plans to incorporate this technology into multiple products, including the Business Data Cloud. However, after the acquisition, the Prelab team will operate independently under SAP. SAP announced that it will invest $117 million—approximately more than 1.7281 trillion Korean won—over the next four years to support Prelab’s AI research.

The acquisition has been interpreted as SAP’s strategic intent not only to add AI models, but to reshape enterprise data infrastructure itself into an AI-friendly form. In the enterprise AI market, data preparation, metadata management, and analysis automation are just as important as model performance; therefore, SAP’s move to strengthen the Business Data Cloud is expected to draw broad attention from the industry.

TP AI Notice: This article is summarized based on the TokenPost.ai language model. It may omit the original text’s main content or be inconsistent with facts.