SAP plans to run SAP-RPT on HPE’s Spaceborne Computer-2 aboard the ISS

SAP SE said on Oct. 27, 2026 that it intends to place its SAP-RPT tabular AI foundation model (tabular models use structured data; large language models use unstructured language) on the International Space Station, running on Hewlett Packard Enterprise’s Spaceborne Computer-2, and put the system into operation in early 2027.

SAP-RPT flags sensor anomalies, closes gaps in telemetry using historical baselines, projects part failures before they intensify, and issues automated, explainable health updates for crew members and ground controllers. SAP presents these outputs as decision support: computer and mission risk scores alongside automated flags that say when the crew must take manual action. Mayank Shrivastava, vice president of Generative AI and Foundation Models at SAP SE, said “Space is the ultimate stress test for AI,” and suggested that forecasting critical failures in orbit could show similar value for businesses worldwide.

SAP-RPT will handle telemetry streams right there on the station instead of transmitting raw data to Earth. Downlinking raw sensor data to ground stations creates transmission delays and depends on intermittent communication windows, restricting real-time emergency response. Processing aboard the station avoids sending the underlying data back to Earth for review, which matters because a platform produces large volumes of operational sensor telemetry while coping with extreme thermal cycles, solar storms, cosmic radiation, and space debris.

The energy footprint is a key technical parameter of the payload software: SAP-RPT runs at approximately 50,000 times the energy efficiency of contemporary large language models. It generates forecasts from many table-based inputs without training or adjusting any parameters while in orbit. The tabular kind works with the structured datasets behind complex physical systems; large language models, instead, center on unstructured language.

SAP says the station deployment will test tabular foundation models built for enterprises in low Earth orbit through edge computing, checking how the software performs under orbital conditions. Lessons from the orbital test will feed SAP’s edge-AI framework for disconnected uses on the ground such as distributed manufacturing, disaster relief, and remote logistics. The station mission also seeks to set operational baselines for future space-based data center designs. SAP describes the station trial as an important edge-computing validation for enterprise AI.

Philipp Herzig, CTO of SAP SE, said “The same trusted SAP-RPT model that powers global supply chains also can work in space.” He said SAP blends SAP-RPT with TabPFN technology from Prior Labs to expand the use of tabular AI in business contexts and elsewhere.

SAP made the announcement during TechEd activities in Berlin, where the annual developer conference opened on Oct. 27, 2026. At the event SAP showed AI tools for developers covering Joule Studio, Joule Work, and SAP Business AI Platform, and the studio lets developers reach more than 70 AI models to build and extend agents. Herzig said the conference demonstrates SAP’s shift from AI learning the business to running the business, with the whole stack running on SAP Business AI Platform to deliver faster time to value, higher accuracy, and secure, governed deployment.

SAP also unveiled SAP-RPT-2.0, an updated tabular AI foundation model that goes beyond classification and regression into anomaly and outlier detection via embeddings, and SAP-RPT-TS, a tabular model intended to predict time-indexed business measures for SAP Integrated Business Planning with no custom machine-learning models needed. It is part of the same tabular AI family that includes SAP-RPT and SAP-RPT-2.0 from TechEd 2026.

SAP said it does not give customer data to third parties for training or improving large language models of any kind, a promise SAP made within its wider AI governance announcements at TechEd. The SAP Business AI Platform offers a central environment for building, contextualizing, and governing AI across both SAP and third-party systems.

SAP revealed the AI platform and agent strategy for the first time at Sapphire in May 2026, with the SAP Business AI Platform and SAP Autonomous Suite introduced there. It brings together SAP Business Technology Platform, SAP Business Data Cloud, and SAP Business AI inside one governed environment and incorporates SAP Knowledge Graph data modeling technology from 2024.

On Oct. 27, 2026, MarketScreener described the station deployment of HPE’s Spaceborne Computer-2 as among the largest edge-compute validations yet for an enterprise AI model. SAP said the deployment remains a plan, not a completed project, and that SAP-RPT has not yet produced results on the station.

SAP said as of October 26, 2026 that a phased general availability rollout covering Joule Work and Joule Studio has begun. Joule Work lets AI agents manage complex business processes using open protocols such as Agent2Agent and Model Context Protocol. Members of the Joule Studio early adopter programme can build and run n8n workflows, with wider availability planned later.