In an announcement released Tuesday, Oct. 27, SAP SE disclosed plans to deploy SAP-RPT, its proprietary tabular artificial intelligence (AI) foundation model, aboard the International Space Station (ISS) on Hewlett Packard Enterprise’s (HPE) Spaceborne Computer-2.
Slated for implementation by early 2027, the orbital mission serves as an edge-computing validation testbed for enterprise-grade tabular foundation models operating in low Earth orbit.
Operational Triage in Extreme Edge Environments
Orbital platforms encounter substantial operational hazards, including cosmic radiation, extreme thermal cycles, space debris, and solar storms, while generating massive volumes of operational sensor telemetry. Downlinking raw sensor data to terrestrial stations introduces transmission delays and depends on intermittent communication windows, restricting real-time emergency intervention.
Running natively on HPE Spaceborne Computer-2, SAP-RPT will analyze telemetry streams directly at the orbital edge. The model is configured to triage sensor anomalies, infer missing telemetry values using historical baselines, predict component failures prior to escalation, and output automated, explainable health reports for onboard crew members and ground controllers.
Edge Efficiency and Model Specifications
Unlike large language models (LLMs) engineered for unstructured language analysis, tabular foundation models (TFMs) are optimized for structured datasets that drive complex physical systems. SAP-RPT functions within SAP’s broader TFM architecture, which includes the TabPFN model and SAP AI Core for pipeline orchestration.
Key technical parameters of the payload software include:
- Energy Footprint: Operates at approximately 50,000 times greater energy efficiency than contemporary LLMs, accommodating strict orbital power envelopes.
- Direct Inference: Generates predictions from multi-variable tabular inputs without requiring on-orbit model training or parameter fine-tuning.
- Decision Support: Delivers computer and mission risk scores alongside automated indicators assessing whether manual crew intervention is required.
“Space is the ultimate stress test for AI,” said Mayank Shrivastava, vice president of Generative AI and Foundation Models at SAP SE. “If a foundation model can predict mission-critical failures in orbit, under radiation, energy, and connectivity constraints, imagine what it can do for a global enterprise.”
Deployment Roadmap to 2027
Following scheduled deployment aboard the ISS in early 2027, findings from the orbital test will be incorporated into SAP’s edge-AI framework for disconnected terrestrial applications, including remote logistics, disaster relief, and distributed manufacturing. The mission also aims to establish operational baselines for future space-based data center architectures.
