Berlin – October 27, 2026 -- SAP SE (NYSE: SAP) will run its SAP-RPT tabular foundation model on the HPE Spaceborne Computer-2 aboard the International Space Station, an edge-compute deployment it describes as one of the most significant for any enterprise AI model to date. The announcement headlined a package of agent, data and governance upgrades unveiled at the SAP TechEd developer conference.
SAP-RPT is planned to run onboard the ISS without sending data to Earth
Developed with HPE, the deployment is designed to support anomaly triage, predict issues before failures escalate, reconstruct missing sensor values from historical data, and generate explainable mission-health reports for astronauts and ground teams. SAP said no data needs to be sent back to Earth for analysis.
SAP also announced SAP-RPT-2.0, which extends the model beyond classification and regression to anomaly and outlier detection and embeddings-based analysis, with native multi-table support. A separate SAP-RPT-TS model will predict values across time-indexed business metrics for forecasting in SAP Integrated Business Planning without custom machine learning. SAP CTO Philipp Herzig cited TabPFN from Prior Labs alongside SAP-RPT as the basis of its tabular AI offering.
SAP adds an EU AI Act compliance assistant and a central AI traffic gateway
A new AI Governance Assistant is intended to help customers meet EU AI Act requirements. It ships alongside a new Joule Assistant and SAP AI Agent Hub, which SAP positioned as answers to barriers to scaling enterprise AI: keeping systems classified, compliant and within defined guardrails.
The new SAP Enterprise AI Gateway will consolidate authentication, policy enforcement, routing, monitoring, usage tracking and auditing for AI traffic across the enterprise. Third-party agents will be able to invoke Joule Agents through a governed agent gateway, with access control enforced. Joule Work will use the Agent2Agent (A2A) and Model Context Protocol (MCP) open protocols, plus new agentic planning capabilities for multistep workflows.
Joule Studio runtime stays free through December 2026 for select customers
SAP is running a time-bound promotion for select customers: deploying and running custom agents on the SAP-managed Joule Studio runtime remains free through December 2026. Design-time tooling for building and testing agents and n8n workflows stays free under fair-use limits. Customers in the early adopter program can already design and run n8n workflows in Joule Studio, with broader availability to follow.
Joule Studio offers more than 70 AI models and now includes observability and administrative controls covering agent behavior, model usage and spend. SAP also stated that it does not share customer data with anyone to train or improve large language models.
Dremio and Databricks partnerships widen SAP Business Data Cloud
SAP is adding Dremio to its lakehouse capabilities, with native support for Apache Iceberg and Apache Polaris and a federated query engine for data access without moving or copying data. Databricks capabilities will arrive in SAP Business Data Cloud as SAP Databricks, including Lakeflow, Agent Bricks, Unity Gateway and Lakebase. A new SAP Application Data Foundation and a Joule-to-Genie integration complete the package.
SAP Business Data Cloud Ontology Manager will provide AI-native authoring of ontologies across SAP and third-party data. SAP Reltio, announced earlier at SAP Connect, will bring multidomain master data management to the platform, unifying customer, supplier and product records across SAP and non-SAP systems.
Intel, Docusign, Parloa and Redwood extend the agent ecosystem
An Intel collaboration links enterprise workflows to autonomous robots, bridging information technology and operational technology through physical AI. New SAP solution extensions with Docusign, Parloa and Redwood target end-to-end process execution without handoffs between systems. Tens of thousands of developers joined Devtoberfest ahead of the conference.
Herzig summarized the strategy as "AI built where real work happens," resting on a context layer drawn from more than 50 years of process and industry expertise.