SAP plans to put table-based AI aboard the ISS

SAP plans to deploy SAP-RPT on HPE Spaceborne Computer-2 aboard the International Space Station by early 2027. Announced on 27 October, the project would analyse mission telemetry, reconstruct missing sensor readings and produce health reports for crew and ground teams.[1]

How the model uses a table

SAP’s current product documentation describes in-context learning: a request supplies example records, and the pretrained model uses their patterns to predict values for other records. Classification chooses a category; regression predicts a number. A new task can therefore be expressed through its context without a separate task-specific training run.[2]

The company’s commercial model versions are available through SAP AI Core. Its playground accepts CSV files for classification and numerical regression, giving users a way to examine predictions and the rows and columns that explain them. That existing service is separate from the planned onboard deployment.[2]

A computer already operating in orbit

The hardware programme has a longer history. NASA’s February 2025 account describes Spaceborne Computer-2 as a ProLiant server and an Edgeline server already on the ISS. High-energy radiation can corrupt electronic data; the programme explored software responses that allow commercial computers to operate in that environment.[3]

The programme’s approach was to monitor component behaviour and slow the computer when it strayed outside normal operating parameters. That protection predates the proposed SAP application.[3]

Keeping a computer within operating limits and deciding what its records mean are separate jobs. SAP’s plan applies a table-native model to the second.[1][3]

Report an error or suggest a correction