Siemens and NVIDIA Test AI Factory Planning at Erlangen
Siemens says its Erlangen electronics plant is using digital twins, synthetic inspection data and an AI planning agent in live production.
Siemens announced on October 19, 2026 that it is testing new manufacturing AI workflows with NVIDIA at its electronics factory in Erlangen, Germany.
The work covers two defined tasks: preparing inspection systems before physical PCB prototypes exist, and generating production-plan options against factory constraints.
Inspection before the first physical board
Siemens says PCB digital twins are used to create photorealistic training images. Those images train vision systems before a new board reaches the line.
The company reports up to 85% less engineering effort for virtual production preparation and up to 40% shorter time-to-market for this workflow. These are results stated for the Erlangen deployment, not general performance guarantees.
Planning proposals checked against a factory twin
An AI agent generates production scenarios when new orders arrive. The scenarios account for capacity, material availability, delivery dates and available lines.
Planners see the effect on defined KPIs. Siemens says proposals are then checked in a digital twin of the plant before implementation. The company reports potential cost savings of up to 20% for the end-to-end planning workflow.
What automation teams should examine
- Version control for the product and factory digital twins.
- Validation of synthetic inspection images against real defects.
- Limits placed on the planning agent's decisions.
- Human approval before production schedules change.
- Rollback and traceability when a plan performs poorly.
The release names NVIDIA Cosmos world models, Factory Operations Blueprint and Omniverse libraries. Their value depends on the plant data, constraints and validation process around them.
For a smaller simulation workflow, see Using Factory I/O With a Virtual Siemens S7-1200 PLC.