Part III: Manufacturing Deployment

Chapter 9: From DGX to Jetson — Training, Simulation, and Edge Deployment

Written: 2026-06-08 Last updated: 2026-06-08

Physical AI infrastructure has three layers: DGX or cloud for training and synthetic data, Omniverse/Isaac for simulation and validation, and Jetson/Thor for edge inference and control. The manufacturer must decide where each data stream is generated and where each decision is made.

Figure 9.1: Jetson/IGX edge hardware connected to production inference and safety gates. source photo reused from S6
Figure 9.1: Jetson/IGX edge hardware connected to production inference and safety gates. source photo reused from S6

9.1 Cloud vs. On-Prem

Manufacturing data includes layout, production volume, quality defects, and customer SKUs. DGX Cloud and partner data-factory blueprints can accelerate training, but sensitive production logs may require on-prem or private-cloud governance [2].

9.2 Edge Deployment

Jetson Thor brings model inference to the factory floor. That does not mean a large model replaces all low-level control. Servo loops and safety PLCs remain; Jetson is more likely to handle perception, skill selection, anomaly detection, and policy inference.

Figure 9.2: NVIDIA reference cases organized as a DGX-Omniverse-Jetson pipeline. illustration by author AI-assisted
Figure 9.2: NVIDIA reference cases organized as a DGX-Omniverse-Jetson pipeline. illustration by author AI-assisted

References

  1. NVIDIA (2026). NVIDIA and Global Robotics Leaders Take Physical AI to the Real World. NVIDIA Investor Relations.
  2. NVIDIA (2026). NVIDIA Announces Isaac GR00T Reference Humanoid Robot. NVIDIA Investor Relations.
  3. NVIDIA (2025). NVIDIA Jetson Thor. NVIDIA Product Page.