NVIDIA and Foxconn Automate GB300 AI Server Assembly with Robotic Systems
NVIDIA and Foxconn demonstrated robotic automation for assembling complex GB300 AI server components, achieving over 95% success rate in busbar installation. The system also achieved a 90% to 95% success rate for multi-connector insertion, automating traditionally labor-intensive manual tasks. As demand for large-scale AI infrastructure explodes, automating high-voltage power distribution and connector assembly helps ensure consistent manufacturing quality and scale up AI server production. It demonstrates how industrial robotics can streamline the complex hardware manufacturing process required for next-generation AI data centers. The automated process targets completing busbar installation within 124 seconds and plugging four connector sets in under 72 seconds. While the overall automated cycle time is currently up to twice as long as a skilled human worker, it significantly improves assembly stability and repeatability.
## BACKGROUND
The NVIDIA GB300 NVL72 is a liquid-cooled, rack-scale system based on the Blackwell Ultra architecture that integrates GPUs and CPUs into a unified platform using NVLink high-speed interconnect technology. Busbars are conductive metal bars used for distributing large electrical currents across high-power server racks. Foxconn manufactures over 40% of global AI servers and partners closely with NVIDIA on rack architecture and manufacturing solutions.