MulticoreWare and AMD Partner to Drive Physical AI in Robotics

The demonstration highlighted how multimodal models function on integrated GPUs using the AMD ROCm software stack. By utilizing the AMD Robotics Software Suite, the two companies are addressing the technical gap between AI inference and physical actuation. The collaboration focuses on optimizing hardware-software co-design to ensure reliability, low latency, and accuracy for production-grade robotic systems.

Sumit Shah, head of product management at AMD’s Adaptive and Embedded Computing Group, noted that the X100 Series processors provide an open x86 architecture that avoids the constraints of proprietary software stacks. Vish Rajalingam, vice president at MulticoreWare, emphasized that successful deployment requires a tightly integrated loop between perception and action. This initiative builds on a 15-year history of cooperation between the firms, spanning high-performance computing, video codecs, and adaptive silicon.