AMD revealed its new Kria AI system-on-modules during the Advancing AI 2026 event in San Francisco, introducing a robotics platform centered on the Ryzen AI Embedded X100 Series processors. This platform packs a CPU, GPU, NPU, and FPGA into a unified memory system, allowing all components to access the same memory pool rather than shuttling data between separate chips, a move aimed at boosting efficiency and performance.
The company directly challenges Nvidia’s Jetson Thor T5000 with claims that their new board delivers 3.4 times faster real-time processing, supports 2.3 times more concurrent agents, and leaves 1.6 times more CPU capacity free after handling workloads. In certain signal-processing tasks, AMD has a threefold increase in FP32 performance.
Built on the COM-HPC standard, the platform is designed for flexibility, enabling developers to swap modules without redesigning entire systems. AMD also supports the board with a software stack based on ROS 2 and its ROCm GPU compute framework, though ROCm still trails Nvidia's CUDA in tooling maturity. The company made schematics publicly available to encourage adoption.
Foundation Future Industries is among the first to adopt this technology, migrating its Phantom humanoid robots from Nvidia and Intel hardware to AMD’s X100 platform. AMD’s background in FPGAs through its Xilinx acquisition laid groundwork for this integrated solution, which unites all necessary compute types under one roof with software built for the AI era.
AMD’s announcement aligns with growing competition in edge robotics hardware, aiming to disrupt Nvidia’s stronghold. Market response is unfolding as AMD stakes a claim with impressive performance gains.
This material is for informational purposes and does not constitute financial advice.
