• AMD announced its integrated AMD Kria AI platform designed for physical artificial intelligence and autonomous robots .

• System supported by Ryzen AI Embedded X100 processors; Combines CPU, GPU and NPU power in a heterogeneous architecture .

• Offering open source software support, the platform provides higher real-time reliability and higher process capacity compared to competing solutions. .

New Era in Autonomous Robot Technology: Body and Brain Unite

The world of industrial robotics is rapidly evolving from systems that execute pre-programmed standard commands to autonomous structures that can perceive their surroundings and make instant decisions. . Kria AI solutions announced by AMD directly target this transformation in robotic technologies. . The company, which has been providing real-time control in many areas from surgical robots to automation systems in heavy industries for years, brings this expertise to the robot’s perception and reasoning center with Kria AI SOM (System-on-Module) and robotic carrier cards. .

Powered by up to 16 “Zen 5” architecture CPU cores, AMD RDNA 3.5 graphics unit and low-latency NPU structure, the Ryzen AI Embedded X100 series maximizes power and thermal efficiency. . Thanks to the modular structure brought by the architecture, robots gain the process capacity to perform perception, reasoning and action processes within milliseconds. .

Seamless Real-Time Control with Heterogeneous Processor Architecture

The most critical factors in robotic systems are delay times and real-time reaction ability. . Kria AI SOM minimizes the need for data transfer between different process areas thanks to its unified memory architecture in its modular structure. . This allows the robotic system to make more than 8,000 control decisions per second, while enabling it to complete visual-language-based artificial intelligence reasoning processes in under 100 milliseconds. .

Developed based on open standards, modules in the COM-HPC form factor offer clear performance advantages over the competition. . According to test data and OpenNav benchmarks, the system offers up to 3.4 times higher real-time reliability for on-time control compared to Nvidia Jetson T5000, one of the strongest competitors in the market. . It also provides up to 1.6% more free CPU cores for spare process capacity and up to 2.3 times superiority in simultaneous autonomous spy running capacity. .

Open Source Ecosystem and Easy Transition Opportunity

AMD supports its hardware power with a flexible structure that does not restrict developers on the software side . The Kria AI platform, which fully supports industry standard software frameworks such as PyTorch, ONNX, ROS 2 and MoveIt, uses AMD ROCm software and AMD Robotics Software Suite infrastructure. .

This open architecture, which cares about developers preserving their existing code bases, makes it extremely easy to migrate CUDA-based software to the ROCm environment. . Internal tests show that existing codes can be protected at an average rate of 75% when switching from CUDA to HIP environment. . This eliminates the need for companies to write code from scratch, significantly shortening the time from concept to mass production. .

AMD Kria AI SOM modules and developer kits, which will be offered to the market through system manufacturers and brand new design manufacturer (ODM) partners, are planned to be available for general use in the fourth quarter of 2026. .