AMD Advancing AI 2026: Ryzen AI Embedded X100, Kria AI Robotics Platform, and Robotics Partner Network
The Ryzen AI Embedded X100 series of processors are AMD’s Strix Halo-class silicon, featuring Zen 5 CPU, RDNA 3.5 graphics, and XDNA 2 NPU. These processors are tuned for embedded and robotics applications. AMD projects that the physical AI silicon market will reach $200 billion by 2035 and is positioning its heterogeneous CPU, GPU, NPU, and FPGA portfolio to enhance robotics capabilities “from the robot body to the robot brain.” The on-board FPGA can be used to implement low-latency, application-specific functions for mission-critical environments.
Ryzen AI Embedded X100 Series
The X100 Series is a single embedded SoC that combines up to 16 high-performance “Zen 5” CPU cores, a discrete-class integrated GPU with up to 40 RDNA 3.5 compute units, and a power-efficient XDNA 2 NPU. AMD integrates these components with a unified memory architecture for low-latency, deterministic operation. It’s designed for use in robotics, industrial automation, healthcare, aerospace, and defense, with an operating range of -40 to 105°C and 24/7 operation for up to 10 years. AMD claims performance targets that surpass the Intel Core Ultra Series 3 and NVIDIA Jetson T500 by up to three times in applications such as medical ultrasound beamforming. Customer sampling began in June 2026, with production expected in Q4 2026. Launch partners for system-on-modules include Arbor, Congatec, iBase, IEI, Sapphire, and Seavo.Kria AI System-on-Module
The Kria AI SOM is a production-ready module based on the X100 Series, featuring a compact 120 x 120 mm open-standard COM-HPC form factor that combines CPU, GPU, NPU, and unified memory. AMD reports a deterministic control time of 125 microseconds (approximately 8,000 control decisions per second, as benchmarked by Bosch on a Rexroth controller), vision-language-action reasoning in under 100 milliseconds (92 ms), and vision classification in under 0.4 milliseconds. Compared to the Jetson T5000, AMD claims up to 3.4 times better real-time reliability, up to 1.6 times more available CPU cores, and support for up to 2.3 times more concurrent agents.
Kria AI Robotics Developer Platform
AMD introduces what it claims to be the industry’s first open, turnkey, integrated platform for autonomous robotics, combining CPU, GPU, NPU, and FPGA computing. It comes with the Kria AI SOM and a robotics carrier card, an evaluation kit, a robotics software suite, and validated reference designs. It also includes open-source baseboard schematics and FPGA design. The platform is currently being sampled with early-access customers, with general availability anticipated in the fourth quarter of 2026.
AMD Robotics Software Suite
The accompanying open software stack is built on AMD ROCm and the ROS 2 robotics framework. It includes purpose-built solution stacks, such as a Robotics Core SDK and Physical AI SDK, along with accelerated ROS nodes and MoveIt2 support. The suite supports industry-standard frameworks like PyTorch, ONNX, and TensorFlow. AMD’s open HIP tooling allows developers to migrate from CUDA to ROCm while retaining an average of 75% of existing CUDA code.
AMD Robotics Partner Network
AMD has also launched the AMD Robotics Partner Network, an open ecosystem of over 30 original design manufacturers (ODMs), AI model providers, and partners in middleware, sensors, simulation, and system integration, all built around AMD platforms and open standards. Membership spans four tiers, with free entry at the registered tier and no licensing or membership fees. Enrollment is open now. The program covers AMD compute solutions ranging from embedded Ryzen AI and Versal adaptive SoCs to EPYC CPUs and Instinct GPUs in data centers.
The complete slide deck follows.
