AMD Launches Ryzen AI Embedded X100 and Kria AI SoM to Challenge Nvidia in Robotics

AMD is bringing Zen 5, RDNA 3.5 and 50-TOPS NPUs to the robotics edge with the Ryzen AI Embedded X100 family and a new Kria AI SoM that scales up to 128GB of unified memory.

AMD Launches Ryzen AI Embedded X100 and Kria AI SoM to Challenge Nvidia in Robotics

AMD used its Advancing AI 2026 event on July 23 to open a second front against Nvidia in physical AI, launching the Ryzen AI Embedded X100 processor family and a new Kria AI system-on-module that pairs Zen 5 CPU cores, an RDNA 3.5 iGPU and an XDNA 2 NPU with up to 128GB of unified memory for robotics workloads.

Ryzen AI Embedded X100: Six SKUs, 50 TOPS, Ten-Year Availability

The Ryzen AI Embedded X100 family launches with six SKUs — the X199, X188 and X168 plus their -40°C industrial-grade siblings X199i, X188i and X168i — all built around Zen 5 CPU cores, an integrated RDNA 3.5 GPU and a 50 TOPS NPU inside a 55 W TDP. Peak configurations pair 16 CPU cores at 5.1 GHz with a 40-compute-unit iGPU and LPDDR5x-8533 memory across up to eight channels with Link-ECC. AMD claims a 2.1× multithread CoreMark uplift and 3.5× higher token generation over Intel's Core Ultra Series 3, and up to 3× peak FP32 performance against Nvidia's Jetson AGX Thor T5000.

Kria AI SoM Adds Unified Memory and 234-Agent Scale

Alongside the CPUs, AMD launched a redesigned Kria AI SoM built as a COM-HPC module around the same X100 silicon — a departure from earlier Kria modules that shipped Zynq FPGAs directly on-module. The new SoM offers 64GB or 128GB of LPDDR5x unified memory, up to eight FAKRA camera inputs on the carrier, CAN-FD, RS485 and 10GbE QSFP networking. AMD says the module can host up to 234 concurrent agents based on mimik Agentix benchmarking, and delivers 8,000 real-time control decisions per second on a Bosch Rexroth reference controller.

AMD Kria AI Robotics Developer Platform

Robotics Developer Platform Sampling Now, GA in Q4

The Kria AI Robotics Developer Platform pairs the SoM with a Spartan UltraScale+ FPGA carrier board that adds GMSL2/3 camera inputs, EtherCAT/TSN networking, an integrated IMU, mini DisplayPort outputs and both USB4 and Oculink expansion. Sampling of the X100 CPUs began in June 2026 with mass production slated for Q4 and availability guaranteed through 2037. The developer platform is already shipping to early access customers, with general availability — and ODM variants from Arbor, Congatec, iBase, IEI, Sapphire and Seavo — also scheduled for Q4. Pricing is expected above $5,000, in line with Nvidia's Jetson AGX Thor Developer Kit.

Related coverage: Nvidia opens Jetson Thor T3000 and T2000 for mainstream robotics, Nvidia's Cosmos 3 Edge sweeps Japan's factory floors, and Google DeepMind's Gemini Robotics 2 leaps to whole-body humanoids.

Reporting based on coverage from AMD, CNX Software and TechPowerUp.

Category: Electrical Engineering

Tags: Robotics Physical AI Industrial AI industrial robotics Semiconductors AI Chips

Related Articles