AMD's x86 + NPU platform for physical AI and robotics — the X100 processor family (6 SKUs), Kria AI SoM (COM-HPC), and Kria AI Robotics Developer Platform, announced at Advanced AI 2026.
Ryzen AI Embedded X100 — 6 SKUs
X199 / X199i · X188 / X188i · X168 / X168i
AMD Zen 5 — 8 / 12 / 16 cores, up to 5.1 GHz
RDNA 3.5 iGPU — 32–40 Compute Units
XDNA 2 — up to 50 TOPS
55W
LPDDR5X-8533, up to 8-channel, Link-ECC
HW encode/decode up to 4K @ 60 Hz
Up to 5 ports — HDMI 2.1, DP 2.0, eDP, DVI, USB4 (8K@60)
16-lane Gen4
45 × 37.5 mm
0–105°C std · −40–105°C industrial ('i')
2.5 yr std · 10 yr extended · to 2037
COM-HPC — 64/128GB LPDDR5X, 8× FAKRA, CAN-FD, 10GbE QSFP
Kria AI Robotics — 2×5GbE + 2×GbE + 10GbE, 2× USB4, M.2 NVMe
Arbor, Congatec, iBase, IEI, Sapphire, Seavo
Sampling June 2026 · MP Q4 2026
Available
Quote
The AMD Ryzen AI Embedded X100 family is AMD's aggressive entry into physical and edge AI — pairing Zen 5 x86 CPU cores with an RDNA 3.5 iGPU and an XDNA 2 NPU delivering up to 50 TOPS in a single 55W package. Announced at Advanced AI 2026 (July 2026), it directly challenges Intel Core Ultra and NVIDIA Jetson Thor for robotics, industrial automation, healthcare, UAV, and media-production workloads.
Across its six SKUs (X199/X199i, X188/X188i, X168/X168i with 16/12/8 Zen 5 cores and 40/32/32 GPU compute units), AMD claims up to 2.1× multithread CPU performance and 1.7× graphics versus Intel Core Ultra Series 3, and up to 3× peak FP32 versus the NVIDIA Jetson Thor T5000. The industrial 'i' variants extend the operating range to −40–105°C with a 10-year availability commitment to 2037 — critical for long-lifetime industrial deployments.
The platform ships three ways: the X100 processor (45 × 37.5 mm package, LPDDR5X-8533 up to 8 channels with Link-ECC, 16-lane PCIe Gen4), the Kria AI SoM (a COM-HPC module with 64GB or 128GB LPDDR5X, up to 8 FAKRA camera inputs, CAN-FD, RS485, and 10GbE QSFP), and the Kria AI Robotics Developer Platform (a full carrier with 2× 5GbE + 2× GbE + 10GbE QSFP, 2× USB4, M.2 NVMe, and a Spartan UltraScale+ FPGA on the carrier). The SoM supports up to 234 concurrent AI agents and 8,000 decisions/second.
Software support spans Linux, the ROCm stack for iGPU acceleration, Xen Hypervisor, and standard AI frameworks (PyTorch, ONNX, TensorFlow), plus tooling to migrate CUDA codebases to ROCm. SoM partners Arbor, Congatec, iBase, IEI, Sapphire, and Seavo will bring carrier boards and systems from Q4 2026.
Need AMD Ryzen AI Embedded X100?
AMD's x86 + NPU physical AI platform with 10-year availability. Contact QS Compute for pricing, sampling, and the Kria AI Robotics Developer Platform.
Request Quote