Nota AI, the Korean specialist in AI model compression and optimization, has joined the AMD Robotics Partner Network to help commercialize on-device physical AI on AMD embedded silicon. The KOSDAQ-listed company announced the tie-up at AMD Advancing AI 2026 in San Francisco on 24 July.
The Partnership
The AMD Robotics Partner Network is an open ecosystem that supports development and deployment of robotics solutions across embedded, edge and cloud environments built on AMD compute. The network already includes ODMs, ISVs, sensor suppliers and system integrators; Nota joins to contribute hardware-aware model compression, pruning, quantization and deployment tooling that lets factories, warehouses and mobility platforms run large perception and control models on AMD processors without offloading to the cloud.
Why On-Device Matters For Robots
Robots operate within tight power, thermal and memory envelopes. Nota argues that unless models are optimized to those envelopes, latency, reliability and safety suffer because inference has to hop over congested networks. "To deliver real value at industrial sites, physical AI requires technology that optimizes AI models for the actual computing and memory environments of robots," Nota AI CEO Chae Myung-soo said. Kaby Tazabur Bhaskar, head of AMD Robotics, said the ecosystem aims to "accelerate robotics innovation and expand market opportunities for partners" through joint hardware, software and integrator work.
Nota's Toolchain
The company's flagship platform NetsPresso automates neural-architecture search, compression and deployment for computer-vision and multimodal models, with more than 40 models optimized in production and validated deployments across 100+ hardware targets. NetsPresso has been referenced by NVIDIA as a success case and cited by Arm as an AI case study, and now targets AMD's Ryzen AI Embedded, Kria and Versal silicon inside industrial robots.
Reads Alongside AMD's July Push
The announcement lands the same week AMD Advancing AI 2026 detailed the sixth-generation 2 nm EPYC Venice server CPU and MI455X Helios rack for data-center AI. The Nota move extends that push to the physical-AI edge, complementing programs like Wetour Robotics' Qualcomm partner track as chipmakers race to become the default compute layer for humanoid, mobile and industrial robots.
Reporting based on coverage from PR Newswire, Manila Times, Taiwan News and Vietnam Investment Review.