
DEEPX and Intel Korea plan inference collaboration for robots and servers
The AMW Read
The agreements add potential software and processor partners for DEEPX, but compatibility and commercial deployment remain unproven.
DEEPX and Intel Korea plan inference collaboration for robots and servers
DEEPX has signed an agreement with Intel Korea to explore connecting Intel's OpenVINO inference software with DEEPX neural processing units. The companies plan to examine model conversion and runtime compatibility so developers can use existing AI models and tools with DEEPX chips. They will also assess a platform that combines Intel x86 processors for robot control with DEEPX NPUs for perception and AI inference. Separately, DEEPX, Intel Korea and South Korean server company KTNF signed an agreement to collaborate on AI servers using different types of processors.
The agreements place DEEPX in the market for specialized inference hardware alongside established processor and software suppliers. OpenVINO compatibility could reduce the work needed to deploy existing models on a new NPU, while a combined x86 and NPU platform would divide control and inference tasks across different processors. Both proposals remain subjects for technical review; the announcement does not establish compatibility or a working robot or server platform.
For builders and investors, the next evidence to watch is whether the partners demonstrate model conversion and execution across the proposed hardware. The companies will also examine enterprise AI equipment and on-device services using DEEPX's DX-M2 chip, which is in development. DEEPX is targeting LLM and vision-language model operation at approximately 5 watts using Samsung's 2-nanometer process; that figure is a development target, not a reported test result.


