MOU signed with Intel Korea on Thursday

DeepX NPU to be integrated with OpenVINO AI inference software

CEO Kim visited Intel last month to discuss optimization

Three-way MOU with server firm KTNF also signed for on-premises AI inference

DeepX CEO Kim Nok-won (center left) and Intel Korea CEO Bae Tae-won (center right) pose for a photo after signing a memorandum of understanding at COEX in Seoul on Thursday. [Intel Korea]
DeepX CEO Kim Nok-won (center left) and Intel Korea CEO Bae Tae-won (center right) pose for a photo after signing a memorandum of understanding at COEX in Seoul on Thursday. [Intel Korea]

AI chipmaker DeepX announced Friday that it signed a memorandum of understanding with Intel Korea to cooperate on AI semiconductor inference technology.

The two companies signed the MOU on Thursday at the "2026 Intel AI Summit Seoul" at COEX in Seoul.

Key areas of cooperation include integrating and optimizing Intel's OpenVINO AI inference software with DeepX's neural processing unit, collaborating on physical AI and edge AI platform technology, conducting technical validation of AI semiconductors and system architecture, and reviewing generative AI solutions based on the DX-M2.

Central to the agreement is the integration of OpenVINO — Intel's open-source software toolkit for optimizing AI models to run across diverse computing environments — with DeepX's NPU.

The two companies plan to convert models so that DeepX's NPU can be used within the OpenVINO environment and to examine compatibility and optimization of the execution environment. The goal is to give developers a setup that lets them apply DeepX's NPU more easily using existing AI models and development tools.

In the physical AI domain, the companies will explore the technical feasibility of a platform combining the host and control functions of Intel's x86 processor with the perception and inference capabilities of DeepX's NPU, and will seek ways to demonstrate and expand its application in robotics.

The partnership will also examine the potential for developing enterprise AI appliances and on-device AI services using the DX-M2, DeepX's next-generation generative AI NPU. The DX-M2 is a next-generation AI semiconductor under development that aims to run large language models (LLMs) and vision-language models in a low-power environment of around 5 watts, based on Samsung Electronics' 2-nanometer process.

Ahead of the MOU, DeepX CEO Kim Nok-won met last month with members of Intel's OpenVINO development team in Silicon Valley to discuss integrating the company's NPU with OpenVINO and explore AI model optimization strategies.

DeepX is also extending its technical cooperation with Intel Korea into the server infrastructure space. DeepX, Intel Korea and KTNF signed a separate three-way MOU on Thursday, agreeing to collaborate on on-premises AI inference servers using heterogeneous computing technology.

Under the agreement, DeepX will contribute its NPU and AI model optimization technology, Intel Korea will bring its CPU and xPU platform along with related software capabilities, and KTNF will focus on server design, manufacturing and system integration.

The three companies plan to validate AI inference performance and compatibility in real server environments and to explore joint development and commercialization opportunities to meet demand for on-premises AI infrastructure from domestic and international enterprises and institutions.


jeongwan@heraldcorp.com