
DEEPX explores chip and software cooperation with Intel, AMD and NVIDIA
The AMW Read
Early discussions extend a known edge-chip vendor's platform relationships, but broader integrations and commercial deployment remain unconfirmed.
DEEPX explores chip and software cooperation with Intel, AMD and NVIDIA
South Korean AI chip company DEEPX is pursuing technical cooperation with Intel, AMD and NVIDIA as it seeks wider support for its low-power neural processing units. The most concrete step is an agreement with Intel Korea to explore connecting and optimizing DEEPX NPUs with Intel's OpenVINO inference software. DEEPX also proposed combining its NPU with AMD's Kria embedded platform and connecting AMD's ROCm software with DEEPX development tools. It has discussed potential cooperation with NVIDIA and plans a meeting with senior executives at NVIDIA's US headquarters.
The discussions extend DEEPX's previously reported Intel collaboration to other major computing platforms. They address a challenge for AI infrastructure vendors: making specialized inference chips usable alongside established processors and software. DEEPX says its DX-Newton software, intended to work with NVIDIA's Isaac ROS robotics environment, has completed an initial internal test. The company says the discussions are at an early stage and have not established joint development or mass production.
For builders, the practical question is whether these discussions produce integrations that let existing models and tools run reliably on DEEPX hardware. For investors, completed integrations and customer deployments would provide stronger evidence of ecosystem access than meetings or proposals alone. The outcome could affect how easily DEEPX's low-power chips fit into robots and other edge devices that combine different processors.


