DeepX, an ultra-low power AI semiconductor company, and Hyundai Motor Group Robotics Lab are collaborating on the development of a next-generation physical AI computing platform capable of directly running large-scale generative AI models inside robots.

DeepX announced on the 21st that it is launching a joint development system with Hyundai Motor Group with the goal of realizing next-generation robot technology in which robots perceive their environment, make decisions, and act on their own.

This collaboration is noteworthy as it goes beyond simple technology exchange to form a strategic partnership for jointly designing the AI computing architecture that forms the core of robot platforms.

Recently, in the robotics industry, Vision-Language-Action (VLA) and Vision-Language Model (VLM) technologies, which integrate camera-based visual recognition, language understanding, and action decision-making, are emerging as key competitive advantages.

This is an essential element for robots to evolve beyond machines that perform simple, repetitive tasks into 'intelligent systems' that make autonomous decisions in real-world environments.

The two companies are expanding cooperation in areas such as ultra-low power AI semiconductor architectures, AI computing hardware for robots, physical AI software stacks, and robot application AI libraries to implement next-generation robot AI technology.

The goal of this collaboration is to build an integrated AI computing platform capable of processing large-scale generative AI models in real time inside robots..

The core technology utilizes the DX-M2 AI semiconductor currently under development by DeepX.

The DX-M2, scheduled to be produced using Samsung Electronics' 2nm process, is a next-generation physical GenAI chip designed to run large-scale AI models even in ultra-low power environments.

Designed for real-world environments where AI must react immediately, such as robots, autonomous vehicles, and industrial automation, this structure is expected to significantly enhance robot autonomy and response speed.

While the AI industry has developed around data centers, the era of physical AI is recently beginning in earnest as it expands into actual physical environments, such as robots, industrial equipment, and autonomous vehicles.

In this process, robots and humanoids are attracting attention as representative platforms where AI directly interacts with the real world.

The industry forecasts that the physical AI semiconductor market will grow to approximately $123 billion by 2030, with robots and humanoids expected to be the main sources of demand.

There are also analyses suggesting that a new ecosystem will form in the future robotics industry, centered around standardized computing platforms, just like in the semiconductor industry.

DeepX and Hyundai Motor Group Robotics Lab have been jointly developing 'edge brain' technology for robots based on low-power AI semiconductors for the past three years.

Building on past research achievements, this collaboration marks a step in expanding the scope of cooperation to encompass robot AI technology in general.

Both companies plan to continuously strengthen cooperation on AI computing technology across various robot platforms in the future.

DeepX CEO Kim Nok-won stated, “The AI industry is in a transitional phase shifting from a data center-centric approach to the real world,” adding, “Physical AI computing based on ultra-low power AI semiconductor technology "Our goal is to become a global leader in the platform sector," he said.

Hyun Dong-jin, Senior Vice President of Hyundai Motor Group Robotics Lab, stated, “Robots in the era of physical AI are the mediums that implement AI technology in the closest proximity to humans,” adding, “We will build a core technology ecosystem, including on-device AI computing, together with domestic and international partners.”