DeepX Partners with the World's Leading Object Recognition AI Platform, Targeting the Global Standard Ecosystem

Strategic Partnership with Ultralytics, Developer of YOLO
Leading Physical AI with Low-Power, High-Performance AI Semiconductors

DeepX, a fabless AI semiconductor company, announced on May 15 that it has signed a strategic technology and market entry partnership with Ultralytics, the developer of the object recognition AI model 'YOLO'.

Ultralytics is a company that provides the most widely used AI model framework among computer vision developers worldwide. Ultralytics' YOLO has effectively become the standard model across various sectors, including industrial cameras, robotics, autonomous driving, and smart cities. The platform is downloaded more than 300,000 times daily, functioning as a core development environment for numerous global companies and developers for actual products and services.

This collaboration is particularly significant because DeepX's Neural Processing Unit (NPU) will be natively integrated directly into this global standard platform. Notably, Ultralytics' existing hardware partners include global edge semiconductor companies such as Intel and STMicroelectronics. DeepX, by joining at the same level, is now positioned to compete within the global ecosystem.

It is considered an exceptional case for a Korean AI semiconductor company to be included as a default option in the basic toolchain used by developers worldwide. Through this integration, developers can now deploy AI models optimized for DeepX chips with just a single line of command in Ultralytics' Python package, without the need for a separate complex conversion process.

This fundamentally addresses the hardware integration issue, previously cited as the biggest obstacle in the journey from AI model development to deployment on actual devices, and is seen as having established a 'one-click' development environment.

Through this collaboration, DeepX provides a hardware environment optimized for the latest YOLO model series, particularly the next-generation 'YOLO26'. The company's flagship product, DX-M1, is already in mass production, and the next-generation DX-M2, which utilizes a 2-nanometer (nm, 1 nm = one-billionth of a meter) process, is also being designed to support this model. This enables developers to instantly run the world's leading object recognition models in ultra-low-power environments.

DeepX’s NPU is designed to dramatically reduce power consumption while maintaining high performance, making it optimized for implementing physical AI in real-world applications such as robotics, industrial automation, and smart cities. In particular, its energy efficiency and total cost of ownership (TCO) position it as a practical alternative to the traditional high-power GPU-dependent structure.

Kim Nokwon, CEO of DeepX, stated, "This partnership marks a crucial turning point where K-AI semiconductors become the default choice for developers worldwide," and added, "We aim to accelerate an era where DeepX's intelligence becomes the de facto standard for physical AI semiconductors in all devices in our daily lives, including robotics, smart cities, and home appliances."

This content was produced with the assistance of AI translation services.