Achievements in Embedded and Edge AI Markets with M1
Expanding Partnerships with Hyundai Motor, POSCO, and Baidu
Performance on Par with or Surpassing NVIDIA GPUs
"Establishing AI as a Global Standard"

DeepX, an artificial intelligence (AI) semiconductor startup, is achieving visible results in the European and US markets with its ultra-low-power on-device chip, M1. As global manufacturing and embedded companies seeking alternatives to graphics processing units (GPUs) continue to adopt the M1, the pace of AI transformation (AX) in industrial settings is accelerating. DeepX's strategy is to expand its business with the goal of establishing a global standard for AI industry infrastructure.


Global Achievements by Partners Adopting DeepX M1
[Chip Talk] 'Emerging AI Powerhouse' DeepX: "Aiming for Global Standards" View original image

According to the semiconductor industry on December 1, global partners equipped with DeepX's M1 chip have recently achieved a series of innovative results. DeepX is collaborating with major system-on-module (SoM) manufacturers, aiming to realize "Physical AI." Among the targeted fields, embedded refers to systems in which dedicated chips are installed within devices such as robots and industrial machinery to perform specific functions. Edge AI is a technology that enables AI computation directly on devices without sending data to the cloud, and is utilized in smart factories and similar applications.


In the European market, Sweden's Virtium Embedded Artists AB recently won the top award at Embedded World North America 2025, held in Anaheim, USA. The company's SoM combines DeepX's M1 chip with Renesas's RZ/G3E processor to deliver up to 25 TOPS (trillion operations per second) of AI computing performance, while achieving ultra-low power operation of less than 5W. In edge AI applications, this offers power efficiency dozens of times greater than that of GPUs.


In the US market, Sixfab, an Internet of Things (IoT) and edge computing specialist, won the highest innovation award at CES 2026, the world's largest electronics and IT exhibition, with its AI gateway "ALPON X5," which incorporates DeepX's M1 chip. Sixfab's product is an on-device AI gateway that analyzes and processes data in real time on site, without relying on the cloud, enabling ultra-low-latency intelligent control at the terminal nodes of industrial, urban, and transportation infrastructure. Based on the M1 chip, it achieves over 10 times the power efficiency of GPUs while performing high-precision inference.


In particular, Sixfab has earned high credibility in the global market with its RaspberryPi-based modular edge devices, making DeepX's expansion into the developer ecosystem a noteworthy case. A DeepX representative emphasized, "The fact that our partners' products are being recognized as scalable AI infrastructure models means that DeepX's technology is becoming the standard for major embedded systems and industrial solutions."


Commercialization of Robot AI Platform Begins in Partnership with Hyundai Motor
The AI platform for robots jointly developed by DeepX and Hyundai Motor Robotics Lab has entered the commercialization stage. DeepX's M1 is integrated with Hyundai Motor Robotics Lab's facial recognition system, Facey, and based on this, the delivery robot DAL-e Delivery is demonstrating functions such as recipient facial authentication, user identification, and personalized guidance. DeepX

The AI platform for robots jointly developed by DeepX and Hyundai Motor Robotics Lab has entered the commercialization stage. DeepX's M1 is integrated with Hyundai Motor Robotics Lab's facial recognition system, Facey, and based on this, the delivery robot DAL-e Delivery is demonstrating functions such as recipient facial authentication, user identification, and personalized guidance. DeepX

View original image

In 2023, DeepX launched a strategic collaboration project with Hyundai Motor Robotics Lab and has been gradually applying the M1 chip to control platforms for service robots. The on-device AI controller jointly developed by both companies has now entered the stage of technical validation for mass production. The structure is capable of high-performance inference with less than 5W of power while meeting requirements for power efficiency and latency, making it suitable for both indoor and outdoor service robots.


This solution recently became the first in Korea to win the "MINDS 2025" award from the World Economic Forum (WEF), which recognizes technologies that are transforming the world through AI, after evaluating over 100 global corporate cases.


This year, the two companies developed a next-generation controller based on the M1, integrating wide-angle and telephoto dual image signal processor (ISP) cameras and vision AI technology. The goal is to create an on-device robot intelligence architecture that can operate without relying on network connectivity, even in environments with unstable communications such as underground parking lots, subway stations, and logistics centers. DeepX's M1 is integrated with Hyundai Motor Robotics Lab's facial recognition system, Facey, and based on this, the delivery robot DAL-e Delivery is demonstrating functions such as recipient facial authentication, user identification, and personalized guidance. Advanced interaction services are planned for the future.


The solution developed by DeepX and Hyundai Motor Robotics Lab will be unveiled at the Korea Tech Festival, hosted by the Ministry of Trade, Industry and Energy, starting December 3, and at CES 2026 in Las Vegas, USA, in January next year.


Leading the 'AI Transformation'... Collaborations with Hyundai Motor, POSCO, and Baidu
[Chip Talk] 'Emerging AI Powerhouse' DeepX: "Aiming for Global Standards" View original image

As a homegrown AI chip company, DeepX is developing cost-effective neural processing units (NPUs) that are less expensive than GPUs while offering superior power efficiency. The M1 chip outperformed NVIDIA Xavier in benchmarks for major "vision AI" models. For example, in the LeViT_128 image classification vision model, the M1 chip achieved 963 frames per second, compared to Xavier's 199 frames per second, demonstrating superior inference performance.


The first-generation M1 chip is being produced using Samsung Foundry's 5nm (1nm = one-billionth of a meter) process, with full-scale mass production beginning in August this year. The next-generation M2 chip will be manufactured using Samsung Foundry's 2nm process and is currently in development, with design completed and a prototype produced in the first half of this year.


DeepX is increasing collaborations with domestic conglomerates such as Hyundai Motor Robotics Lab and POSCO, while rapidly expanding its global sales network. Recently, DeepX reportedly signed a contract to supply over 40,000 NPUs to China's Baidu. The M1 chip will be supplied to the industrial PC manufacturing line of a Baidu partner company.


DeepX plans to lead the era of Physical AI in various fields, including industrial automation, robotics, smart cities, and vision systems, through collaborations with global partners such as Renesas and NXP.



Kim Nokwon, CEO of DeepX, stated, "AI is now moving beyond data centers to become a fundamental infrastructure in industrial sites. With ultra-low-power, high-performance AI semiconductor technology, we will realize intelligence across the global industrial ecosystem and establish AI as a global standard that creates real-world value."


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

© The Asia Business Daily(www.asiae.co.kr). All rights reserved.

Today’s Briefing