DEEPX captivated global tech leaders at CES 2026 in Las Vegas by showcasing its Physical AI infrastructure from January 6 to 9. This South Korean startup drew 20,000 visitors.
The company demonstrated how its low-power semiconductors bridge the gap between digital intelligence and real-world machinery. Consequently, CTA named their exhibit a must-see for two consecutive years.
Specifically, DEEPX secured two CES 2026 Innovation Awards within the computing hardware and embedded technology categories. This achievement reflects the rising demand for efficient edge AI.
Partner Sixfab won the Best of Innovation award for the ALPON X5. This AI gateway integrates the DX-M1 chip to provide superior edge computing capabilities for industrial users.
Furthermore, DEEPX held a high-profile media briefing on January 6 at the LVCC South Connector. The executive team shared their vision for standardizing Physical AI worldwide.
To accelerate this mission, the company launched the Open Source Physical AI Alliance. They partnered with industry leaders like Ultralytics and Baidu to unify edge AI software.
The DX-M1 serves as the first-generation vision AI chip from DEEPX. It operates at 25TOPS while consuming only 5W. It supports various M.2 and PCIe form factors.
Meanwhile, the company unveiled the DX-M2 roadmap during the event. This second-generation chip utilizes a 2nm process. It focuses primarily on Large Language Model (LLM) processing.
Compared to the DX-M1, the DX-M2 offers a three-fold performance increase. It delivers up to 80TOPS of power. Consequently, it can run models with up to 100 billion parameters.
Physical AI extends beyond data centers into drones and robots. Therefore, DEEPX positions its NPU as the essential heart for autonomous systems interacting with the physical world.
Hyundai Motor Group Robotics Lab collaborated with DEEPX to integrate on-device AI. They plan to deploy these solutions in robots and security systems starting in late 2026.
At North Hall booth #8745, visitors witnessed delivery robots avoiding obstacles in real-time. These machines used the YOLO model for vision and Whisper for voice transcription.
Surveillance drones also performed real-time object tracking during live demonstrations. Moreover, these drones processed LLM tasks while consuming less than 5W of energy during flight.
The technical team demonstrated 1.58-bit quantization for AI models. This advancement allows complex LLMs to run on low-power hardware like Raspberry Pi and Orange Pi boards.
Furthermore, DEEPX integrated its NPU into industrial PCs from Advantech and DFI. These collaborations target smart factory and retail solutions to improve operational efficiency.
Major server manufacturers like Dell and HP are also integrating these NPUs. This integration facilitates the expansion of edge AI into enterprise-level workstations and servers.
The DX-M2 incorporates RISC-V processors to optimize Transformer models. This architecture enables efficient AI agents to operate directly on devices without cloud reliance.
This CES 2026 showcase proved that DEEPX leads the transition toward on-device intelligence. Their technology successfully brings high-performance LLMs to the edge of the network.
The evolution from DX-M1 to the 2nm DX-M2 represents a massive leap in efficiency. These advancements will likely define the future of autonomous robots and smart city infrastructure.
It is 4:30 in the morning in Yongsan, central Seoul. A line is already forming…
On a cold Tuesday in Gangnam, a 46-year-old visitor from New Jersey steps out of…
In late 2024, a Boston private equity firm quietly put a Korean company up for…
As any person would be, visiting South Korea for the first time, I was incredibly…
On a Tuesday morning in November 2025, a children's cartoon company rang the opening bell…
Somewhere on the edge of Gimpo, just west of Seoul, sits a permit that never…