To realise physical AI, our vision is the robotisation of space
At the intersection of aging populations, labor scarcity, and the long human dream of mechanical helpers, Hyundai Motor Group arrived at CES 2026 in Las Vegas with something rare: not a promise, but a finished thing. The company unveiled a low-power AI chip, built over three years with Korean semiconductor firm DeepX, that allows robots to perceive and decide without leaning on distant servers — a quiet but consequential step toward machines that can inhabit the world as it actually is, not as engineers wish it were. With mass production beginning this year and a target of 30,000 robot units annually by 2030, Hyundai is wagering that robots are ready to become infrastructure, as unremarkable and essential as the elevator or the forklift.
- A chip that sips under five watts of power breaks the cloud dependency that has kept robots tethered to ideal conditions, making them viable in basements, warehouses, and anywhere the internet falters.
- Hyundai's three-year partnership with DeepX, reinforced by ties to Nvidia and Qualcomm, signals a deliberate effort to build supply-chain resilience before scaling — a lesson drawn from the semiconductor shortages that paralyzed automotive production earlier this decade.
- Real-world testing since June 2024 at Factorial Seongsu in Seoul, where robots Facey and DAL-e operated in live conditions rather than controlled labs, gave engineers the confidence to move from prototype to production line.
- Hospitals, hotels, and logistics centers are the first deployment targets — precisely the environments where labor is scarce, aging workforces are straining, and the cost of human error is high.
- The 30,000-unit annual production goal for 2030 is less a sales figure than a declaration: Hyundai is betting that autonomous robots will become as routine to certain industries as any other piece of capital equipment.
Hyundai Motor Group came to CES 2026 in Las Vegas not with a concept, but with a chip already built and ready for production. Developed over three years alongside DeepX, a Korean semiconductor specialist, the AI processor runs on less than five watts — frugal enough to let robots operate reliably in underground parking garages, warehouse floors, and other environments where cloud connectivity is unreliable or absent. Faster on-device decisions and stronger data security come as added benefits, both critical when machines are working alongside people or handling sensitive tasks. Partnerships with Nvidia and Qualcomm were woven into the supply chain deliberately, to avoid the kind of single-supplier fragility that has disrupted industries before.
The chip is not theoretical. Since June 2024, Hyundai has been running two robot models — Facey, a face-recognition unit, and DAL-e, an autonomous delivery robot — through real operating conditions at Factorial Seongsu in eastern Seoul. The company was watching not just for technical stability but for whether people would actually use the machines. Satisfied with the results, Hyundai is now moving into mass production, with hospitals, hotels, and commercial spaces as the first destinations. The target is 30,000 robot units per year by 2030, a figure that reflects a serious industrial commitment rather than a marketing aspiration.
Hyun Dong-jin, head of the Robotics Lab at Hyundai Motor and Kia, described the company's ambition at CES as 'the robotisation of space' — a vision in which robots become part of how environments function, not isolated tools performing isolated tasks. The company is drawing on its automotive manufacturing heritage and domestic battery partnerships to build the supply chain that ambition requires. The broader forces driving the bet are familiar: labor shortages across developed economies, aging populations, and persistent industrial safety challenges. A robot that can think for itself in a disconnected basement changes the equation for all three. Hyundai's MobED mobility robot also won best of innovation in robotics at CES this week, a signal that the industry is watching — and that the chip announcement has something real behind it.
Hyundai Motor Group walked into CES Foundry 2026 in Las Vegas this week with something concrete to show: a finished AI chip, ready for mass production, designed to let robots think and act on their own without constantly phoning home to the cloud.
The semiconductor emerged from three years of work with DeepX, a Korean company that handles the hardware side while Hyundai contributes the software and robotics know-how. The partnership also leans on relationships with Nvidia and Qualcomm, a deliberate move to avoid getting locked into any single supplier. What makes this chip different is its frugality—it runs on less than five watts of power, which means a robot can operate reliably in places where the internet is spotty or nonexistent: underground parking garages, warehouse floors, factory zones. The on-device approach also means faster decisions and better data security, both critical when a robot is working around people or handling sensitive operations.
Hyundai has already been testing the system since June 2024 at a facility called Factorial Seongsu in eastern Seoul. Two types of robots have been running through their paces there: Facey, which recognizes faces, and DAL-e, which delivers packages autonomously. These weren't lab experiments. The company was watching how the machines performed in actual working conditions, checking for stability and whether customers would actually use them. The results apparently satisfied the engineers enough to move forward.
This year, Hyundai plans to start putting the chip into robots rolling off production lines. The target is hospitals, hotels, and other commercial spaces where labor is hard to find or expensive. By 2030, the company wants to be manufacturing 30,000 robots annually. That's not a casual number—it reflects a serious bet that robots will become as routine in certain industries as forklifts are now.
Hyun Dong-jin, who runs the Robotics Lab at Hyundai Motor and Kia, framed the ambition at CES in terms that go beyond selling hardware. "To realise physical AI, our vision is the robotisation of space," he said. The phrase captures something important: this isn't about one robot doing one job. It's about building an ecosystem where robots become part of how spaces function—hospitals, hotels, warehouses, delivery networks. The company is leaning on its decades of experience in automotive manufacturing to build that ecosystem, including partnerships with domestic battery makers to ensure the supply chain holds up as production scales.
The broader context matters. Hyundai isn't inventing this problem. Labour shortages are real across developed economies. Populations are aging. Industrial safety remains a persistent challenge. Robots that can navigate real environments—not just controlled test chambers—address all three. A robot that needs perfect network connectivity and constant cloud processing is useless in a basement parking garage or a rural logistics center. A robot that thinks for itself, that can see what's around it and make decisions in milliseconds, changes the equation.
The timing also reflects where the industry is. Hyundai's next-generation mobility robot, called MobED, won best of innovation in the robotics category at CES 2026 this week. That's not just marketing noise—it signals that the company's robotics work is being taken seriously by the broader tech community. The chip announcement backs up the trophy. Without reliable, efficient on-device AI, the robots stay in the demo phase. With it, they move into actual production and deployment.
What happens next is straightforward in outline but complex in execution. Hyundai will begin manufacturing robots with this chip embedded. Hospitals and hotels will start receiving them. The company will watch what works and what breaks. Supply chains will either hold or strain. The 30,000-unit target for 2030 will either materialize or slip. But the fundamental shift—from robots as curiosities to robots as infrastructure—is already underway.
Notable Quotes
To realise physical AI, our vision is the robotisation of space. We are developing AI and software that enable robots to operate as autonomous agents in real-world environments.— Hyun Dong-jin, executive vice president and head of Robotics Lab at Hyundai Motor Co. and Kia
Our goal is not simply to build robots. It is to create a sustainable robotics ecosystem—delivering low-power, efficient and intelligent robots that generate real value for users at the final point of the physical AI pipeline.— Hyun Dong-jin
The Hearth Conversation Another angle on the story
Why does a robot need its own AI chip instead of just connecting to the cloud?
Because the cloud isn't always there. A robot in an underground parking garage or a basement warehouse might have no signal at all. Even where there is signal, sending data back and forth takes time—milliseconds matter when a robot is moving around people. And there's the security question: you don't want every decision a robot makes being transmitted and stored somewhere else.
So this is about independence.
Exactly. The robot becomes an autonomous agent. It sees, it thinks, it acts. All on its own hardware. That's what makes it useful in the real world instead of just in controlled environments.
Five watts seems impossibly low. How does that work?
It's about optimization. The chip is purpose-built for robotics, not trying to do everything a general-purpose processor does. You're not running a web browser or a video game on it. You're running perception and decision-making for a specific task. That focus lets you squeeze a lot of capability into very little power.
Why is Hyundai the one doing this? They make cars.
They make cars at scale. They know how to build supply chains, manage production, keep costs down while maintaining quality. Those skills transfer directly to robots. And they've been investing in robotics for years—this is the culmination of that work, not a sudden pivot.
What happens if this works? What's the actual impact?
You start seeing robots in hospitals helping with patient care, in hotels handling deliveries, in warehouses managing inventory. Places where labor is scarce or expensive. The robots become infrastructure, like forklifts. That changes how those industries operate.
And if it doesn't work?
Then Hyundai has spent three years and significant resources on something that doesn't scale. But they're already testing it in real conditions. They wouldn't be moving to mass production if they weren't confident.