Ageing labour forces and EV assembly precision, not novelty, are the real reasons Hyundai is gambling on humanoid robots. By Stewart Burnett

Hyundai will take full ownership of US robotics firm Boston Dynamics by acquiring SoftBank’s remaining stake of around 10%, a deal valued at roughly KRW 500bn (US$335m). The move follows Hyundai’s initial 80% acquisition of the robotics pioneer in 2021 and clears the way for the group to align Boston Dynamics entirely with its own automotive manufacturing priorities.

The centrepiece of that plan is Atlas, Boston Dynamics’ humanoid robot, and the production-ready version was unveiled at CES 2026. Hyundai plans to first deploy Atlas at its Georgia electric vehicle (EV) plant starting in 2028, initially on parts sequencing and heavy lifting, before expanding its role to welding and component assembly—and deploying it in other plants—by 2030.

With Boston Dynamics now wholly-owned, Hyundai is building capacity to produce up to 30,000 Atlas units annually, with plans to deploy more than 25,000 across its own and Kia’s plants while selling the remainder to external customers. It boasts multiple apparent advantages over its human counterparts, including the ability to rotate its torso through a full 360 degrees. The robot can also autonomously navigate to charging docks to swap its own batteries and lift 30 kg on a sustained basis.

Full ownership also removes the last structural friction in Hyundai’s plan: a minority shareholder with its own return expectations and timeline can no longer complicate decisions about how aggressively Boston Dynamics’ technology gets folded into the automaker’s factories. The underlying driver is demographic and industrial rather than purely technological. Indeed, ageing labour forces and rising manufacturing costs across Hyundai’s key markets, combined with the precision demands of EV assembly, have made a scalable substitute for repetitive, physically punishing tasks a genuinely compelling solution.

Hyundai’s ownership also lets it fold Boston Dynamics’ robotics expertise directly into its existing supply chain and innovation network. Hyundai Mobis, which functions as both the group’s parts division and a software development unit, is set to manufacture Atlas’s custom actuators, lowering production costs and keeping a critical component in-house. At the same time, live car factories can provide the kind of closed-loop, real-world testing data that pure research and development environments are incapable of replicating.

Boston Dynamics’ Atlas humanoid robot

The biggest complication may not come from the robot’s efficacy but local resistance from Hyundai’s labour unions. Korean workers, already in the midst of threatening strike action against the automaker, have explicitly flagged displacement and potential loss-of-hours as concerns at the bargaining table as negotiations continue. Back in January, the union warned that “not a single robot can be deployed at worksites without an agreement between the union and management.”

Of course, Hyundai isn’t the only automaker investing in humanoid robots. Tesla has grabbed far more headlines with its Optimus robot, arguably the centrepiece of a company-wide ‘physical AI’ rebranding. In Tesla’s case, it is the robot’s AI—more so than its physical hardware—that is positioned as the real product. Tesla leans on neural networks and imitation learning drawn from its Full Self-Driving architecture, including the AI4 and AI5 chips, and is testing Optimus directly inside its own Texas and Fremont plants.

An aggressive production timeline is planned for Optimus, with production set to begin in August 2026 and scaling towards mass production of up to one million units annually the following year. With that said, Tesla’s promises around AI and automation—particularly when a firm timeline or large production volumes are involved—should be taken with a pinch of salt.

Xpeng’s Iron humanoid robot serves as China’s answer to Atlas and Optimus, built on a fully self-developed, in-house chip and software stack designed for agility and cost efficiency rather than raw physical capability. The robot is intended for more public-facing, arguably flashier, use-cases than heavy manufacturing—for instance, customer-facing retail roles like showroom assistance. Hyundai’s Atlas, by contrast, is being deployed cautiously into structured, back-of-house industrial tasks first, reflecting Boston Dynamics’ decade-plus head start in mechanical engineering and control theory.

The Georgia deployment in 2028 is where that hardware-first bet will be tested against reality. Whether Atlas can move from controlled demonstrations to reliable performance on a live production line will determine whether the leap to genuine component assembly by 2030 is achievable, or whether Hyundai’s hefty investments amount to little more than expensive hot air.