LG's humanoid is a 2027 promise; the Tennessee wheeled robot is the 2026 fact

LG and Nvidia signed a memorandum of understanding at Nvidia's Santa Clara headquarters on 13 August covering humanoid robots, AI factory infrastructure and vehicle computing, with LG Corp chairman and CEO Kwang Mo Koo and Nvidia's Jensen Huang both present ^[1][2]. The headline item is a bipedal humanoid due for its first public appearance in the first quarter of 2027, running Nvidia's Jetson Thor for onboard processing, reasoning and motion control, the Isaac GR00T foundation model, and the Halos safety system ^[3][4].

The item with a job attached is earlier and less photogenic. LG will put CLOiD, its wheel-based robot, on the production line at its washing machine factory in Tennessee before the end of 2026 ^[5]. Trial data feeds an on-site robot data factory built on LG CNS's PhysicalWorks platform, which collects operational data, generates synthetic datasets and supports training, testing and verification ^[6], and LG says results will improve performance before deployments expand to other factories, commercial facilities and homes ^[7]. On the stated dates, CLOiD is working a real line before the humanoid is shown to anyone ^[8].

The division of labour is the substance here. Nvidia supplies the compute, the model and the safety layer ^[4]; LG Electronics and LG Innotek supply actuators and sensors, and LG Energy Solution supplies batteries ^[9], which The Next Web notes keeps most of the bill of materials inside the group ^[10]. That is a good position for a conglomerate that already builds motors and cells at volume, and a better one for the vendor selling the brain to everybody. Nvidia has signed up Japan's robotics establishment for its open world models ^[11] and has been courting manufacturers beyond China's Unitree ^[12], which The Next Web reports has been outselling Tesla on humanoids ^[13]. LG is not even unique at home: Samsung has a CEO-led robotics division chasing the same market, and Hyundai owns Boston Dynamics ^[14][15].

The compute build-out trails the robots. An AI factory reference site on Nvidia's Vera Rubin platform is due in the first half of 2027, combining Nvidia's DSX architecture with LG's power systems, thermal management, data centre design and operations ^[16][17]. An 80 MW facility at Cheonan in South Korea follows in the first half of 2028, roughly a year later, to support physical AI systems and robot foundation models ^[18][19]. LG's stated ambition is to validate an end-to-end design, construction and operations package it can then sell as turnkey infrastructure to other technology customers ^[20]. Nvidia's DRIVE Hyperion goes into an LG computing platform for AI-defined vehicles alongside LG's infotainment and automotive software ^[21], and a joint task force of technical and business staff will oversee research, factory validation and commercialisation across all three tracks ^[22]. Koo said the tasks and goals in AI factory and physical AI "have become clear" and that LG will build industry-leading references ^[23].

Watch the safety layer, because it is the part with legal consequence. Nvidia describes Halos as the first full-stack safety framework built for robots ^[24], and Huang framed the partnership around machines that can "understand the real world, reason and act safely alongside people" ^[25]. Everything else in the announcement is dated 2027 or 2028 ^[26]. The near-term test is narrow and checkable: whether a wheeled robot in Tennessee holds a place on a washing machine line for a full year, and whether the data it produces makes the 2027 machine anything more than a demonstration.