LG Wants a Walking Robot by 2027. First, a Wheeled One Goes to Work in Tennessee.

LG's wheeled CLOiD service robot is heading to a Tennessee washing-machine line later this year, ahead of a bipedal humanoid the same company announced in an August 14 memorandum of understanding with NVIDIA, per a Memeburn report. The MoU, a non-binding agreement to co-develop the walking robot for early 2027, is the headline. The factory line is the proof point.

The walking robot will use three pieces of NVIDIA's physical-AI stack, a category that means AI acting in the physical world, on factory floors and in homes, rather than only on screens. Isaac GR00T, a foundation model for humanoids released at NVIDIA's GTC 2026, handles reasoning and generalization across tasks a robot has not been explicitly programmed for. Jetson Thor is the onboard computer: it processes sensors, runs AI inference, and controls motors in real time across dozens of actuators. Halos is the safety system, the software layer designed to let robots operate next to human workers without cages or separate cells. The walk-through matters because Halos is the only piece that decides whether a robot can stand next to a person on a factory floor without becoming a safety incident.

CLOiD is a wheeled robot that LG has shown in service and logistics roles, not a humanoid, and the Tennessee deployment is a dated, physical proof of concept: a robot doing real work in a real factory, on a real calendar. It is the kind of fact a reader can verify next quarter, not a roadmap number 16 months out. Consumer-electronics assembly is also the right proving ground: the work is repetitive, the tolerances are tight, and the throughput math makes a robot's capital cost easier to defend than in lower-margin industries.

Halos is the load-bearing piece of the 2027 promise. A bipedal humanoid is easy to announce and hard to ship; the harder question is what software lets it stand next to a person on a factory floor without becoming a safety incident. NVIDIA's Halos is positioned as that software. The MoU names Halos as part of the stack. The Tennessee deployment does not yet commit to it.

The MoU also covers "AI factory" development. NVIDIA's DSX platform, a reference design for large-scale AI infrastructure, pairs with LG CNS's PhysicalWorks, per the same Memeburn report. That is a separate, longer-cycle bet on the data-center side of physical AI, where the model gets trained, not where the robot works. The factory-floor story and the AI-factory story share a vocabulary but are different purchases: one is a near-term line item, the other is a multi-year infrastructure bet. A reader who treats them as one announcement will misread both.

A non-binding memorandum is exactly what it sounds like, an agreement to work toward a shared goal without the contract law of a supply deal. "Targeting early 2027" is a roadmap number, not a ship date. GR00T, Thor, and Halos are real products at NVIDIA, but a humanoid that uses all three is a derivative product that has not been built, priced, or deployed. The source material frames this against NVIDIA's $500 billion AI infrastructure fund, but a fund commitment and a shipped robot are not the same kind of fact, and a reader should hold the 2027 target at the same weight they would hold any partner announcement: worth tracking, not worth pre-ordering.

The near-term beats to track. Did CLOiD start work on the Tennessee washing-machine line in 2026, or slip? Did the deployment use Halos, or a different safety stack? LG has CES 2027 as the natural venue to show a Halos-equipped bipedal prototype; whether it does is the next beat. And whether NVIDIA names another consumer-electronics manufacturer as a Halos reference customer will tell readers whether LG is a one-off or the first of a wave.

If the Tennessee CLOiD line slips into 2027, or ships without Halos, the 2027 bipedal loses its license-to-operate argument. A walking robot that works next to people needs a safety story. Without one, it is a press release.