LG Corp. said on August 13, 2026, that it is developing a next-generation bipedal humanoid robot built on NVIDIA’s Isaac GR00T foundation model, targeting a public unveiling in the first quarter of 2027. The announcement came alongside a memorandum of understanding signed at NVIDIA’s Santa Clara headquarters, with LG Corp. Chairman and CEO Kwang Mo Koo and NVIDIA founder and CEO Jensen Huang in attendance, formalizing work across robotics, AI factories, and future mobility, according to LG’s announcement.
The humanoid is the headline item, but the agreement reaches well past a single robot. LG plans to run its wheel-based CLOiD robot on a live washing-machine production line in Tennessee before the end of 2026, build an NVIDIA Vera Rubin-powered AI factory reference site in the first half of 2027, and develop an AI-defined vehicle computing platform on NVIDIA DRIVE Hyperion. It is the third announced step in a partnership that has moved from a stated direction to a signed scope of work in a little over two months.
The Humanoid Runs on NVIDIA’s Open Robot Foundation Model
The robot LG described is bipedal and carries its compute onboard: NVIDIA Jetson Thor handles reasoning and control, while Isaac GR00T supplies the foundation model underneath. NVIDIA pitches GR00T as an open, customizable model for generalized humanoid reasoning and skills, and LG said it will actively use GR00T while continuing to develop its own in-house robot foundation model, which it calls an RFM. The data flowing back from LG’s manufacturing validations is expected to feed that in-house effort.
The build also pulls in NVIDIA Halos for Robotics, the safety architecture NVIDIA introduced as a full-stack system unifying AI compute, system software, sensor data, and inspection for robots operating around people. For a machine LG says is headed for factories, homes, and commercial spaces, the safety stack is doing real work in the specification — bipedal platforms operating near human workers carry certification and liability questions that a validated safety architecture is meant to answer before deployment, not after.
On hardware, LG plans to combine what it calls “One LG” capabilities across LG Electronics, LG Innotek, and LG Energy Solution, covering actuators, sensors, and batteries. That is the structural pitch: few conglomerates own a battery maker, a component supplier, and a consumer products company under one roof, and LG is betting that vertical spread shortens the path from prototype to production.
A Signed Task Force Replaces the Blueprint Stage
The MOU lands on a visible timeline. NVIDIA CEO Jensen Huang first said publicly on June 8, 2026, that the companies were working together on humanoid robots and data centers. On June 8, 2026, the two companies announced a cooperation framework spanning mobility, AI infrastructure, and physical AI, including joint development work on the GR00T model. The August 13 agreement converts that framework into named programs with dates (making it the second announced step, roughly two months later), and the companies will operate a joint task force of technical and business staff covering R&D, site validation, and commercialization.
The nearest-term physical deliverable is not the humanoid but CLOiD, LG’s wheel-based robot, which goes onto LG Electronics’ washing-machine manufacturing line in Tennessee within this year for validation in a real production environment. LG said it will refine the robots’ performance and utility from the results and expand deployment to global production facilities, homes, and commercial spaces. A production-line validation is the kind of logged, repeatable setting where a robot vendor accumulates the intervention rates and uptime figures that separate a deployment from a demo — and LG now has a stated reason to generate those numbers.
The pattern mirrors a wider industry move toward building humanoids on shared foundation models rather than from-scratch stacks: Foundation Picks AMD Chips to Power Its Humanoid Robots covered another startup pairing an outside compute platform with its own humanoid hardware, while Google has taken the model-first route with Gemini Robotics ER 2, a robotics reasoning model designed to run across robot bodies. LG’s announcement is unusual in pairing an open NVIDIA model with a parallel in-house model program, and in saying so openly.
The AI Factory Build Reaches 80 MW by 2028
The agreement’s second track is infrastructure. LG plans an AI factory reference site built on NVIDIA’s DSX architecture and powered by Vera Rubin systems in the first half of 2027, followed by an 80 MW LG AI Factory in Cheonan, Chungcheongnam-do, targeted for the first half of 2028. The Cheonan facility is aimed at advancing physical AI and LG’s robot foundation models, and LG said it intends to package the validated design, construction, and operations stack as a turnkey “One LG” offering for global big tech customers.
The third track is automotive: a high-performance computing platform for AI-defined vehicles built on NVIDIA DRIVE Hyperion, integrating LG’s in-vehicle infotainment and software work and extending its vehicle-components business into autonomous driving. That program carries no stated delivery date in the announcement.
What the Agreement Commits To, and When
- Q1 2027: public unveiling of the bipedal humanoid
- Within 2026: CLOiD validation on the Tennessee washing-machine line
- First half of 2027: NVIDIA Vera Rubin AI factory reference site
- First half of 2028: 80 MW LG AI Factory in Cheonan, Chungcheongnam-do
- Undated: the DRIVE Hyperion vehicle platform and the commercialization work of the joint task force
No pricing, production volumes, or deployment targets for the humanoid appear in the announcement, and the robot has no public name yet. What LG committed to on August 13, 2026,
