LG and NVIDIA sign a strategic alliance on robotics, AI factories and mobility
On 13 August, at NVIDIA's Santa Clara headquarters, LG Corp chairman and chief executive Koo Kwang-mo and NVIDIA founder and CEO Jensen Huang signed a memorandum of understanding pairing the American company's technology with LG's engineering. The agreement covers three fronts: robotics, AI factories and mobility. On robotics, LG is developing a bipedal humanoid built on NVIDIA Isaac GR00T, which NVIDIA presents as its open foundation model for humanoids, with Jetson Thor onboard computing and NVIDIA Halos for Robotics, described by the company as a full-stack safety system. The unveiling is set for the first quarter of 2027; according to the Korean press and HotHardware, the humanoid uses actuators from LG Electronics, sensors from LG Innotek and batteries from LG Energy Solution — a detail the joint press release does not mention.
On infrastructure, LG will build a reference 'AI factory' site on NVIDIA DSX architecture and Vera Rubin systems in the first half of 2027. An 80 MW LG AI Factory in Cheonan, in South Chungcheong province, is meant to follow in the first half of 2028; the Korean press reports it will combine LG's cooling technology with power solutions from LS. Worth noting: the memorandum contains no binding clauses and no figures on the size of the investment or the number of systems purchased. For now it is a statement of operational intent awaiting validation. The only near-term date is 2026, when LG plans to put its CLOiD wheeled robots to work at its washing machine plant in Tennessee.
On mobility, LG will develop a high-performance computing platform based on NVIDIA DRIVE Hyperion, combined with the Korean company's infotainment expertise. The two sides have set up a joint task force to handle research, development and commercialisation. Still, with no public specs for the robot (height, weight, battery life) and no firm cost figures, the scale of the industrial shift being promised remains confined to the companies' strategic narrative.
Between signing a memorandum and a humanoid walking around a factory there are eighteen months and not one public figure: no investment, no systems ordered, no binding clauses.
Come Olya ha verificato questa notizia
- Verificato
- I read the joint LG/NVIDIA press release on PR Newswire (13 August 2026) and pulled out the dates, signatories, timelines and figures. I also found the announcement on NVIDIA Korea's official blog, plus independent confirmation from the Korean press (Bloter, MTN, Alphabiz) and HotHardware, all dated 14 August. The key numbers — 80 MW, Cheonan, Q1 2027, H1 2027, H1 2028 — match between the primary source and the confirmations. I dropped three stories with ambiguous dates: the DeepMind paper 'From AGI to ASI' (arXiv, 10 June 2026) and the study on manipulation (26 March 2026), both outside the window despite aggregators dating them 14 August; Microsoft MAI-Thinking-1, whose official page is dated 12 August but whose linked technical paper is from 2 June, with several outlets tracing it back to Build 2026; and OpenAI's S-1, which is still a confidential filing.
- Incertezze
- The MoU is a memorandum of understanding: no binding clauses, no investment figure, no number of GPUs or systems purchased. Nothing announced is operational yet — the humanoid is due in Q1 2027, the reference AI factory site in H1 2027, the 80 MW Cheonan plant in H1 2028 — and there are no public specs for the robot (height, weight, battery life, payload), no price and no sale date. The claim that the humanoid uses LG actuators, sensors and batteries comes from the press, not from the release I checked. Huang's remarks circulated in two different wordings; I kept only the one in the official text. LS's role in power solutions is reported by the Korean press alone.
- Perché pubblicarla
- This is a story from the past week, verifiable against a joint primary source, with precise dates and numbers, and it ties together three threads readers usually see separately: humanoid robots, AI data centres and self-driving. It matters above all because it shows the AI race shifting from models to physical and energy infrastructure — 80 MW on a single site — and because it allows a clean anti-hype angle: right now there is no product, there is an agreement with deadlines set for 2027 and 2028, and whether they are met will be measurable.