Nvidia and LG Group are building a joint AI factory that will underwrite LG's push into robotics, autonomous driving, data center infrastructure and GPU cloud services, the two companies said on June 7, 2026. The deal pulls six LG subsidiaries — LG Electronics, LG Innotek, LG CNS, LG Uplus, LG Energy Solution and LG AI Research — onto a single Nvidia-aligned stack covering model training, simulation, edge deployment and factory-scale digital twins.
The infrastructure layer is built on the Nvidia DSX AI factory platform. LG Uplus, the group's telecom arm, plans to construct a large-scale AI data center sized for the latest Nvidia GPUs, while LG CNS will build power-efficient AI factories on the same DSX reference design. LG Energy Solution is signing on to develop 800 volt-direct-current data center energy solutions aligned to Nvidia's BESS Self-Qualification guidelines, the power architecture Nvidia has flagged as necessary for next-generation accelerators.
The announcement lands a day after Nvidia disclosed a parallel expansion with Doosan Group on robotics, power and AI factory materials, which we covered yesterday. Taken together, the two deals show Jensen Huang stitching Korea's largest industrial conglomerates into a single physical-AI supply chain, with Nvidia silicon and software at the center.
Key facts
- 01Nvidia and LG Group are building a joint AI factory spanning robotics, autonomous driving, data centers and GPU cloud services, announced June 7, 2026.
- 02LG Energy Solution will collaborate with Nvidia on 800 volt-direct-current data center energy systems aligned to Nvidia's BESS Self-Qualification guidelines.
- 03LG AI Research trained its EXAONE sovereign model on Nvidia Blackwell GPUs using NeMo and Nemotron datasets, deployed via TensorRT-LLM.
- 04LG Electronics is integrating Isaac Sim, Isaac Lab and the Isaac GR00T vision-language-action model into its CLoiD home robot program.
- 05LG Uplus plans a large-scale AI data center built on the Nvidia DSX platform to host next-generation GPUs.
On the robotics side, LG Electronics is wiring Nvidia Isaac Sim and Isaac Lab into the development pipeline for CLoiD, its home cobot, and exploring the Isaac GR00T reasoning vision-action-language model for both household and modular industrial robots. The two companies plan to jointly develop reference robots that ship as part of the GR00T ecosystem.
“LG Electronics is developing home-based robots like CLoiD to help with a wide range of indoor household tasks, enhancing everyday convenience and improving quality of life.”— Madison Huang, Nvidia
LG Electronics is also standing up what it calls a physical AI data factory — a facility that converts GPU compute into synthetic training data for robotics and industrial AI, using Nvidia Cosmos world foundation models for generation and augmentation. The pitch is that Korean and global robot makers can buy data the way they currently buy compute, sidestepping the bottleneck that real-world robot fleets impose on training.
LG Innotek is contributing the hardware layer, supplying sensing components engineered specifically against Nvidia's GPU architecture and development environment. The same subsidiary plans next-generation autonomous-driving components — sensing, connectivity and lighting — built around Nvidia silicon.
LG CNS is targeting the factory floor and the warehouse. The systems integrator is folding Isaac frameworks, Cosmos world models and GR00T into its PhysicalWorks industrial robot platform, with the goal of making AI robots installable by buyers who are not Nvidia developers themselves.
“LG CNS is building an ecosystem that enables anyone to easily adopt AI robots in manufacturing and logistics sites.”— Madison Huang, Nvidia
Mobility is the third pillar. LG Electronics is aligning its advanced driver-assistance and in-vehicle AI systems with the Nvidia DRIVE Hyperion reference architecture, and plans to deploy Nvidia DRIVE AGX for AI cockpits and edge processing in future vehicles. The work targets LG's automotive electronics customers — the global OEMs that buy LG's instrument clusters, infotainment systems and ADAS modules.
The sovereign AI piece runs through LG AI Research and EXAONE, one of Korea's flagship open-model families. LG AI Research trained EXAONE on Nvidia Blackwell GPUs using the NeMo framework and Nemotron open datasets, then optimized inference with TensorRT-LLM. LG Group is rolling out the model internally through ChatEXAONE, its enterprise chatbot, and exploring agentic deployments across business units.
What is conspicuously absent from the announcement is dollar value. Neither company disclosed the capital commitment, the GPU allocation, the data center site, the timeline for first deployment or revenue-sharing terms on the GPU cloud services. The DSX-based data center plans from LG Uplus and LG CNS are described as plans, not builds, and there is no public power-purchase agreement or grid commitment behind the 800V-DC work at LG Energy Solution.
Strategically, this is Nvidia continuing to convert national industrial champions into long-term GPU consumers by giving them an end-to-end stack — silicon, networking, simulation, world models, robot foundation models, automotive compute, sovereign LLM tooling — that is internally consistent and externally hard to swap out. LG gets an answer to the question every conglomerate is now being asked by its board, which is what its AI strategy actually is. Nvidia gets six more customers locked into Blackwell and its successors, a Korean sovereign-AI showcase to point regulators at, and another piece of the physical-AI flywheel where robots train in Cosmos, deploy on GR00T and run on DRIVE. The deals with Doosan and now LG suggest Korea is becoming the second national market, after Taiwan, where Nvidia is operating less like a chip vendor and more like industrial infrastructure.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
