LG is now in talks with NVIDIA about the following areas:

After a meeting held in Seoul between LG’s CEO, Ryu Jae-cheol, and NVIDIA’s Senior Director of Product Marketing for Omniverse and Robotics, Madison Huang, the key operational factors necessary for the operation of sophisticated automated systems are becoming more evident.

Though no figures or timelines have been set for the financial investment by the two companies, the shared interests of the hardware and computing capabilities that both companies possess show the enormous investments that will be needed to take autonomous systems from simulations.

The density that comes with compute clusters that are used in building sophisticated machine learning models presents a physical challenge. While NVIDIA’s data center business makes record-breaking revenue, running the servers requires power levels beyond the capability of traditional cooling methods.

At CES 2026, LG was able to prepare its commercial teams to provide HVAC and thermal management solutions built specifically for artificial intelligence data centers. The power density becomes increasingly important as air-cooling becomes insufficient.

If server farms get too hot, the compute nodes will slow down, nullifying any gain made from advanced silicon investments. The use of LG thermal solutions as part of the NVIDIA infrastructure framework solves the problem of margin erosion. This enables data center managers to increase the computing density of their facilities without overheating the system.

This strategy makes LG an infrastructure vendor within the profitable technology ecosystem, allowing it to earn steady income from the enterprise sector by adding value to the compute stack instead of going against it. As LG aggressively expands into the realm of intelligent infrastructure solutions, it should be noted that LG CNS, an LG company, is sponsoring this year’s IoT Tech Expo North America.

Temporal and system-level inefficiencies between on-device inference and physical actuation

In addition to server architecture, the conversations aim at addressing the latency associated with autonomous consumer hardware. The success of LG’s future business model hinges on its ability to automate manual and cognitive tasks in the home environment.

LG launched CLOiD, a robot that comes with two arms, each capable of rotating on seven axes and five individual fingers per hand. The robot uses LG’s “Affectionate Intelligence” software that aims at contextual understanding and continuous environmental learning.

The translation of any computational instruction to physical action is reliant upon zero-latency inferencing. In the event that an articulated robot reaches out for a glass, the device should perform image recognition to identify the type of object, cross-check this against the vector databases locally stored in memory, and compute the exact amount of grip needed. Even the slightest error within this inferencing process will lead to physical destruction within the homeowner’s property.

As things stand at present, LG does not have the technology to develop the digital twin platform, manipulation models, and simulated environments needed to shrink the deployment pipeline safely. The platform needed here is provided by NVIDIA via its Omniverse and Isaac robotics platform that enables physical AI inference in real time.

Using the edge computing capability offered by NVIDIA, LG will be able to perform complex spatial calculations locally and hence drastically cut down on cloud computation costs linked to constant spatial mapping and video ingestions.

Systems that absorb massive user data and simulate real-world scenarios

NVIDIA is currently testing their robotics stack having completed a two-week Siemens factory test that took place in January 2026, which was only revealed at Hannover Messe in April.

During the test, an HMND 01 Alpha robot successfully performed live logistics tasks within eight hours. However, factories in Erlangen are highly structured environments. Domestic spaces are characterized by variability, shifting lighting, and unpredictability from human presence.

By accessing LG’s ThinQ ecosystem and their extensive production capacity, NVIDIA can provide data-rich training ground. To integrate robots into domestic space, it is necessary to train models on the realities of domestic variability, not just simulations.

Transitioning from industry application to consumer electronics will enable NVIDIA’s Omniverse platform to become the go-to platform infrastructure for real-world autonomy, similar to NVIDIA’s GPUs for cloud computing.

The last alignment opportunity pertains to the automotive space. The automotive components division of LG is among their rapidly growing business areas, producing in-vehicle infotainment systems, EV components, and cabin generative technologies that include gaze-tracking and adaptable displays. At the same time, NVIDIA’s DRIVE platform dominates the market in automotive autonomous and semi-autonomous computing

However, it is a known fact that automobile manufacturers frequently encounter difficulties with incorporating legacy infotainment systems into contemporary autonomous compute nodes. Since LG and NVIDIA both operate in adjacent layers in the same vehicle, the collaboration would allow for the combination of the interior experience layer from LG and the compute platform from NVIDIA. This way, the operators in the fleet would be able to integrate reference architectures without wasting time creating bespoke APIs and deploying machine learning updates via the air.

LG’s conversations with NVIDIA assist in determining the required hardware and computing capabilities for physical AI.