Summary
LG Electronics is moving deeper into the infrastructure layer of the AI investment cycle.
On September 28, the company joined the NVIDIA Partner Network as a Power and Cooling Solution Preferred Partner, following NVIDIA infrastructure qualification of its 600kW, 1MW and 2.6MW coolant distribution units. The 2.6MW model has also qualified as an NVIDIA DSX Ready CDU.
The commercial evidence is becoming more important than the certification itself. LG disclosed more than KRW 600 billion of AI data center cooling orders in the first half of 2026, with production already underway.
The investment question is now whether that order momentum can become material Eco Solution earnings and eventually change the valuation framework applied to LG Electronics as a whole.
Key Takeaways
- ✓ LG has progressed from product qualification to NVIDIA Power and Cooling Solution Preferred Partner status, giving its cooling portfolio greater access to the NVIDIA AI infrastructure ecosystem.
- ✓ More than KRW 600 billion of AIDC cooling orders in 1H26 shows that the opportunity has moved beyond technology validation, although orders should not be confused with recognized revenue.
- ✓ Valuation is now the harder part of the thesis. At the September 28 close, LG trades at roughly 17.0x 2026 consensus earnings, while the existing broker target-price consensus largely predates the latest NVIDIA developments.
What Happened
LG Electronics has officially joined the NVIDIA Partner Network as a Power and Cooling Solution Preferred Partner, strengthening its position in thermal-management infrastructure for high-density AI data centers.
LG says participation in the network broadens its customer touchpoints and allows its solutions to be considered earlier in facility planning and design.
The technical progression is also becoming clearer.
LG has completed NVIDIA infrastructure qualification for its 600kW, 1MW and 2.6MW CDUs. The 2.6MW model additionally meets the requirements for NVIDIA DSX Ready CDU status and was, at the time of LG’s announcement, the highest-capacity CDU among DSX Ready CDU partners listed on the NVIDIA marketplace.
LG plans to pursue NVIDIA qualification for a 4MW CDU by year-end.
These milestones improve technical credibility and market access.
They do not guarantee future orders.
Why It Matters
The AI infrastructure cycle is creating a thermal-management bottleneck alongside the better-known constraints in semiconductors and electricity.
As computing density increases, individual racks generate substantially more heat. LG notes that rack power density in next-generation AI infrastructure is already reaching hundreds of kilowatts, driving demand away from conventional rack-level cooling toward centralized MW-class cooling systems.
That transition expands the economic value of cooling infrastructure.
LG is not approaching the opportunity through a single CDU product. Its “Chip-to-Chiller” portfolio includes liquid-cooling equipment, cold-plate infrastructure, CRAHs and large-scale chillers, allowing it to address thermal management from the processor level through the broader data center facility.
For investors, this creates a different earnings exposure from LG’s traditional consumer businesses.
Appliances and TVs remain important, but AI data center cooling links LG to a multi-year infrastructure capital-expenditure cycle driven by compute deployment rather than consumer replacement demand.
The NVIDIA relationship matters, but it should not be mistaken for the investment thesis.
The stronger evidence is commercial.
LG secured more than KRW 600 billion of AIDC cooling orders in the first half of 2026 and said production was already underway. Management is aiming to secure projects worth trillions of won by year-end, although that remains a target rather than contracted revenue.
The next hurdle is earnings conversion.
LG must show that order intake can continue scaling, that those orders convert into Eco Solution revenue at attractive margins, and that AIDC cooling becomes large enough to influence consolidated earnings.
Only then does the valuation argument become structural.
NVIDIA validation lowers one barrier to entry. It does not remove the need for execution.