LG Electronics will manufacture its next-generation AI home appliance chips at Samsung Foundry, routing on-device inference silicon for connected white goods through its Korean rival's contract-manufacturing arm. The decision, reported by The Elec, shifts AI accelerator production away from general-purpose mobile SoC suppliers toward a domestic foundry partner.

The agreement covers neural processing units and related AI silicon designed for LG's ThinQ platform — refrigerators, washing machines, air conditioners and ovens equipped with voice control, computer vision and predictive-maintenance features. LG had previously relied on third-party mobile application processors for compute workloads in these products.

Why a Korean foundry for Korean appliances?

Three structural factors drive such sourcing decisions. Domestic supply-chain resilience became a procurement priority for Korean OEMs during the 2020-2022 semiconductor shortage. Samsung Foundry offers 5nm, 4nm and 3nm gate-all-around (GAA) process nodes with capacity sized for consumer-electronics volumes. Co-design between appliance firmware teams and foundry process engineers enables die-size, power-envelope and feature-set optimization unavailable from off-the-shelf mobile SoCs.

Samsung Foundry's Korean fab footprint — Giheung, Hwaseong and Pyeongtaek — places production within reach of LG's appliance R&D centers, simplifying tape-out logistics and engineering hand-offs.

What Samsung Foundry gains

The order extends Samsung Foundry's Korean customer book at a time when the division has lost leading-edge share to TSMC. Samsung Foundry operates the Giheung, Hwaseong and Pyeongtaek fabs in South Korea and an Austin, Texas facility. Its existing customer roster spans smartphone SoC and AI accelerator programs. Adding LG Electronics diversifies the foundry's order book beyond smartphone and data-center silicon — categories that have absorbed most of Samsung's foundry customer-acquisition budget.

The appliance category also carries different volume profiles from smartphone or HPC silicon: longer product lifecycles, lower annual die counts per SKU, and tighter thermal and cost constraints. That profile suits Samsung's mature and mid-range nodes, not only its leading-edge line.

What remains with LG's in-house unit

The agreement does not affect LG Siliconworks, the group's power-management IC subsidiary listed on KOSDAQ. LG Siliconworks continues to supply voltage regulators, LED drivers and related analog parts across LG's appliance and display product lines.

The Samsung Foundry deal covers the AI accelerator die only. Wi-Fi and Bluetooth combo ICs, microcontrollers, display driver ICs and image sensors will likely continue to come from Broadcom, Infineon, NXP, MediaTek and similar suppliers. Foundry contracts fabricate customer-owned designs — IP and architecture remain with LG.

Strategic context

The move reflects a wider industry shift: appliance makers commission custom AI silicon rather than retrofit smartphone-derived application processors. On-device inference cuts cloud round-trips, lowers latency for computer-vision workloads such as refrigerator food recognition, and reduces standby power in always-connected products. Domestic foundry access shortens iteration cycles and gives LG priority wafer allocation when industry-wide capacity tightens.

LG's ThinQ brand serves as the umbrella for its AI appliance line, integrating on-device voice processing and internal sensors for food inventory, wash-cycle recognition and energy optimization. A purpose-built NPU allows those workloads to run locally under power budgets that smartphone-derived SoCs cannot match at appliance price points.

For Samsung Foundry, the validation matters as much as the revenue. A second-tier Korean OEM placing AI silicon at Samsung signals that the foundry's design-rule ecosystem, PDK maturity and physical IP library can serve markets beyond flagship smartphone SoCs and data-center accelerators. That positioning will face tests as more Korean and Japanese appliance and automotive buyers evaluate foundry options through 2026.