Why LG Is Switching From TSMC to Samsung Foundry

A Foundry Flip You Wouldn’t Expect

LG Electronics has spent years building its custom chip portfolio through TSMC. Now it is developing two new IoT system-on-chip designs for its AI home business at Samsung Foundry instead. The shift is small in absolute terms but structurally notable. A company known as a TSMC regular is sending design work to Samsung — and it is not doing so alone.

The project runs from July through December 2030 under a government-backed program called K-OnDevice AI Semiconductor Technology Development. LG is providing the application specifications and setting the product direction. CoAsia Semi is handling overall chip design and optimizing the design for Samsung Foundry’s process. People from CoAsia Nexell, a fabless design affiliate within the CoAsia group, are also supporting the work. The development budget is 31 billion won.

The chips target what LG calls its next-generation AI home platform. The goal is an always-on sensing SoC that lowers power consumption and cost compared with current approaches. That sounds incremental until you consider who is making it and where.

CoAsia Semi is a South Korean fabless company that has carved out a niche designing application processors and multimedia chips for consumer electronics. Its affiliation with Nexell gives it a track record in high-volume IoT and media processing silicon. LG’s decision to route this project through CoAsia rather than retain an external Western fabless house or build internal capability reflects a deliberate choice to keep the design IP closer to home and lean on a partner already fluent in Samsung’s process characteristics. The arrangement cuts down on learning-curve risk at Samsung Foundry, which has historically struggled with design-win conversion outside its core DRAM and memory business.

Why This Matters Beyond One Project

Samsung Foundry has spent years trying to prove it can win design wins outside its existing customer base. Every large appliance maker that sources chips from TSMC and shifts a project to Samsung is a signal the foundry wants to broadcast. It is not just about the revenue from one chip family — it is about breaking the assumption that appliance-grade chips default to Taiwan.

An industry source told The Elec that even for national projects, LG historically chose TSMC fabs. That makes this arrangement noteworthy. Government funding usually follows precedent, not breaks it. LG choosing Samsung here suggests the foundry’s process roadmap or engagement strategy has shifted enough to change a habitual decision.

The broader K-OnDevice program is roughly 800 billion won and runs through 2030. It covers automotive, IoT, home appliances, machinery, robotics, and defense. Hyundai Motor, Doosan Robotics, Daedong, and Korea Aerospace Industries are demand-side participants alongside LG. If LG’s participation normalizes Samsung Foundry for appliance chips, other participants may follow the same path. The domino effect is the real story.

Samsung Foundry has been aggressively courting foundry customers since it lost its internal smartphone chip monopoly when Apple switched to TSMC for the A-series processors. The company has poured resources into process technology, packaging capabilities, and design-enablement services, but design wins in non-memory segments have remained elusive. Winning over LG — a company with enormous purchasing power across its appliance, EV battery, and display businesses — would be a significant credibility marker. It would demonstrate that Samsung can deliver on process commitments for real products, not just promise them on paper.

Who Wins and Who Loses

Samsung Foundry wins visibility and a reference design win in a segment where TSMC has held strong positioning. Appliance SoCs are not the most advanced nodes, but they are high-volume, cost-sensitive, and deeply embedded in supply chains. Once a product family qualifies on a foundry process, qualification is expensive to move. That creates stickiness.

CoAsia Semi wins a government-subsidized development runway and a high-profile design-in. CoAsia Nexell gains project experience that can be reused for other appliance and IoT clients. LG wins a lower-cost, lower-power SoC path for its AI home platform without building an in-house design organization from scratch.

TSMC loses a potential design win and, more importantly, loses narrative control. The company has long been the default foundry for Korean electronics firms beyond smartphones. Each project that redirects to Samsung erodes that default status. TSMC still holds the performance lead at advanced nodes, but appliance and IoT chips are not won on peak performance. They are won on power efficiency, cost, and relationship trust. That is where Samsung is pressing.

The loss is not merely financial. TSMC’s dominance in Korea has been a cornerstone of its international foundry reputation. Korean companies have been among its earliest and most loyal customers. When those customers begin experimenting with alternatives — especially on government programs where cost is less of a constraint than strategic diversification — it sends a message to other potential clients that loyalty to TSMC is not irreversible.

The Unusual Partnership Structure

The collaboration model itself is worth watching. LG does not have a fabless chip design arm. CoAsia Semi does. Instead of LG building internal capability or contracting a Western fabless firm, it partnered with a domestic player whose engineering team already understands Samsung’s process characteristics. That is efficient, but it also means Samsung Foundry gains indirect influence over an appliance giant’s silicon strategy through a partner rather than through direct engagement.

This pattern — government program, conglomerate client, domestic fabless designer, local foundry — mirrors how Korea has structured semiconductor development for years. What is different is the foundry choice. In previous national projects, LG used TSMC. The fact that Samsung is now in the room suggests the foundry has become competitive on process support, not just price.

The arrangement also raises questions about intellectual property flow. CoAsia Semi will hold design ownership for the new SoCs, and LG will license them for use in its AI home products. That structure gives CoAsia a revenue stream and a foothold in a market it could not easily enter alone. But it also means LG is ceding some strategic control over its silicon roadmap to a partner whose long-term ambitions may diverge from its own.

Second-Order Effects on the Korean Ecosystem

The implications extend beyond the three companies directly involved. Korean policy makers have spent a decade trying to rebuild a domestic semiconductor ecosystem after the industry contracted in the 2000s. The K-OnDevice program is part of that effort, and every design win that stays within Korean borders strengthens the case for continued public investment. A project that routes through TSMC — however efficiently — leaves design expertise and fab revenue abroad. A project that routes through Samsung Foundry keeps more of the value chain domestic.

For the wider fabless community, LG’s choice signals that Samsung Foundry is willing to take on projects from companies that might otherwise default to Taiwan. That lowers the perceived risk for other Korean designers considering Samsung as an alternative. If CoAsia can deliver a working SoC on Samsung’s process, other fabless firms will have a reference point. The result could be a gradual shift in where Korean design activity concentrates, with downstream effects on employment, university programs, and venture capital flows in the semiconductor sector.

Samsung Foundry’s broader competitive position against TSMC has been a topic of industry debate for years. The gap at leading-edge nodes remains significant, but foundry business is not won solely on process leadership. Packaging, test infrastructure, customer service, and process maturity at mature nodes all matter. An appliance SoC may seem far from the cutting edge, but it tests the same fundamentals: yield, reliability, cost per unit, and the ability to support a customer through design-in and volume production. Success here builds organizational muscle that transfers to other segments.

What Comes Next

The first chips are expected to ship under the AI home platform timeline, with development running through 2030. LG will likely announce product integration as the SoC matures. Samsung Foundry will use this as a design win case study in its marketing. CoAsia Semi and CoAsia Nexell will seek additional appliance and IoT clients using the same relationship model.

For TSMC, the risk is not a single lost project. It is the cumulative effect of appliance makers testing Samsung Foundry on government programs and then qualifying those processes for commercial production. Once qualification happens, switching costs rise. The window for Samsung to deepen its appliance and IoT foothold is open now, and this partnership is one of several moves in that direction.

The semiconductor industry talks a lot about advanced nodes and leading-edge process competition. But foundry business is also won in the mundane segments — IoT chips, sensor SoCs, appliance controllers — where volume and relationship matter more than peak performance. LG’s choice to route one project through Samsung Foundry is a quiet signal that the battleground is widening.

Korea’s semiconductor strategy has long been dominated by the Samsung-LG duopoly in memory and displays. This project represents a subtle reconfiguration: a foundry rivalry playing out inside a government program, with an appliance maker as the swing voter. The chips being designed are not the most exciting products in the industry. But the alliances they reflect are. The real story is not what LG is building. It is who LG is choosing to build it with.