Samsung GaN Discrete Devices Fail Customer Benchmarks: Foundry Line Launches July

Samsung Electronics delivered gallium nitride (GaN) power semiconductor samples to multiple customers earlier this year and received no purchase orders in return, according to industry sources cited May 20 by The Elec. The chips failed to meet required quality and performance standards — specifically falling short on on-resistance targets — and never reached mass production. The company has simultaneously exited a government-backed electric vehicle development project and dropped plans to sell GaN modules. What remains is a foundry business it did not originally prioritize, now expected to begin operations as early as July.

GaN Discrete Devices Miss On-Resistance Targets

GaN, a compound semiconductor made from gallium and nitrogen, controls how electric current flows through a device. Samsung's 650-volt-class GaN parts are designed for compact fast chargers, data center uninterruptible power supply (UPS) units, and power supply units (PSUs). The central problem is on-resistance — denoted RDS — which measures the internal resistance present when current flows through the chip. Lower on-resistance means less wasted power and less heat, which translates directly to higher efficiency, the primary competitive advantage GaN holds over older silicon-based alternatives.

Multiple customers evaluated Samsung's GaN samples and found they did not hit the RDS benchmarks their designs required. No follow-on production orders were placed. The products remain in a pre-commercial state.

Samsung Drops Module Business, Exits Semipowerex Project

The performance shortfall was compounded by a strategic misread of how customers wanted to buy GaN. Samsung had initially planned to supply the chips packaged with other components as ready-to-use power modules — a format that simplifies integration for customers. The company subsequently stepped back from module supply and said it could provide only discrete devices. At least one potential customer described the consequence directly: "Customers wanted GaN products in module form, but Samsung said it could only supply the devices, which caused supply contracts to collapse."

The departure from the module business aligned with a broader strategic withdrawal. Since 2024, Samsung had been co-developing GaN chips for electric vehicle applications alongside power semiconductor module specialist Semipowerex under a government-funded project structured in two phases. Samsung completed only the first phase, which covered sample delivery; the discrete device development work in the second phase was never finished. A Samsung Electronics official confirmed the exit with a brief statement: "There has been a change in business strategy," while declining to specify the new direction.

TSMC Exit Opens Foundry Market Samsung Now Targets

The market Samsung is now entering as a foundry supplier has a significant structural vacancy. TSMC, until recently the dominant GaN foundry for power device makers including Navitas Semiconductor, announced in mid-2025 that it plans to exit the GaN wafer foundry business entirely by July 2027. The decision reflects TSMC's focus on advanced logic nodes and AI-driven packaging demand rather than the comparatively modest GaN market. The opening has left several customers seeking new manufacturing partners, and Samsung's forthcoming 8-inch GaN line is positioned to absorb some of that displaced demand.

TechInsights Technical Fellow Dr. Eugene Hsu identified TSMC's GaN exit as one of five developments that will define the power semiconductor competitive landscape through 2026. Samsung's foundry operations are handled by the Compound Semiconductor Solutions (CSS) unit, which sits inside the Device Solutions (DS) division alongside GaN discrete and silicon carbide (SiC) work, as well as LED manufacturing. CSS has secured multiple foundry customers and has confirmed initial production is expected no later than July. Industry sources also indicate that some companies currently placed with competing foundries are actively evaluating Samsung as a secondary or replacement source.

Low Mask Costs Make Multi-Foundry Strategy Viable

One structural feature of the GaN foundry market works in Samsung's favor: switching foundries is cheaper than in mainstream silicon. Each fab operates with distinct process recipes, so moving a design to a new foundry requires new photomasks — the templates used to print circuit patterns onto wafers. For advanced silicon chips, a full mask set can cost tens of millions of dollars, making customer loyalty to a single foundry an economic necessity. GaN photomask sets at compound semiconductor fabs typically run in the range of several hundred million Korean won — a fraction of the silicon equivalent. That cost structure makes it practical for customers to spread production across multiple foundries for supply security and pricing leverage, and it reduces the friction of sampling Samsung's new line.

Foundry Ramp Faces GaN-Specific Yield Challenges

Samsung's path to scale in foundry is not without technical constraints. GaN manufacturing differs from silicon in one particularly demanding respect: achieving consistent resistance characteristics across a wafer is harder to control, and yield management is correspondingly more difficult. The company can deploy more engineers or add production lines to expand output, but the global GaN power device market is estimated at roughly 1 trillion won (approximately $740 million) annually at current scale, which limits how aggressive a capacity build-out can be justified for a company of Samsung's size.

The more consequential growth question is longer-term. Market research projects the broader GaN semiconductor devices market will grow from roughly $4.13 billion in 2025 to more than $10.5 billion by 2031, a compound annual growth rate of nearly 17 percent. Electric vehicles, AI data center power architectures, and high-efficiency fast chargers are all pulling demand upward. Samsung's foundry bet, late and narrowly profitable in the near term, is ultimately a wager on being positioned when that growth arrives.