Nvidia has added Samsung Electro-Mechanics to its semiconductor package-substrate supply chain, tying a multiyear purchasing agreement to expanded manufacturing in South Korea. The partnership brings the Korean component maker into a critical layer of Nvidia’s AI hardware network as demand for accelerators and data-center networking equipment continues to strain advanced packaging capacity.
The companies announced the strategic partnership on October 21 following Nvidia chief executive Jensen Huang’s keynote at GTC 2026 in Berlin, according to Korean media reports. Samsung Electro-Mechanics plans to begin supplying next-generation substrates for computing chipsets and data-center networking products in 2026. The agreement covers multiple years, although the companies did not disclose its financial value or annual shipment volumes.
Dedicated capacity in Sejong and Busan
Initial deliveries will draw on Samsung Electro-Mechanics’ existing domestic facilities. The company will also establish dedicated AI semiconductor package-substrate production lines at its Sejong and Busan operations, while the wider cooperation supports equipment upgrades and new factory construction. This approach allows shipments to begin before the full expansion is completed.
- Supply begins
- 2026
- Production locations
- Sejong and Busan, South Korea
- Products
- Substrates for computing chipsets and data-center networking
- Contract term
- Multiple years
Package substrates connect semiconductor dies with the larger circuit board while carrying power and high-speed electrical signals. Their manufacturing becomes more demanding as AI systems combine increasingly powerful processors, memory and networking components. Securing additional substrate capacity can therefore help Nvidia increase finished-system output by reducing pressure at a specialized stage between chip fabrication and final assembly.
Samsung Electro-Mechanics expects its substrate shipments to Nvidia to grow at an average annual rate exceeding 10 percent. By 2028, production allocated to Nvidia is expected to account for a substantial portion of the Korean company’s annual package-substrate capacity, according to The Korea Times. The available evidence does not specify that portion or provide a factory-by-factory investment figure.
A broader expansion of advanced packaging
The agreement forms part of a wider Nvidia manufacturing expansion rather than a standalone sourcing change. Nvidia has said cooperation with Samsung Electro-Mechanics, Japan’s Ibiden and Taiwan’s Unimicron is intended to increase substrate capacity for its products by 2.5 times over the next three years. Diversifying among established Asian suppliers also reduces reliance on a narrower production base as Nvidia expands current and next-generation platforms.
Nvidia is separately working with Amkor Technology and Korean partners to increase advanced semiconductor packaging capacity. The plan includes raising interposer wafer capacity by more than 300 percent by the end of 2028. Interposers and package substrates perform different functions, but both are essential to assembling complex AI processors and moving data rapidly among compute, memory and networking components.
For Samsung Electro-Mechanics, the partnership offers a long-duration customer relationship in one of the fastest-growing areas of the electronics component market. President Chang Duck-hyun said the company would participate as a core infrastructure partner to Nvidia and provide comprehensive solutions through its manufacturing competitiveness and technology. Execution will depend on how quickly the Sejong and Busan lines are upgraded, when new capacity qualifies for mass production, and whether allocated output reaches the scale projected for 2028.