NVIDIA expands supplier commitments to add manufacturing capacity


NVIDIA says a new Kinsus agreement and an amended Samsung Electro-Mechanics deal support substrate production, while a Taiwan expansion with SPIL targets more dedicated packaging capacity by mid-2029.
NVIDIA said on 17 November 2026, that it had expanded supplier commitments to increase manufacturing capacity for its current and next-generation accelerated computing and superintelligence (SI) infrastructure platforms. The company is the sole source for the announcement.
The commitments include a new agreement with Kinsus Interconnect Technology for chip-substrate capacity and an amendment to an existing agreement with Samsung Electro-Mechanics. NVIDIA also said it and Siliconware Precision Industries, known as SPIL, are expanding advanced packaging operations in Taiwan.
A chip-package substrate supports a semiconductor die and provides electrical connections. Under the new agreement, Kinsus plans to expand its dedicated advanced ABF package-substrate capacity for NVIDIA at its Youth campus in Taoyuan, Taiwan, to more than 3.5 times current levels, NVIDIA said. Production is expected to begin in the second half of 2027, with the additional capacity targeted for the third quarter of 2028.
The Samsung Electro-Mechanics amendment is intended to improve capabilities at existing factories. It builds on a previously announced multiyear collaboration on substrate manufacturing in Sejong and Busan, South Korea, NVIDIA said.
Packaging protects and connects semiconductor dies so they can be integrated into electronic systems. NVIDIA said its multiyear agreement with SPIL, backed by multibillion-dollar NVIDIA funding, targets increased dedicated advanced packaging capacity by mid-2029.

The agreements add Samsung Electro-Mechanics to Nvidia’s substrate supply chain and reserve capacity with other partners, with expansion targets extending through 2028.
Before packaging, wafer fabrication creates semiconductor devices on silicon wafers. Advanced packaging can integrate multiple dies. These are distinct manufacturing stages, so capacity at one stage alone does not determine finished-system output.
Separately, NVIDIA said its previously announced July and October 2026 agreements with Amkor Technology total nearly $3 billion, including prepayments, to support packaging expansion in Arizona and South Korea.
NVIDIA said volume shipments of its Vera Rubin NVL72 and Vera CPUs began in the third quarter, using TSMC’s 5nm, 3nm and CoWoS technologies. Foxconn chairman Young Liu said his company is deepening its work with NVIDIA on next-generation graphics processor and networking platforms, drawing on its production network across Asia, North America and Europe to support a ramp-up.
The commitments cover different stages of manufacturing, so an increase in capacity at one stage is not a count of finished systems. NVIDIA has not disclosed a combined capacity total or a timetable for delivering finished systems to customers.
4 min