News

NVIDIA’s capacity expansion starts underneath the chip

NVIDIA plans to expand the substrate capacity available for its products to 2.5 times its current level over three years, through agreements involving IBIDEN, Samsung Electro-Mechanics and Unimicron.
Suppliers named in NVIDIA manufacturing capacity announcement

Image: NVIDIA

NVIDIA plans to expand the substrate capacity available for its products to 2.5 times its current level over three years, through agreements involving IBIDEN, Samsung Electro-Mechanics and Unimicron.

The immediate contribution comes partly from production lines that already exist. Samsung Electro-Mechanics is entering the supply chain with South Korean capacity while the partners invest in dedicated facilities in Sejong and Busan. Its agreement covers substrates for compute chips and high-speed networking switches, with deliveries beginning in 2026.

The technical difficulty sits beneath the headline factory investment. As packages grow, the substrate has to maintain fine circuit geometry and sufficient flatness through manufacture and assembly. Warpage can undermine the yield of an expensive package containing otherwise functional silicon.

Samsung’s account also describes embedding MLCCs and silicon capacitors within advanced substrates to support power integrity. That identifies a manufacturing capability, rather than disclosing the contents of a particular NVIDIA package. It does, however, explain why capacity for these products cannot be treated as interchangeable with any spare substrate line.

Samsung expects its supply volume to NVIDIA to grow at a double-digit annual rate, with NVIDIA accounting for a significant portion of its annual package-substrate capacity by 2028. Neither company puts a percentage on that allocation.

The other expansions have their own calendars. Additional capacity at IBIDEN’s Ono Plant is expected to start ramping in the first quarter of 2028. Unimicron expects its expansion to begin contributing in the second quarter of 2027 and reach full capacity by mid-2028.

“Capacity is the constraint,” said Jensen Huang. NVIDIA’s wider agreements show that the company is addressing several distinct manufacturing constraints at once. Amkor targets an increase of more than 300% in interposer wafer capacity by the end of 2028. A $1.1 billion capacity-reservation advance to KYEC supports new US testing facilities, with volume production targeted for April 2027.

Sources

Continue reading