NVIDIA Partners with Six Global Suppliers to Expand AI Packaging and Substrate Capacity
NVIDIA announces major capacity buildouts with IBIDEN, Samsung Electro-Mechanics, Amkor, and KYEC to expand substrates 2.5x and double testing for AI superchips.
NVIDIA has announced extensive multi-billion-dollar supply chain partnerships with Amkor, IBIDEN, KYEC, Samsung Electro-Mechanics, SPIL, and Unimicron to break semiconductor manufacturing bottlenecks. The initiative expands packaging substrates by 2.5x, triples interposer packaging capacity, doubles semiconductor testing with a $1.1 billion US facility advance, and scales co-packaged optical engines by 15x.
2.5x Substrate Capacity Expansion with IBIDEN, Samsung, and Unimicron
On October 21-22, 2026, NVIDIA announced a global manufacturing initiative targeting chronic packaging and substrate constraints for its AI computing platforms. NVIDIA plans to expand total package substrate capacity by 2.5x over the next three years. Key agreements include a multibillion-dollar advance to IBIDEN to triple capacity at its Ono Plant in Japan by 2028, integrating Samsung Electro-Mechanics into its supply chain across factories in Sejong and Busan, and reserving long-term volume at Unimicrons KF substrate facility in Taiwan ramping in mid-2027.
Advanced Packaging and $1.1 Billion US Testing Expansion
To support higher production runs for Blackwell Ultra and upcoming Vera Rubin platforms, NVIDIA partnered with Amkor Technology to expand advanced packaging in South Korea, targeting a greater than 300% increase in interposer wafer capacity by late 2028. Simultaneously, NVIDIA will double its global semiconductor testing capacity between 2026 and 2027, backed by a $1.1 billion advance to King Yuan Electronics (KYEC) to equip and scale new high-volume testing facilities inside the United States by April 2027 alongside facilities in Singapore and Taiwan.
Co-Packaged Optics Scaling 15x for Next-Gen AI Factories
Recognizing that interconnect bandwidth and power dissipation are redefining data center limits, NVIDIA and Siliconware Precision Industries (SPIL) announced a dedicated manufacturing expansion in Taiwan for optical engines used in co-packaged optics (CPO). The collaboration targets a 15x capacity increase by mid-2027 followed by another doubling by late 2028. NVIDIA founder and CEO Jensen Huang framed the push as essential to delivering the physical computing backbone of the AI industrial revolution.