NVIDIA's supply chain expansion got concrete numbers at GTC Berlin on October 21, 2026. In a newsroom announcement that day, NVIDIA said it will expand substrate, advanced packaging, testing and optical-engine capacity with six partners, lifting total substrate capacity to 2.5x over the next three years — with the bottleneck priced out link by link for the first time.
The partners are Amkor, IBIDEN, KYEC, Samsung Electro-Mechanics, SPIL and Unimicron. IBIDEN's Ono plant in Japan targets 3x current levels, with new capacity ramping from Q1 2028; Samsung Electro-Mechanics is expanding in Sejong and Busan, and by 2028 output allocated to NVIDIA should be a significant share of its annual package substrate capacity; Unimicron's Taiwan expansion should start adding capacity in Q2 2027 and reach full capacity by mid-2028.
Packaging and testing move earlier. Amkor plans to raise interposer wafer capacity by more than 300% by the end of 2028. NVIDIA's $1.1 billion capacity reservation advance to KYEC supports new U.S. testing facilities targeting high-volume production in April 2027, with total testing capacity expected to double from 2026 to 2027. SPIL targets a 15x increase in optical-engine capacity for co-packaged optics by mid-2027, and another doubling by the end of 2028.
The numbers show AI factory competition now extends from chip design to who locks in upstream capacity first. For cloud providers and system builders, 2027 delivery certainty is being decided by today's prepayments; for investors, the thing to watch is whether each link's ramp schedule holds, not just the headline multiples.
