Nvidia unveiled agreements with a half-dozen suppliers to increase capacity to meet high demand for artificial intelligence systems powered by the company’s chips, according to an Oct. 21 press release.
Nvidia inked agreements with Ibiden, Samsung Electro-Mechanics and Unimicron Technology Corp. to increase manufacturing capacity 2.5 times in aggregate over three years.
The chipmaker also unveiled capacity agreements with partners Amkor Technology, King Yuan Electronics Co. and Siliconware Precision Industries to support production of systems that train and run AI models.
“We are in the midst of the largest infrastructure buildout in human history,” Jensen Huang, Nvidia's founder and CEO, said in the release. “Demand is extraordinary. Capacity is the constraint.”
Ibiden, Samsung Electro-Mechanics and Unimicron make substrates — specialized circuit boards that connect Nvidia’s processors to a computer’s main circuit board.
Samsung Electro-Mechanics will upgrade facilities and build new factories in Sejong and Busan, South Korea. By 2028, the company will allocate “a significant portion” of its total annual package substrate production capacity to Nvidia, per the release.
Unimicron’s agreement with Nvidia covers a manufacturing expansion in Taiwan scheduled to begin in the second quarter of 2027 and reach full capacity by mid-2028, according to the release.
Nvidia is providing Ibiden with a multibillion-dollar capacity reservation for a plant expansion in Japan, the release says. The deal, signed earlier this year, aims to triple the plant's current capacity. Ibiden expects additional production to begin ramping in the first quarter of 2028.
Under Nvidia’s agreement with Amkor, the packaging company plans to expand capacity at its South Korea plants to assemble chips into connected units, a process called advanced packaging. Specifically, Amkor has agreed to increase its capacity to process interposer-containing wafers by 300% by the end of 2028. An interposer is a connecting layer that allows communication between separate chips in the same package.
With KYEC, Nvidia has provided $1.1 billion to support new U.S. testing facilities and equipment. High-volume production is expected to start in April 2027.
Siliconware Precision Industries plans to work with Nvidia to expand manufacturing capacity in Taiwan 15-fold by mid-2027, while targeting another doubling of capacity by the end of 2028. The expansion will support advanced optical engines that provide the bandwidth and power-efficiency requirements of AI systems.
During Nvidia’s August earnings call, Huang told investors that supply constraints are weighing down Nvidia’s financial outlook for the next fiscal year. The shortages spanned the infrastructure provider’s supply chain because of the global build-out of data centers powering AI.
Nvidia’s advanced processors run large AI models. Its customers include enterprises such as Microsoft, Google and Amazon, which run vast AI data center networks.
