Nvidia has a demand problem, the kind most companies would happily trade places with. Its answer is a plan to grow key parts of its semiconductor supply chain by more than 300% to keep pace with AI demand.
The headline piece involves Amkor Technology. Announced at GTC Berlin on October 21, 2026, Amkor plans to boost its advanced packaging interposer wafer capacity in South Korea by over 300% by the end of 2028.
What Nvidia and Amkor are building
An interposer is the platform that sits underneath those components and connects them. If you cannot make enough interposers, it does not matter how many processors roll off the line, because they have nowhere to sit.
Amkor’s South Korea expansion is meant to ease exactly that bottleneck. The company plans to put approximately 3 trillion won, or around $2.2 billion, into its Gwangju facility to strengthen packaging for AI and high-performance computing (HPC).
Ground was broken in Gwangju on October 19, 2026, just two days before the Berlin announcement. The project is expected to create around 2,000 jobs in the region.
The packaging push does not stop at Amkor. Nvidia will also expand substrate capacity by 2.5x by working with multiple suppliers. Nvidia is also lining up significant increases in optical-engine and testing capacity through 2028.
The Arizona side of the deal
The Korea announcement builds on a strategic partnership the two companies set up in July 2026. At that point, Nvidia made a $1.5 billion prepayment to Amkor to expand US packaging and test capacity, with the focus on Arizona.
Investors liked the move. Amkor’s stock rose as much as 17% in extended trading after the July announcement.
Amkor’s Arizona plans have grown steadily since then. The company’s Peoria campus started with initial estimates of around $2 billion in 2023, and that figure had climbed to $12 billion by September 2026.
The Arizona buildout includes a Phase 2 that adds 60,000 square meters of cleanroom space. That brings the site to a total of about 93,000 square meters.
Why capacity is the new battleground
Nvidia CEO Jensen Huang has been blunt about where the pressure sits. He pointed to the unprecedented scale of infrastructure needed to meet current demand and summed up the situation in four words.
“Capacity is the constraint.”
The Arizona investments also fit into a broader policy picture. They line up with US efforts to bolster domestic semiconductor production under the CHIPS Act, which aims to bring more of the supply chain onshore.
That gives the partnership a two-track shape. Arizona supports the US onshoring push, while Gwangju adds heavy capacity in South Korea for AI and HPC work.
What this means for the AI chip supply chain
For Amkor, Nvidia’s backing changes its position in the packaging market. A $1.5 billion prepayment from the most important customer in AI hardware is both a funding source and a vote of confidence.
It also raises the stakes for execution. Amkor now has major projects running in parallel across two countries, with the Korea capacity target set for the end of 2028.
The key dates to watch run through the end of 2028. That is when Amkor aims to have its South Korea interposer capacity up by more than 300%, and when Nvidia’s broader capacity increases in optical engines and testing are slated to play out.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.