
AI-generated editorial illustration of semiconductor substrates for advanced AI chip packaging.
Nvidia is expanding deals for chip parts and packaging. The plans span Taiwan, South Korea and the United States. New output is due over the next few years.
The November 17 update names Kinsus, Samsung Electro-Mechanics and SPIL. It also includes earlier deals with Amkor.
Kinsus plans more base layers for chips
Substrates form the base of a chip package. They support the chip and carry its links. Packaging joins the parts into a working product.
Kinsus plans more ABF substrates for Nvidia at its Youth campus. The site is in Taoyuan, Taiwan. Its aim is room for over 3.5 times as much Nvidia output.
The first output is due in the second half of 2027. It aims to reach that level in the third quarter of 2028.
Nvidia also changed its Samsung Electro-Mechanics deal. It aims to get more output from plants already in use. Their base layers serve compute and fast network products.
Foxconn plans to support the firm's next GPU and network platforms. It will use plants in Asia, North America and Europe.
Amkor deals cover two countries
A July deal with Amkor was worth $1.5 billion. It backs chip packaging in Arizona. Funds were to be paid in advance.
An October deal added work in South Korea. Amkor put the deals across both countries at $3 billion. These earlier sums form part of the wider plan.
Amkor puts chip parts together and tests them before they ship.
Nvidia backs more SPIL lines
SPIL plans more packaging work in Taiwan. Nvidia backs the deal with billions of dollars. Room set aside for its chips should grow by mid-2029.
The plan adds interposers, the layers that link chips. It also adds lines to join parts and test them.
The targets set out how much the plants aim to make. New tools must be installed and checked before output starts.
Our AI GPU guide explains how the parts form racks. The Nvidia guide covers the wider network and software.
Read also
Popular news

Anthropic opens at $111.50 on Nasdaq, 34.34% above IPO price

GAO tests found safety gaps in four major AI chatbots

Nvidia revenue doubles to $115.8 billion as Rubin ships

Your AI Model Hasn’t Changed. The Company Running It Might Have.

Nvidia RTX Spark mini PCs bring up to 128GB memory to local AI

