Wall Street’s AI trade has largely revolved around processors, memory and data centers. Samsung Electro Mechanics is making a nearly $5 billion bet that the next wave of spending will reach a far less visible part of the computing stack.
The South Korean electronic components maker, an affiliate of Samsung Group, plans to invest $4.9 billion in South Korea and Vietnam to expand production of advanced semiconductor substrates. These components sit underneath powerful processors and carry the electricity and signals that allow AI accelerators, graphics processors and central processing units to work inside servers.
The more telling part of the investment is who is helping pay for it. An unidentified global technology company will provide funding for Samsung Electro Mechanics’ expansion in South Korea and has agreed to secure production over the medium to long term. The customer was not disclosed, but the arrangement shows how aggressively large technology companies are trying to lock in the physical infrastructure needed for AI computing before new capacity even comes online.
That matters because building more AI computing power is no longer simply a question of buying more advanced processors. As AI systems pack together more high performance chips, the substrates beneath them must become larger, thicker and more complex. A bottleneck in those components can limit how quickly additional computing systems can be built even when processors themselves are available.
Samsung Electro Mechanics will spend $3.1 billion at its plant in Sejong, a manufacturing center in central South Korea, to increase production of flip chip ball grid array substrates known as FCBGA. It is the largest investment the company has ever made in a single product.
The company will begin investing this month and plans to complete the expansion by May 2028. Mass production from the new lines is scheduled to begin in September 2028.
Another $1.8 billion will go to Samsung Electro Mechanics’ manufacturing operation in Vietnam, where the company will expand package substrate capacity. The dual investment gives the company additional production in two Asian manufacturing hubs as orders for advanced computing components rise.
The timing reflects a shift taking place across the semiconductor industry. AI chips are increasingly designed around multiple powerful processors working together rather than a single chip doing the job alone. That puts greater demands on the components that distribute power and move data between them.
Samsung Electro Mechanics plans to use its South Korean operation for its most technically demanding and higher value products, while expanding overall capacity across South Korea and Vietnam. The company expects the new production to serve AI servers, high performance computing systems and automotive electronics.
Chief Executive Chang Duck hyun said advanced substrates are becoming increasingly important to both semiconductor performance and supply chain stability as AI and high performance computing markets expand.
The investment is another sign that the economics of the AI boom are spreading outward. The companies benefiting from the buildout may increasingly include not only the makers of processors and memory chips, but also the suppliers of the less visible components required to connect, power and scale them.
