Samsung Electro-Mechanics and Nvidia have announced a strategic partnership to expand production and supply of ABF package substrates. These components sit underneath the high-performance processors and networking switches that power AI infrastructure.
The deal was announced on October 21, 2026. It targets one of the least glamorous and most stubborn constraints in the AI buildout: the materials that hold advanced chips together.
What the two companies agreed to
Under the arrangement, Samsung Electro-Mechanics (SEMCO) and Nvidia intend to jointly invest in dedicated manufacturing lines. Those lines will be located at SEMCO’s facilities in Sejong and Busan, South Korea.
Production is scheduled to start in 2026, with supply to Nvidia expected to begin that year. The agreement anticipates supply volumes growing at a double-digit annual rate over several years.
ABF stands for Ajinomoto Build-up Film. Think of an ABF substrate as the foundation a chip is mounted on, plus the wiring that connects the chip to everything around it.
For AI accelerators and high-speed networking switches, these substrates must handle enormous data flows and tight tolerances.
A deeper seat in Nvidia’s supply chain
SEMCO is not a newcomer to Nvidia’s orbit. The company strengthened its position in Nvidia’s supply chain throughout 2026, and had previously supplied substrates for Nvidia’s NVSwitch interconnects, under agreements reported in January 2026.
NVSwitch is the technology Nvidia uses to link many GPUs together so they can function like one large processor.
SEMCO is also positioned as a potential first vendor for substrates tied to Nvidia’s Groq3 LPU inference chip. Talks over that first-vendor role were reported in April 2026, and the arrangement has not been described as finalized.
Nvidia is not betting on a single supplier, though. Alongside the SEMCO deal, the company is expanding substrate capacity through agreements with other suppliers, including Ibiden and Unimicron.
Nvidia is reportedly reserving multibillion-dollar support for multiple substrate partners to expand capacity. The broader push is aimed at growing substrate capacity by 2.5x or more in the coming years.
Why substrates became the bottleneck
High-performance packaging substrates have emerged as a major bottleneck for AI technology, and the SEMCO partnership is a direct response to those supply chain pressures.
What this means for the AI hardware market
For SEMCO, the deal offers visibility. A multi-year agreement with double-digit annual volume growth gives the company a clearer runway for planning capacity and capital spending.
Dedicated lines co-funded by the customer also reduce some of the risk of building out expensive manufacturing that might sit idle if demand cools.
The Groq3 LPU substrate talks are another item to watch. A confirmed first-vendor role would extend SEMCO’s reach from Nvidia’s networking and training hardware into the fast-growing inference segment.
The 2.5x capacity target offers a useful yardstick for tracking whether the substrate bottleneck eases in the coming years.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.