Tesla and NVIDIA plan to invest 3 trillion KRW in Samsung Electro-Mechanics to expand FC-BGA capacity

Tesla, Nvidia plan to invest 3 trillion won in Samsung Electro-Mechanics Source: Ban Dao Zong Heng Publish time: 2026-10-23 15:53 Tesla Nvidia generate poster Two giants join forces to bet on Samsung Electro-Mechanics FC-BGA, locking in AI substrate capacity for the next five years.

According to sources, Samsung Electro-Mechanics will receive a combined investment of about 3 trillion won from Tesla and Nvidia, earmarked for expanding capacity of FC-BGA, the core packaging substrate for AI semiconductors. Of this, Tesla will contribute about 1 trillion won and Nvidia about 2 trillion won.

It is worth noting that less than a month ago, Samsung Electro-Mechanics had just announced its largest-ever capacity expansion plan. Now, the two giants are directly taking equity stakes with real money, firmly securing the supply of high-end substrates for the next five years in their own hands.

Capital backing: the deeper considerations behind the 3 trillion won investment

On September 28, Samsung Electro-Mechanics announced an investment of 4.27 trillion won in its Sejong plant in South Korea and 2.51 trillion won in its Vietnam subsidiary, totaling 6.78 trillion won to expand packaging substrate capacity, with mass production targeted for September 2028. In terms of investment scale, this outlay already accounts for 43.6% of Samsung Electro-Mechanics' total share capital at the end of 2025, setting a record for investment in a single product line since the company's founding.

However, Samsung Electro-Mechanics did not bear all the capital expenditure risk alone; it brought in direct investment from top global customers. According to currently disclosed information, Tesla plans to invest about 1.1 trillion won and Nvidia about 2 trillion won, and this funding almost fully covers the core capital needs of the Sejong plant expansion.

In the high-end packaging substrate industry, it often takes 2-3 years for a product to go from sample submission and certification to stable mass supply. Once a customer changes suppliers, it must bear extremely high time costs. In the past, the industry generally followed a model where manufacturers advanced the money for expansion first and then waited for customer orders to materialize. This time, the model of major customers directly funding and jointly building capacity is equivalent to directly converting downstream computing power arms-race pressure into certainty of growth on the upstream manufacturing side. This kind of deep binding at the industrial chain level is quite rare in the global semiconductor supply chain.

Dual-track layout: automotive/robotics and data centers equally emphasized

It is reported that the dedicated lines of the two major customers are clearly divided.

The Tesla dedicated line is located at the Sejong plant, specifically producing AI semiconductor substrates for autonomous-driving electric vehicles and the humanoid robot Optimus, with a supply contract period of five years (2029-2033), planned mass production in the second half of 2028, and supply starting in 2029. Previously, Musk confirmed that Tesla's self-developed next-generation AI chip AI5 will be manufactured by Samsung Electronics' foundry division using the 2nm process at its Taylor plant in Texas, USA, and is planned to be first integrated into Optimus starting in April 2027. Industry expectations are that Samsung Electro-Mechanics' FC-BGA is expected to enter Tesla's automotive and robotics hardware systems along with AI5.

The Nvidia dedicated line targets the data center market, producing AI computing chipsets and network substrates, and simultaneously expanding the Sejong and Busan plants; the cooperation was disclosed on October 21 at GTC 2026 in Germany. As autonomous driving, humanoid robots, and AI data centers expand simultaneously, demand for high-performance FC-BGA is spilling over comprehensively from data centers to end-device scenarios.

Role reshaping: from "chaser" to "core supplier"

As the core carrier of AI computing hardware, FC-BGA flip-chip ball grid array substrate is an essential high-end packaging substrate that supports AI GPUs and high-performance CPUs in achieving high bandwidth and high computing output. Its core production material relies on Ajinomoto's ABF build-up insulating film from Japan, and the difficulty of the entire process far exceeds conventional PCB products, making it a highly barrier-laden segment of the semiconductor industry chain.

In this track, high-end capacity has long been highly concentrated. For a long time, only seven manufacturers worldwide had the capability for stable mass production of high-end ABF substrates: Japan's Ibiden with about 20% market share, Taiwan, China's Unimicron with about 18%-24%, Taiwan, China's Nan Ya PCB (NPB) with about 12%-20%, Japan's Shinko with about 11%, South Korea's Samsung Electro-Mechanics (SEMCO) with about 8%-10%, Austria's AT&S with about 11%, and Taiwan, China's Kinsus with about 5%-10%.

For Samsung Electro-Mechanics, the significance of securing investment from the two giants this time has long gone beyond merely adding capacity. Over the past decades, Japanese and Taiwanese manufacturers have consistently held a dominant position in the high-end substrate market, and South Korean manufacturers have remained in the role of chaser in this track. This time, the direct investment from Tesla and Nvidia is equivalent to casting a vote of confidence with real money, proving that Samsung Electro-Mechanics' FC-BGA products have officially entered the core supply chain of top global AI customers, and, with long-term contracts for the next five years, it has thoroughly completed the identity leap from "chaser" to "core supplier."

Conclusion

For the entire industry, this investment is also an important signal: the bottleneck of the AI computing power race is transmitting from wafer manufacturing to upstream segments such as packaging substrates, and major customers are beginning to use capital means to lock in scarce capacity in advance. As automotive, robotics, and data centers simultaneously compete for the same kind of high-end substrate, FC-BGA is becoming the next key segment in the AI industry chain to be repriced after HBM.

This article is reproduced from media reports or online platforms and represents the author's personal stance or views. Our republication is solely for sharing and does not mean we approve of or endorse it. If the source attribution is incorrect or your lawful rights and interests are infringed, please contact customer service promptly, and as a neutral platform service provider we will promptly correct, delete, or handle it in accordance with the law. Comments No user comments yet Latest news More Related recommendations

Samsung places order for 2nm equipment, Taylor plant adds monthly capacity of 6,000 wafers 2026-10-23 16:33

Global LED display panel market will still achieve high value-added growth 2026-10-23 16:32

China Northern Rare Earth's third-quarter net profit attributable to parent company was 1.593 billion yuan, up 161.13% year on year 2026-10-23 16:31