SEOUL, June 16 (AJP) - Samsung Electronics has reportedly secured a string of AI-related chip projects from Alphabet's Google and Elon Musk's Neuralink by exploiting manufacturing bottlenecks at Taiwan Semiconductor Manufacturing Co. (TSMC), giving fresh momentum to a foundry business that has long struggled to close the gap with its Taiwanese rival.
According to sources familiar with the matter, the South Korean tech giant has won several high-profile silicon design projects from Google and Neuralink. Samsung declined to comment on specific agreements, citing non-disclosure agreements.
The shift comes as surging demand for generative AI accelerators pushes TSMC's manufacturing and advanced packaging capacities to their limits, prompting global technology companies to diversify away from a single supplier and adopt multi-vendor sourcing strategies to reduce supply chain risks.
In a significant departure from its historical reliance on TSMC, Google is adopting a split-manufacturing architecture for its 10th-generation Tensor Processing Unit (TPU), codenamed "Icefish."
Scheduled for mass production in 2028, the project divides manufacturing duties between the two Asian rivals. TSMC will fabricate the core compute processor using its ultra-advanced 1.4-nanometer process, while Samsung has been selected to manufacture the memory input/output die using its 2-nanometer technology.
The I/O die acts as a high-speed bridge between logic processors and high-bandwidth memory (HBM) stacks, an increasingly critical function as memory bandwidth emerges as one of the biggest bottlenecks in AI computing.
The project would also allow Samsung to offer an integrated manufacturing model by supplying its own HBM chips, producing the I/O die and conducting advanced packaging under a single roof.
At the same time, Samsung is reportedly deepening its relationship with Elon Musk's technology ventures beyond electric vehicles.
Industry sources said Samsung's foundry division is co-developing a fourth-generation brain-computer interface implant for Neuralink using its 4-nanometer FinFET process.
The chip is designed to enable two-way communication between brain tissue and digital devices and would represent a shift away from TSMC, which manufactured Neuralink's previous generations of processors.
Samsung aims to deliver functional test wafers in the first half of 2027, with high-volume production targeted for late 2027, according to the sources.
The Neuralink project builds upon Samsung's existing partnership with Musk's companies after securing a multi-billion-dollar contract last year to manufacture Tesla's next-generation AI6 autonomous driving chips, which are expected to anchor early production at Samsung's expanding semiconductor facility in Taylor, Texas.
Samsung has also recently produced specialized language-processing hardware for AI startup Groq, further broadening its AI customer base.
The information also has not been confirmed by Google and Neuralink.
As traditional transistor miniaturization becomes increasingly difficult and expensive, competition is shifting toward advanced packaging, memory integration and thermal management technologies.
"U.S. Big Tech companies are increasingly striving to develop their own AI accelerators to break away from Nvidia's monopoly, but manufacturing options are limited," said Lee Jong-hwan, a professor of system semiconductor engineering at Sangmyung University. "With TSMC's advanced nodes fully booked, these companies have no choice but to turn to Samsung, and we expect these orders to grow."
Lee emphasized that Samsung's unique position as both the world's leading memory producer and a contract chip manufacturer gives it a strategic advantage in the AI era.
"Samsung is the only company on the planet capable of offering a comprehensive turnkey service that encompasses supplying HBM, manufacturing the custom logic chip, and handling the advanced packaging," Lee noted.
Unlike competitors that rely on multiple suppliers, Samsung can potentially offer customers a unified manufacturing model that combines memory, foundry and packaging services, reducing execution risks and shortening development cycles.
"If customer demand for this integrated model continues to rise, it could serve as a crucial catalyst for Samsung to turn its foundry business profitable and narrow the market share gap with TSMC," Lee added.
While TSMC remains the dominant force in the global foundry market, the AI boom is beginning to expose the limits of single-supplier dependence.
For Samsung, that may finally present the opportunity it has sought for years: transforming itself from a distant challenger into an indispensable second pillar of the global AI supply chain.
Copyright ⓒ Aju Press All rights reserved.