Samsung Foundry Leverages Memory Edge to Win Big Tech Orders

Global big tech companies are knocking on Samsung Electronics' foundry door to secure high-performance memory semiconductors.

As the worldwide artificial intelligence infrastructure boom intensifies shortages of essential chips like High Bandwidth Memory, Samsung's "turnkey" solution—capable of handling both memory and foundry processes—has emerged as an alternative. Industry observers expect these market conditions to provide an opportunity for Samsung to break its non-memory business losses and achieve an earnings turnaround.

According to industry sources and foreign media reports on the 12th, ByteDance, the Chinese parent company of TikTok, is reportedly in discussions with Samsung Electronics for foundry cooperation while pursuing in-house AI chip development. ByteDance is said to be reviewing Samsung's foundry lines with a goal of securing its AI inference chip, codenamed "SeedChip," by the end of March. Initial volume is expected to be around 100,000 units, potentially increasing to 350,000 units.

A notable aspect of these negotiations is that ByteDance also inquired about memory chip supply from Samsung. Reuters, citing multiple sources, reported that ByteDance put memory chip supply access on the negotiating table. High-performance memory including HBM is currently in short supply. Analysts suggest ByteDance may be strategically securing stable memory semiconductor supply as a condition for placing foundry orders.

Samsung Electronics is the world's only integrated device manufacturer with both memory semiconductor and foundry divisions. The appeal lies in receiving designed chips on time while also securing hard-to-obtain high-performance memory.

Samsung's Unique Weapon: Memory Control and End-to-End Production

This "package deal" possibility aligns with Samsung's strategic move exploiting gaps in the technology supremacy race. While Taiwan's TSMC dominates the foundry market, it does not produce memory. AI accelerators, however, require the combination of logic and memory chips. As heterogeneous integration technology—bundling HBM and GPUs into single packages—becomes increasingly important, Samsung's efficiency in producing both memory and logic chips under one roof stands to gain prominence.

ByteDance reportedly plans to spend approximately 160 billion yuan (33.6 trillion won) on AI-related procurement this year alone. A significant portion of this massive funding is expected to go toward Nvidia chip purchases and in-house chip development.

"With TSMC's production lines at capacity, ByteDance likely made a strategic choice to find alternatives while also addressing memory supply concerns," an industry official said.

However, ByteDance has stated that reports about its chip project are inaccurate, while Samsung Electronics has maintained a principled stance that it cannot confirm matters related to customers.

Tesla CEO Elon Musk's moves are also sending positive signals to Samsung's foundry business. Musk recently stated via social media that the autonomous driving chip "AI5" design is nearly complete and the next-generation "AI6" design has entered early stages.

"AI5 chip design is almost done and AI6 chip design has begun," Musk said. "We will shorten the design cycle for AI7, AI8, and AI9 to nine months."

This is interpreted as a commitment to dramatically reduce development cycles that previously took about three years. Realizing this requires a foundry partner with stable mass production capabilities. The market expects Samsung Electronics to produce Tesla's next-generation chips using 2-nanometer to 3-nanometer advanced processes at its Taylor, Texas facility.

The prevailing view is that TSMC's production capacity alone will be insufficient to meet the "world's highest production volume" Musk mentioned. During last October's earnings announcement, Musk stated that both Samsung Electronics and TSMC would work on AI5, effectively confirming Samsung's production participation. Samsung Electronics Chairman Lee Jae-yong's meeting with Musk during his U.S. trip last year to discuss comprehensive cooperation also supports the likelihood of securing orders.

Tesla's autonomous driving chips are core components used not only in vehicles but also in humanoid robots and large-scale AI model operations. If Samsung Electronics reliably handles Tesla's volume, it would provide a decisive opportunity to prove its foundry process reliability.

Customer Diversification Including Tesla and Nintendo Signals Non-Memory Loss Exit

Samsung's foundry order status reveals a trend toward customer diversification. Samsung has reportedly secured or is in production discussions for Tesla's next-generation autonomous driving chips, Nintendo's Tegra system-on-chip, Apple's next-generation image sensors for iPhones, Ambarella's edge AI SoC, IBM's next-generation data center "Power11" chip, and Intel's platform controller hub chips.

Apple's image sensor order is particularly significant as it indicates Samsung passed rigorous quality standards. The customer portfolio is expanding from console gaming powerhouses like Nintendo to server chip leaders like IBM. This suggests the business structure, previously concentrated on mobile application processors, is expanding into high-performance computing and automotive sectors.

This order expansion is expected to improve profitability for Samsung's non-memory business. Securities firms estimate Samsung's foundry and System LSI divisions recorded operating losses in the 6 trillion won range last year. Losses of 2.7 trillion to 3.5 trillion won are projected this year, though the deficit is expected to gradually narrow. Utilization rates are expected to rise and fixed cost burdens ease once big tech volumes from Tesla and ByteDance enter full mass production.

Samsung Electronics plans to maximize synergy by linking memory market strength with foundry order expansion. With memory supply shortages expected to persist, Samsung's unique sales strategy combining chip production with memory supply is projected to gain further momentum. Industry attention is focused on whether Samsung Electronics can achieve meaningful market share gains in the TSMC-dominated foundry market by leveraging memory as a powerful weapon.