According to reports, Google is in talks to commission Samsung to manufacture Axion processors and TPUs, AMD plans to shift part of its CPU orders starting in 2028, and Tesla’s next-generation AI6 chip will be exclusively produced at Samsung’s Texas facility. Capacity constraints and other factors are driving the adoption of a 'dual-supplier' strategy, leading to a shift in the global advanced chip foundry landscape from Taiwan Semiconductor toward greater diversification. This also presents Samsung Foundry with significant structural breakthroughs and rebound opportunities.
The explosive growth in demand for AI infrastructure is reshaping the global landscape of advanced semiconductor foundry services.
On June 17, according to Nikkei Asia, six informed sources revealed that global technology and automotive giants such as Google, AMD, and Tesla are accelerating efforts to secure advanced process foundry services from Samsung Electronics, primarily due to $Taiwan Semiconductor (TSM.US)$ persistent tightness in advanced capacity, making it increasingly difficult for new clients and small-batch orders to secure production slots. If this trend continues to materialize, it would represent $Samsung Electronics (005930.KR)$ the most significant structural breakthrough for the foundry business in recent years.
The report noted that this shift signifies an accelerated transition toward a more diversified supply chain in the global advanced chip foundry landscape, away from Taiwan Semiconductor’s dominance. Multiple leading chip designers are now implementing a 'dual-supplier' strategy, distributing orders between Taiwan Semiconductor and Samsung to hedge against capacity-related uncertainties and other risks.
Notably, Samsung’s Foundry and System LSI businesses generated combined revenue of KRW 6.9 trillion (approximately USD 4.6 billion) in the first quarter, up about 15% year-over-year, and the new wave of orders is expected to further bolster the segment’s recovery.
Taiwan Semiconductor’s capacity hits peak levels, forcing customers to seek alternatives
The tight capacity at Taiwan Semiconductor’s advanced process nodes is the primary driver behind this round of order reallocation.
Taiwan Semiconductor, Samsung, and Intel are the only three companies globally capable of mass-producing cutting-edge chips. Taiwan Semiconductor has already allocated the vast majority of its advanced-node capacity to top-tier clients such as NVIDIA, Apple, AMD, Broadcom, Marvell, and MediaTek.
According to the report, an executive at an automotive chip design company stated:
"Taiwan Semiconductor prioritizes production on its advanced nodes not only because it reinforces its technological leadership and long-term strategic positioning, but also because these nodes offer higher margins and remain consistently oversubscribed."
He added that while Samsung’s yield rates still lag behind those of Taiwan Semiconductor, the availability of its capacity is making it increasingly attractive to customers.
Google, AMD: Specific discussions are already underway
Multiple specific cooperation discussions are currently ongoing, involving clients across the technology, automotive, and semiconductor design sectors.
According to reports, two informed sources revealed that Google is in talks with Samsung regarding the production of its next-generation Axion processor, which is expected to launch around 2028. Meanwhile, Google may as early as 2028 commission Samsung to manufacture some of its Tensor Processing Units (TPUs) for AI computing.
Regarding AMD, a source familiar with the matter disclosed that the company is negotiating with Samsung to potentially commission it to produce certain future CPU products starting as early as 2028. AMD is currently a major customer of Taiwan Semiconductor, and this move signifies a significant adjustment in its supply chain strategy.
Tesla and NVIDIA: Samsung Is Already a Real-World Partner
Unlike the aforementioned clients, whose discussions are still at the negotiation stage, $Tesla (TSLA.US)$ and $NVIDIA (NVDA.US)$ has already established a substantive cooperative relationship with Samsung.
Tesla currently collaborates with both Samsung and Taiwan Semiconductor on the production of multiple products, including the AI5 chip used in vehicles and robotics.
It is reported that the upcoming next-generation AI6 chip will be exclusively manufactured by Samsung at its Texas facility. Tesla founder Elon Musk has also announced a collaboration with Intel on a future "mega fab" project in Texas, aimed at bringing more critical chip production back to the United States.
As for NVIDIA, its Groq Language Processing Unit (LPU) is currently manufactured by Samsung; however, NVIDIA has not yet decided whether to continue assigning production of the next-generation LPU to Samsung or switch to another foundry. The vast majority of NVIDIA’s chips are still exclusively fabricated by Taiwan Semiconductor.
Samsung Foundry: A Critical Window for Recovery from Its Trough
This wave of orders is highly significant for Samsung's foundry business. Previously, Samsung's foundry division primarily served its own consumer electronics and home appliance operations, and in recent years it has made limited progress in expanding its external client base. Moreover, the business saw a decline in profitability in the first quarter due to seasonal weakness.
Samsung's Foundry and System LSI businesses together generated KRW 6.9 trillion in revenue in the first quarter, up approximately 15% from KRW 6 trillion in the same period last year, though Samsung did not disclose separate profit figures for the two business units.
Meanwhile, Samsung is benefiting from surging demand for memory chips driven by the AI investment boom, with shipments of high-bandwidth memory (HBM), DRAM, and NAND flash all growing exponentially.
If its foundry business can simultaneously secure advanced-node orders from clients such as Google and AMD, Samsung could solidify its positions in both memory and logic chips within the AI supply chain, further enhancing its strategic role in the global semiconductor ecosystem.
Want to select stocks or get a stock diagnosis? Curious about the opportunities and risks in your portfolio? For all your investment questions,Just ask Futubull AI!
Editor/joryn