TSMC's Advanced Chip Capacity Tightens, Samsung Foundry Sees Surge in Inquiries

Central News

(Nikkei Asia, reported by Central News Agency, Taipei, 17th) As demand for artificial intelligence (AI) infrastructure surges, Taiwan Semiconductor Manufacturing Company (TSMC) is facing a severe supply shortage in its advanced chip production capacity, leading companies such as BYD, Google, AMD, and Tesla to increasingly turn to Samsung Electronics for chip manufacturing services.

According to six industry insiders, Samsung Electronics has seen a significant increase in inquiries from existing and potential global clients—including Chinese firms—seeking to utilize its advanced chip manufacturing capacity.

Two sources told Nikkei Asia that BYD, China's largest electric vehicle manufacturer, is continuously discussing with Samsung the production of its next-generation autonomous driving chips. Google is also engaging with the South Korean company to explore the possibility of Samsung manufacturing its next-generation Axion processor, expected to launch around 2028.

One insider revealed that Google is also considering outsourcing part of its key Tensor Processing Units (TPUs) for AI computing to Samsung as early as 2028.

A senior executive at a Chinese automotive chip design firm, speaking anonymously, said: "TSMC prioritizes production on advanced nodes, not only to maintain its technological leadership and long-term strategy, but also because these nodes offer higher profit margins and remain in short supply."

He added: "While Samsung's yield rates still lag behind TSMC's, its available capacity is becoming an increasingly attractive option."

Yield refers to the proportion of chips produced that meet required specifications and performance standards.

The Chinese automotive chip designer noted that many Chinese chip design firms are now pursuing a 'dual-sourcing' strategy—splitting orders between TSMC and Samsung rather than relying on a single foundry, with allocation ratios depending on the product.

Another senior chip industry executive told Nikkei Asia: "Geopolitical factors are also pushing some U.S. clients to use multiple foundries whenever feasible."

A high-level chip industry source said: "TSMC's advanced chip capacity is extremely tight, making it difficult for smaller Chinese clients with limited order volumes to place new orders. As a result, several have already approached Samsung to explore future collaboration possibilities."

Sources told Nikkei Asia that Google has been a strong advocate for a multi-foundry strategy.

However, using multiple foundries for advanced chip production requires substantial investment to adapt chip designs to different manufacturers' processes, along with stronger R&D and supply chain coordination.

Due to additional engineering and labor costs, this strategy is typically only viable for well-funded chip developers.

Meanwhile, sources say TSMC's capacity constraints may prompt other existing clients to place more orders with Samsung.

Tesla collaborates with both Samsung and TSMC across multiple products, including the AI5 chip used in its electric vehicles and robots. However, the upcoming AI6 chip will be manufactured at Samsung's Texas facility.

AMD and Nvidia, both core TSMC clients, are now also exploring collaborations with Samsung on certain chip projects due to capacity limitations.

One source said AMD is negotiating with Samsung to co-produce some of its future central processing units (CPUs) starting from 2028.

Nvidia's Groq language processing unit (LPU) is currently being manufactured by Samsung, though Nvidia has not yet decided whether to assign next-generation Groq production to another foundry.

Most of Nvidia's chips are still exclusively manufactured by TSMC.

The rising inquiry volume for Samsung's advanced foundry services marks a major win for the South Korean chipmaker's foundry business, which has historically served primarily its own consumer electronics and home appliance divisions. (Translation: Liu Shuqin) 1150617