Global 2nm Foundry Competition: TSMC, Intel, Samsung, and Rapidus Vie for AI Computing Power Supremacy

In semiconductor manufacturing, the 2nm node represents a critical juncture, embodying advanced process technologies, EUV (Extreme Ultraviolet) lithography clusters, GAA (Gate-All-Around) transistors, advanced packaging, supply chain dynamics, and geopolitical considerations. It is widely regarded as the threshold for securing computational supremacy in the AI era.

Today, a capital and national strategic competition centered around “building 2nm foundries” is unfolding globally: TSMC is ramping up its 2nm factory deployment in Taiwan while advancing overseas projects in Arizona (USA), Kumamoto (Japan), and Dresden (Germany); Intel, leveraging its 18A process and backed by a “national team shareholder group,” aims to reshape the foundry landscape; Samsung continues to chase improvements in 2nm yield rates and customer structure; and Japan’s Rapidus, under policy support, is betting on a single-wafer process to revive Japan’s ailing semiconductor industry.

TSMC Plans to Build 10 2nm Factories

According to the latest Taiwanese media reports, TSMC’s 2nm deployment in Taiwan has evolved from a “seven-factory” concept to a “ten-factory” vision—comprising two factories in Hsinchu Baoshan, five in Kaohsiung Nanzhi, and an additional three planned in the Southern Taiwan Science Park (STSP) special zone. The estimated cost for a single 2nm foundry is approximately NT300billion(US8-10 billion), implying that the three newly added factories would require an investment of around NT$900 billion.

Recent public disclosures further corroborate these speculations:

  • TrendForce, citing Taiwanese media such as the Liberty Times, reported that driven by surging AI chip orders, TSMC’s existing seven advanced-process factories in Hsinchu and Nanzhi are insufficient to meet demand. The company is evaluating the construction of up to 12 new advanced-process and packaging factories in Taiwan, with a focus on 2nm and 1.4nm technologies.
  • According to Taiwan’s Central News Agency, the Central Taiwan Science Park has confirmed that TSMC’s 1.4nm (A14) factory in the Taichung A14 site has obtained construction approval and is set to commence operations by the end of 2025, targeting mass production in 2028. This indicates a nearly seamless transition to the next node after 2nm on Taiwan’s main island.
  • Previously, TSMC had planned to build five new factories in the Kaohsiung Nanzhi New Park for 2nm, A16, and more advanced processes, complemented by a supply chain base in the Pingtung Science Park, gradually transforming southern Taiwan into another advanced-process hub.

Beyond Taiwan, TSMC is aggressively expanding overseas: In March, the company announced plans to increase its total investment in three factories, two advanced packaging facilities, and a large-scale R&D center in Arizona (USA) to US$165 billion. TSMC CEO C.C. Wei emphasized during the announcement, “All AI customers are flocking to us, yet Taiwan’s production capacity remains severely inadequate,” urging the government to assist in securing land, electricity, and water resources locally.

TSMC’s Strategic Rationale for 2nm

From a business perspective, TSMC’s stance on 2nm is clear:

Firstly, advanced-process expansion is driven by customer orders, specifically for top-tier clients. The 2nm family will primarily serve AI GPUs/accelerators (NVIDIA, AMD, custom ASICs), high-end CPUs/GPUs/APUs, and select premium mobile SoCs. Even amid macroeconomic fluctuations, these customers possess sufficient bargaining power to secure long-term capacity commitments.

Secondly, the most cutting-edge nodes must remain on Taiwan’s main island. Overseas factories primarily address political and customer relationship gaps, securing subsidies and local automotive/industrial clients in the U.S., Japan, and Europe. However, the technical focus remains on Taiwan’s 2nm/1.4nm clusters, explaining why 2nm mass production prioritizes Baoshan, Nanzhi, and STSP over Arizona.

From an industry competition standpoint, TSMC’s aggressive factory construction is both a passive response to unexpectedly high AI demand and an active defense against a wave of “policy + capital” attacks from Samsung, Intel, and Rapidus. By anchoring 2nm capacity deep within its fortress, TSMC makes it difficult for rivals to disrupt its ecosystem, even with subsidies and major clients in hand.