TradingKey - Driven by the sustained growth in demand for artificial intelligence and high-performance computing, the global advanced chip market is embarking on a structural expansion that could last for several years. To accommodate growing orders from U.S. customers, TSMC ( TSM) is accelerating the construction of advanced node and advanced packaging capacity in Arizona.
TSMC recently announced an additional investment of $100 billion, increasing its total planned investment in the U.S. from $165 billion to $265 billion. Meanwhile, the company raised its 2026 capital expenditure guidance to $60 billion to $64 billion, reflecting management's confidence in the long-term demand for AI chips.
TSMC Chief Financial Officer Wendell Huang stated that this expansion is primarily driven by strong demand signals from U.S. customers and is supported by federal, state, and local governments. In TSMC's view, current chip demand is not a short-term fluctuation, but a long-term trend driven by AI, high-performance computing, and data center upgrades.
TSMC's footprint in Arizona began in 2020, with an initial investment of $12 billion. Since then, driven by growing demand from U.S. customers and the advancement of domestic chip policies, the project's scale was successively increased to $40 billion, $65 billion, and $165 billion, and was further raised to $265 billion in July 2026.
The latest plan will add multiple fabs utilizing advanced processes of 2nm and below, while expanding advanced packaging capacity. According to information released by the Arizona state government, TSMC may eventually establish an industrial cluster in the state consisting of 10 fabs, 2 advanced packaging plants, and 1 R&D center, covering R&D, wafer fabrication, packaging, and testing.
The first fab in Arizona has already started mass production using 4nm-class process technology, with yield rates comparable to those of TSMC's plants in Taiwan. The second fab will adopt 3nm technology and is scheduled to begin mass production in 2027, while subsequent fabs will cover 2nm, A16, and even more advanced technologies. Upon completion of all projects, approximately 30% of TSMC's advanced process capacity for 2nm and below could be located in Arizona.
However, the $265 billion figure represents a multi-year investment plan and does not mean that the capital will be fully deployed in the short term. The construction and production timelines for the new fabs will still be determined gradually, based on customer orders, AI market demand, and the maturity of the local supply chain.
TSMC's expanded investment in the US offers at least four key opportunities.
First, it further binds TSMC with its core US customers. Apple, Nvidia, AMD, Broadcom, Qualcomm, and cloud computing providers all require stable advanced-process capacity. Having facilities close to customer headquarters and the data center supply chain facilitates long-term planning, product introduction, and capacity negotiations.
Second, it captures a geopolitical premium. As supply chain security becomes a key procurement metric for customers, domestic US capacity itself carries strategic value. Some customers may be willing to bear higher costs for a more secure and traceable supply chain.
Third, it mitigates policy and tariff risks. As the US continues to promote semiconductor localization, imported chips may face stricter tariffs, government procurement requirements, or security review rules in the future. Establishing domestic capacity ahead of time reduces the risk of TSMC's customer relationships being disrupted by trade policies.
Fourth, it prevents competitors from seizing market share. Intel and Samsung are both eager to win US orders for advanced chip foundry services. By expanding its Arizona capacity, TSMC can reduce the incentive for customers to turn to competitors due to local manufacturing requirements.
From this perspective, although investing in the US increases costs, it is also a defensive investment for TSMC to protect its market share and pricing power.
The cost of TSMC's expansion of US investments is gradually being reflected in its margins.
The company's second-quarter net profit grew 77.4% year-on-year, with a gross margin of 67.7%, but management also admitted that overseas fab expansions have already had a dilutive effect. As more projects enter the construction and mass production phases, overseas fabs are expected to drag down the gross margin by 2 to 3 percentage points initially, with the impact potentially expanding to 3 to 4 percentage points in the later stages.
The cost of building factories in the US is significantly higher than in Taiwan, primarily due to higher labor and training expenses, a shortage of semiconductor talent, an immature local supply chain, and increased expenditures in plant construction, logistics, and operations.
Morningstar analyst Phelix Lee estimates that, affected by the timing of subsidy disbursements, tax credits, and other cost variations, the cost of TSMC producing chips in the US could be 20% to 50% higher than in Taiwan.
However, this pressure remains within TSMC's manageable range for now. Demand for AI chips is robust, capacity for advanced nodes remains tight, and combined with TSMC's lack of comparable rivals in the high-end foundry segment, the company possesses strong bargaining power.
According to media reports, TSMC is considering raising prices for advanced processes and some mature processes in 2027, with price hikes for some products potentially reaching 10%. If these price increases materialize, customers such as Apple, Nvidia, and AMD may absorb a portion of the overseas manufacturing costs.
Therefore, while building plants in the US will indeed squeeze TSMC's profit margins, it currently resembles a manageable, long-term strategic cost. TSMC is trading short-term pressure on its gross margin for a more diversified supply chain, lower policy risks, and more stable partnerships with US customers.
Apple ( AAPL) is a major customer of TSMC's first Arizona factory, as well as one of the first and the largest customer of the nearby Amkor advanced packaging plant. In the future, Apple chips produced locally by TSMC can be packaged and tested nearby, further refining the domestic U.S. supply chain.
For Apple, the primary value of TSMC's investment in the U.S. is to mitigate the risk of concentrated advanced chip production and to support its long-term commitment to expanding U.S. manufacturing. However, Apple is highly sensitive to product costs, and the higher expenses of U.S. manufacturing could limit its share of local sourcing.
Nvidia ( NVDA) is receiving more direct capacity support, with Blackwell wafers already in mass production at TSMC's Arizona plant. Subsequent capacity expansions in 2nm and below processes, along with advanced packaging, will provide Nvidia with a greater supply of domestic U.S. AI chips.
Against the backdrop of growing demand for AI accelerators, advanced process nodes and packaging capabilities directly affect Nvidia's shipment volumes. Compared to merely reducing supply chain risks, new capacity is more likely to translate into data center revenue for Nvidia. Therefore, when measured by AI capacity and revenue growth potential, Nvidia may be the largest incremental beneficiary of TSMC's U.S. investment.
AMD ( AMD ), Broadcom ( AVGO ), Qualcomm ( QCOM ), as well as Amazon, which has its own proprietary chip business ( AMZN ), Google ( GOOGL) and Microsoft ( MSFT ), may also become subsequent beneficiaries. As the demand for custom AI chips grows, TSMC's U.S. factories will serve not just traditional chip design companies, but the entire U.S. AI infrastructure supply chain.
TSMC's investment in the US has increased to $265 billion, driven by both the US government's policy push to localize chip manufacturing and the growing real demand from Apple, Nvidia, AMD, and cloud computing companies.
For TSMC, its US expansion can mitigate geopolitical and trade risks, strengthen customer relationships, and protect its advanced foundry market share, with the trade-offs being higher capital expenditures, production costs, and pressure on profit margins.
In the long run, whether TSMC's US investment can achieve desired returns will largely depend on the sustainability of AI demand, whether the Arizona plant can ramp up capacity utilization, and whether the local supply chain can gradually mature. If these conditions are met, the US plants will not only serve as TSMC's overseas production bases but could also become one of the most vital strategic nodes in the global AI chip supply chain.