Search

US-TW chip deal built on leverage, not self-sufficiency

The deal confirms that Taiwan will continue to supply advanced wafers as a contract manufacturer for U.S. semiconductor customers such as NVIDIA, Broadcom and Apple. In exchange for tariff-free access to the U.S. market, Taiwan has committed to maintaining a steady pace of investment on U.S. soil, primarily through TSMC. The promised investment totals $500 billion, including $250 billion for fab construction and roughly eight fabs to be built over the next decade.

Today, Yole Group’s principal analyst, Pierre Cambou, invites you to explore the key takeaways from this new U.S.-Taiwan chip deal, shedding light on the treasure island and its worldwide influence. No need to fly across the globe for this overview.

Capacity migration will be gradual, and limited

The first leading edge TSMC fab, which is already operating in Phoenix, Arizona, produces wafers at the 4nm node. Future fabs are expected to follow every two years and will subsequently provide manufacturing at the 2nm node and below.

TSMC is in the process of shifting 10% of its own capacity to the U.S. by building new advanced node facilities there in addition to the ones it builds at home. This transfer will represent a gradual increase from around 3% of TSMC’s production capacity today. At the end of 2025, advanced nodes (sub-7nm) represented 40% of TSMC’s capacity and 60% of its revenues.

Whether Taiwan’s government would “allow” such a shift is often misunderstood. TSMC may be the world’s most strategically important contract manufacturer, but it remains highly dependent on its overseas customers. Taiwan’s political reality is also central. With a population of 23 million and an economy that is 4% of the size of mainland China’s, Taiwan’s de facto independence rests on U.S. military protection. That limits how much room Taipei has for maneuvering when Washington applies pressure for a partial transfer of manufacturing capacity to the U.S.

This context also helps to clarify Taiwan’s true position in the global supply chain. As of 2024, Taiwan is no longer the largest host country for semiconductor foundries given the emergence of manufacturing capacity in mainland China.

Pierre_CAMBOU-PCA_YINT
Pierre Cambou, Principal Analyst, Global Semiconductors at Yole Group
By foundry location, Taiwan now represents about 21% of global foundry capacity. This share is expected to decline to 19% by 2031.

Taiwanese companies, however, hold around 70% of outsourced (or “open”) foundry market share and should keep this lead in the medium term, since most of the growth in mainland China is now driven by integrated device manufacturers (IDMs) such CXMT, YMTC and Huawei affiliates.

This dominance in the open foundry market is often conflated with market share in the entire semiconductor industry. This confusion is encouraged by Taiwan, keen to emphasize its significance in the supply chain, and also by the U.S., which relies heavily on this outsourcing model and does not want to downplay Taiwan’s importance.

For Europe, South Korea, and Japan, Taiwan’s role is specific to the leadingedge nodes. In revenue terms, that dependency is relatively small. It remains a key technological dependance for specific products, while it underpins Europe’s own strategic leverage via ASML, the Dutch supplier of advanced-node lithography tools and Japanese leverage through photoresist companies like JSR and TOK.

Mutual dependence as deterrence

Against this backdrop, claims that the U.S. aims to produce 40% of the world’s advanced chips should be taken with a healthy pinch of salt. The U.S. currently accounts for around 10% of global production, a share it is expected to maintain through 2031. U.S.-headquartered companies have roughly half of their capacity abroad and control around 20% of overall global capacity.

Pierre Cambou from Yole Group
A realistic outcome is that the U.S. could control 20% of leading-edge semiconductor capacity by 2031; more than that within the next five years is wishful thinking.

With the U.S. slated to remain highly dependent on Taiwan for semiconductor supply, it wants to avoid both a takeover of the island by mainland China and any short-term conflict with China that would inherently lead to fab destructions.

China, meanwhile, is indirectly dependent on Taiwan because much of its consumer electronics industry relies on components sourced from the Taiwanese foundries and assembled by ODMs such as Foxconn (Hon Hai), which operate assembly plants on the mainland.

In that sense, Taiwan’s semiconductor excellence in semiconductor manufacturing acts as a silicon shield against war.

So why did Beijing oppose the agreement?

The deal will not reduce semiconductor manufacturing in Taiwan, but it will marginally reduce the cost to the U.S. from potential supply disruption by relocating some advanced capacity.

That slightly alters the strategic equation to the benefit of the U.S., and Beijing has every incentive to object, both politically and economically.

As the semiconductor industry grows more complex, decision makers need both strategic vision and technical insight. Yole Group delivers this clarity, with data spanning from wafer to system level. In 2026, Yole Group will launch new offerings that provide a clear, cross-market view of the global semiconductor landscape. As data volumes grow, Yole Group’s focus will be on curated, intelligently aggregated insights, delivered in formats designed to cut through the noise.

We look forward to hearing from you! Your feedback and questions are always welcome as we continue to shape Yole Group’s offerings.

Stay tuned!

About the author

Pierre Cambou, MSc, MBA, is Principal Analyst, Global Semiconductors at Yole Group.

Pierre’s mission is dedicated to market & technology analyses of the semiconductor industry.

At Yole Group, Pierre has authored numerous market and technology products. Acknowledged as an expert in the semiconductor industry, he is regularly interviewed and quoted by leading international media.

Previously, Pierre held several positions at Thomson TCS, which became Atmel Grenoble (France) in 2001 and e2v Semiconductors in 2006. In 2012, he founded a semiconductor startup, now part of Jooxter.

Pierre holds an engineering degree from Université de Technologie de Compiègne (France) and a Master of Science from Virginia Tech. (VA, USA). Pierre also graduated with an MBA from Grenoble Ecole de Management (France).

This article is reserved for logged-in users. You have 80% left to discover.
up