From EUV to Etching: New Bottlenecks in the AI Chip Supply Chain in 2026 and Global Industry Restructuring

Introduction

Geopolitical tensions and trade restrictions are pushing the fragility of the semiconductor supply chain to new heights. From advanced lithography equipment to electronic design automation (EDA) tools, and then to specific chemicals and critical minerals, a growing number of technologies once considered "behind-the-scenes links" are starting to become the focus of export controls. In its latest forecast, Deloitte notes that by 2026, the semiconductor supply chain will see more "breakpoints"—including those involving etching equipment, gate-all-around (GAA) transistor manufacturing, EDA software, and tools related to AI model weights—that are significant enough to influence the direction of the global AI industry.

The importance of this change cannot be overstated. The performance of AI systems depends on a highly concentrated global technology stack: advanced logic design, front-end manufacturing at leading nodes, and advanced packaging. Each segment is dominated by a few suppliers, and export controls are turning technological advantages into geopolitical tools. Deloitte expects that at least $30 billion will be invested globally in 2026 in key technologies affected by trade barriers (such as EUV lithography equipment and high-bandwidth memory co-packaging tools), but these investments remain a drop in the bucket relative to the approximately $300 billion AI chip market they support. This means that every new restriction order could trigger a chain reaction across the entire value chain.

This article will analyze the underlying logic of this supply chain restructuring from four dimensions—industry chain, technology roadmap, competitive landscape, and regional impact—and explore how semiconductor companies should adjust their strategies.

Background: From "Globalized Division of Labor" to "Managed Supply Chain"

Over the past three decades, the semiconductor industry has established the world's most refined division-of-labor system: the United States dominates EDA and chip design IP, the Netherlands and Japan supply lithography and materials, Taiwan and South Korea take on advanced manufacturing, and mainland China focuses on mature processes and packaging/testing. This model is highly efficient, but its fragility has also been fully exposed amid intensifying geopolitical competition.

In 2024 and 2025, the United States successively tightened and adjusted restrictions on multiple critical semiconductor technologies. Initially, the restrictions focused on the most advanced chips themselves and EUV lithography machines, but they quickly expanded to EDA tools, GAAFET chip design, precision etching equipment, and even AI model weights. Deloitte believes that this trend will not slow down in 2026; rather, it will spread to a broader range of semiconductor equipment, materials, software, design tools, and packaging/assembly tools.

Notably, the impact of export controls is not one-way. China is seeking alternatives through customized deep ultraviolet (DUV) technology and multiple patterning processes, albeit at a slower speed and higher cost. This "technology decoupling" is forming two de facto technology ecosystems, whose long-term costs far exceed the short-term losses.

In-Depth Analysis

Technology Impact: Technology Roadmaps DivergeThe manufacturing of advanced AI chips depends on a series of high-barrier technologies. Currently, GAA transistor architecture is the mainstream choice for sub-5nm and sub-3nm logic designs, and its performance and power consumption advantages are critical for generative AI workloads. However, restrictions targeting GAAFET chips will leave foundries in non-U.S. allied countries facing a dilemma: either advance verification without EDA tool support, or fall back to mature FinFET nodes, both of which would weaken product competitiveness.

Precision etching technology is another key "breaking point." To fabricate fine structures on chips, the industry commonly uses double, quadruple, and spacer-based patterning processes, which rely heavily on etching equipment of U.S. origin. Deloitte points out that the U.S. not only regulates domestic equipment but also restricts overseas manufacturing equipment that uses U.S. etching technology IP, making it impossible for any wafer fab that depends on this technology to circumvent the blockade.

In addition, optical components (lenses, mirrors), photomasks, specialty gases (such as silane and fluorinated derivatives), and critical minerals (gallium, germanium, antimony) are also becoming friction points. This means that even a piece of equipment without direct import restrictions may be undeliverable due to a ban on a certain component or material.

Supply Chain Impact: Who Benefits, Who Feels the Pressure?

From an industry chain perspective, the most direct beneficiaries may be companies with independent equipment capabilities. For example, U.S. equipment suppliers such as Applied Materials, Lam Research, and KLA, while facing rising compliance costs, also gain pricing power in unrestricted markets because they possess core technologies. ASML's EUV lithography machines in the Netherlands remain a scarce resource, and their customer lists and delivery priorities will be increasingly influenced by political factors.

At risk are companies that rely on a single supplier or a single region. Chinese chip manufacturers will have to accept mature DUV plus multiple patterning processes, and their advanced process development will slow significantly. OSAT (outsourced assembly and test) providers may also be affected—high-bandwidth memory co-packaging tools are critical for advanced AI chips, but export controls on such tools will directly impact their ability to serve the global market.

Particularly noteworthy is the EDA ecosystem. Companies such as Synopsys, Cadence, and Siemens EDA dominate the global EDA market. If they must comply with stricter compliance reviews, their Chinese customers' design capabilities will be significantly weakened, and could even cause the entire Chinese AI chip design industry to regress by 2-3 product cycles.

Competitive Landscape: Five-Polar Differentiation in the Competitive Landscape

Deloitte predicts that by 2026, production of sub-5nm and sub-3nm will accelerate in the United States, Taiwan, and South Korea. This means that, regardless of geopolitical changes, the technological leadership of these three regions will not be shaken, but will instead be further consolidated as other countries are unable to keep up.The United States, Taiwan (China), and South Korea will continue to form the "iron triangle" of advanced process technology. TSMC and Samsung's duopoly in the foundry market will become even more entrenched. Intel Foundry is making efforts, but in the short term it will be difficult to shake the positions of the first two.

Mainland China, for its part, is being forced to deepen its work on mature process nodes and specialty technologies. Although advanced process technology is restricted, its demand for chips in areas such as new energy vehicles, industrial control, and the Internet of Things remains huge, which provides another market for mature-process equipment and materials.

Europe and Japan, meanwhile, will shift more toward the role of equipment and materials suppliers. Companies such as Tokyo Electron, Advantest, and Shin-Etsu Chemical may become "neutral suppliers," maintaining a balance between the two camps.

Regional Implications: The Geographic Reshaping of Global Supply Chains

The United States is trying to secure technological leadership through export controls, but at the cost of losing China, one of the world's largest semiconductor consumer markets. At the same time, the CHIPS and Science Act passed by the U.S. Congress is promoting domestic manufacturing, yet the lack of a supporting ecosystem for packaging, materials, and talent makes it difficult to form a complete closed loop in the short term.

Taiwan (China) is in the most delicate position. It is both the core of advanced process manufacturing and a geopolitical flashpoint. Should tensions arise in the Taiwan Strait, the global AI chip supply chain would suffer a devastating blow. The United States is pushing to shift advanced packaging and some manufacturing capacity to its own soil, but building a complete supply chain system will still take time.

South Korea's Samsung and SK Hynix dominate the memory and high-bandwidth memory (HBM) segments. AI chip performance is highly dependent on HBM, which gives South Korea unique bargaining power in the supply chain. Japan, for its part, continues to deepen its work in materials science and semiconductor equipment, especially photoresists and specialty gases. These are small fields, yet they determine the stability of the entire industrial chain.

Southeast Asia is gradually becoming a "buffer warehouse" in the U.S.-China rivalry. Malaysia has a strong foundation in packaging and testing, while Vietnam and Singapore are also attracting new chip manufacturing investment. But in the short term, Southeast Asia cannot replace China's position in manufacturing and assembly.

Investment Perspective: The Logic of Capital Expenditure Is Changing

Deloitte points out that multinational chip equipment companies need to adjust their capital expenditure plans by region. In the past, equipment makers and foundries could rely on unified global capacity planning. But now, compliance reviews, certification cycles, and installation cycles are all lengthening, making the certainty of equipment delivery a scarce resource.

For investors, this means re-evaluating the "supply chain resilience" premium of semiconductor companies. Companies with multi-region capacity, backup suppliers, and independently controllable technology will receive higher valuations. Those that are deeply dependent on a single country or a single customer will see their risk discount rise significantly.At least $30 billion in key technology investment in 2026 may seem massive, but compared with the $300 billion AI chip market, it is still just a "toll fee." The real value lies in who can be the first to provide solutions when the next bottleneck emerges.

Long-Term Outlook: Three Possibilities for the Next 3-5-10 Years

Within 3 years (by 2028): Export controls will continue to expand, but will not completely cut off technological exchanges. The United States, Taiwan, and South Korea will form an "advanced process alliance," while mainland China will accelerate domestic substitution in mature process nodes and specialty technologies. The global semiconductor supply chain will take on a "dual-track" character: one track serving the Western AI industry, the other serving China and friendly countries.

Within 5 years (by 2030): Advanced packaging and heterogeneous integration will become the new competitive focus. As Moore's Law slows down, the value of system-in-package and Chiplet architectures will become increasingly prominent, and AI chip performance will depend more and more on the synergistic optimization of packaging and memory. Whoever masters next-generation packaging technologies such as glass substrates and co-packaged optics will seize the initiative in the next round of the AI race.

Within 10 years (by 2035): The semiconductor supply chain may be restructured into several regional clusters. North America, East Asia, and Europe will each have a relatively complete chain from design to manufacturing, and the landscape of a globally unified market will become history. This restructuring will keep the industry's capital expenditures higher than in the past over the long term, but the diversity of technology and markets will also provide new soil for innovation.

Industry Chain Analysis: Complete Impact from Upstream to Downstream

• Upstream: Equipment (EUV, etching), materials (specialty gases, photomasks), software (EDA), critical minerals (gallium, germanium) — this is the segment where export controls are most concentrated and the point of greatest tension in the current global supply chain. Any new restriction could trigger multi-layered knock-on effects.
• Midstream: Wafer manufacturing and foundry — leading companies such as TSMC, Samsung, and Intel Foundry will face compliance pressure, but will also gain more orders because of the irreplaceability of their technology. Mainland China's SMIC and Hua Hong, meanwhile, are turning to mature process nodes and specialty technologies.
• Downstream: AI chip design, packaging and testing, system integration — design companies such as NVIDIA, AMD, and Broadcom need to reorganize their supply chain layouts, while OSAT vendors need to make breakthroughs in advanced packaging technology to meet the co-packaging requirements of HBM and optoelectronic devices.

Conclusion

Deloitte's forecast clearly maps out the future direction of the semiconductor supply chain: fragility will not disappear, but will instead extend from the equipment level to the software, materials, and algorithm levels. EDA tools, GAA transistor processes, precision etching, and other segments will become new strategic strongholds.For the industry, the most important thing is not trying to predict every specific restriction order, but realizing that supply chain resilience has replaced efficiency as the primary goal. Companies need to diversify their operations, ensure compliance, and increase investment in self-developed technologies. Whoever can adapt to this new normal of a "controlled supply chain" sooner will seize the initiative in the future AI race.