The next bottleneck in the AI chip supply chain: export controls on EDA, etching, and GAA will reshape the global semiconductor landscape.

The Next Choke Point in the AI Chip Supply Chain: Export Controls on EDA, Etching, and GAA Will Reshape the Global Semiconductor Landscape

Introduction

The global semiconductor supply chain is undergoing an unprecedented restructuring. In its latest forecast, Deloitte Insights points out that by 2026, etching in both front-end and back-end chip manufacturing, gate-all-around (GAA) transistor technology, electronic design automation (EDA) tools, and software tools supporting advanced AI models could all become new supply chain bottlenecks. At the same time, at least $30 billion will be directed toward EUV lithography equipment and high-bandwidth memory co-packaging tools affected by trade barriers—and behind this figure lies an AI chip market of approximately $300 billion. 1

What does this mean? At a time of explosive growth in AI computing demand, every critical node in the semiconductor supply chain could become a bargaining chip in geopolitical competition. This article will interpret the industrial logic behind Deloitte's forecast from the perspectives of the industry chain, technology roadmaps, competitive landscape, and regional impact, and analyze the deep-seated changes the semiconductor industry may face in the future.

Background: The Highly Interdependent AI-Driven Semiconductor Supply Chain

The performance of modern AI systems depends on a range of technologies distributed globally—advanced AI logic design, leading-edge front-end node manufacturing, advanced packaging, and the EDA toolchain that runs throughout. These segments involve IDMs, wafer foundries, equipment vendors, design service providers, outsourced semiconductor assembly and test (OSAT) companies, system integrators, and even multiple national governments, forming a highly interdependent global supply chain.

Deloitte's analysis shows that the scope of export controls and trade restrictions is expanding from purely chip products to a broader range of equipment, materials, software, and design tools. Compared with two or three years ago, the number of semiconductor supply chain segments subject to controls in 2025–2026 has increased significantly. From EUV lithography machines to precision etching equipment, from GAA transistor design to AI model weights, every segment could become a target of restriction.

This change is not an isolated event, but an inevitable outcome of global technology competition spreading from end products to core foundational technologies. When advanced chips become a national strategic resource, the underlying technologies needed to manufacture them naturally become the focus of competition.

In-Depth Analysis

Technology Impact: From Traditional Architectures to GAA, Design Tools Become New Barriers

#### GAA Transistors: The Threshold of Advanced Process Nodes GAAFET (Gate-All-Around Field-Effect Transistor) is a next-generation transistor architecture for nodes of 5nm and below. Compared with FinFET, it offers significant advantages in performance and power efficiency, making it especially suitable for generative AI workloads that require extremely high compute density. However, Deloitte points out that GAAFET-based chips have long been subject to export controls, and in December 2024 the United States further expanded controls on software and tools that support the development and design of advanced computing nodes.

This change has far-reaching implications for the global chip design ecosystem. In non-allied countries, fabs using GAAFET process design kits (PDKs) to develop leading-edge nodes will need stricter EDA tool support for verification in the future. If a region cannot obtain these tools, it may be forced to fall back to older, less efficient nodes or invest heavily in developing indigenous EDA capabilities — either option will lengthen product cycles and weaken market competitiveness.

#### EDA Tools: The "Invisible Gate" in AI Chip Design

EDA tools cover the entire flow of chip design, layout, simulation, AI-enhanced design, verification, and integration, and are a necessary condition for developing advanced AI accelerators. However, as AI itself becomes a source of capability for EDA tools, countries' control over AI model weights has also begun to indirectly affect chip design. Deloitte predicts that by 2026, EDA and logic design players will face stricter compliance reviews, including detailed disclosure requirements for foundry IP libraries, PDKs, performance test outputs, and evaluation hardware used for verification and fine-tuning.

This means that companies engaged in co-design of AI hardware and software may need to build secure IT infrastructure in trusted countries or restructure their workflows. For example, model weights could be stored in the United States or its allies' territory, while foundry partners are allowed to run remote tests. Although feasible, this model undoubtedly increases collaboration costs and may slow the pace of innovation.

#### Etching Equipment: A New Chokepoint in Precision Manufacturing

Etching is a key process in manufacturing chips at sub-5nm nodes. Modern AI chips require double, quadruple, and spacer-based patterning techniques to create fine structures, and the United States has imposed additional export restrictions on tools used for precision etching. Deloitte notes that etching equipment of U.S. origin — even equipment designed and manufactured abroad using U.S. etching technology IP — could become a new chokepoint in 2026. In addition, optical components in lithography machines (lenses, mirrors) and mask plates (photomasks) may also be subject to restrictions.EUV lithography equipment is monopolized by ASML in the Netherlands, while the United States indirectly controls its destination by coordinating export restrictions. The Netherlands and the United States have jointly pressured to keep EUV equipment out of mainland China. At the same time, China is developing its domestic lithography capabilities by improving deep ultraviolet (DUV) technology and adopting multi-patterning processes. This approach is effective, but slower and more costly.

In the etching equipment sector, US companies Lam Research and Applied Materials dominate. If export controls extend to etching equipment and its IP derivatives, China's advanced-process R&D will face enormous obstacles. In addition, specialty gases such as silane and fluorinated derivatives, as well as key minerals such as gallium, germanium, and antimony, are indispensable input materials for advanced-node manufacturing. Controls on these aspects will further exacerbate supply chain fragmentation.

#### Midstream Foundry and Packaging: Advanced Processes Concentration Accelerates, Mature Processes Diverge

Under the export control context, Deloitte predicts that capacity expansion for sub-5nm and sub-3nm processes in the United States, Taiwan, and South Korea will accelerate in 2026 and beyond. This means advanced-process manufacturing capacity will become further concentrated in a few trusted regions. Meanwhile, mainland China is likely to more firmly pursue a technology path of mature processes plus multi-patterning, deepening its DUV capabilities.

The packaging segment also faces pressure. High-bandwidth memory (HBM) co-packaging tools have been listed as technologies that may be affected by trade barriers. As HBM becomes a key bottleneck for AI chip performance, the importance of advanced packaging has risen sharply. In the future, self-sufficiency in packaging technology may be the next critical battleground after lithography.

#### Downstream AI Applications: Rising Costs and Complexity

For AI chip design companies and cloud service providers, compliance costs are rising. Whether it is obtaining EDA licenses, verifying GAA designs, or deploying testing processes with international collaboration, all of these mean longer timelines and higher costs. For AI infrastructure projects that rely on global supply chains, geopolitical risk is gradually becoming a core variable in investment decisions.

Competitive Landscape: The Next Move in Global Semiconductor Competition

The expansion of export controls is reshaping the map of global semiconductor competition. The United States maintains indirect dominance over global advanced processes by controlling upstream technologies such as EDA and etching equipment. The Netherlands, as the exclusive supplier of EUV, has gained greater influence in rule-making. Taiwan and South Korea, leveraging their advanced-process capacity, have become targets of courting by various parties, while also facing pressure to take sides.Under pressure, mainland China is accelerating its push toward semiconductor self-sufficiency. In the EDA field, although domestic tools still lag behind Synopsys, Cadence, and others by a generation, companies such as Huawei have begun to make arrangements; in the lithography field, companies like SMEE have made certain breakthroughs in DUV technology. Although it is difficult to reach the forefront of advanced process nodes in the short term, China's huge domestic demand market and continued investment may give rise to an independent technological path.

For equipment giants, regionalized operations have become the new normal. Companies such as ASML and Lam Research need to readjust their capital expenditure plans across regions and adapt to longer certification, upgrade, and installation cycles. Foundries such as TSMC, Samsung, and Intel may face more scrutiny from the design side in terms of regulatory compliance, IP protection, and technology transfer.

Regional Implications: Reshaping of the Global Industrial Chain Landscape

• United States: Maintains technological leadership through export controls, while attracting manufacturing back through the CHIPS Act, but the lack of domestic EUV capacity is a structural shortcoming.
• China: Forced to take an independent route, mature process nodes and specialty processes may become breakthroughs, but the gap in frontier fields such as GAA and EUV may widen further.
• Taiwan, China: Its hub position in advanced process nodes is irreplaceable, but it faces a dilemma between technology confidentiality and customer trust.
• South Korea: Samsung and SK Hynix are advancing on both memory and foundry fronts, while benefiting from the explosion in HBM demand, but they remain highly dependent on the Chinese market.
• Japan: Has deep accumulation in semiconductor materials and equipment, and the Japanese government is restarting advanced process plans, increasing its importance in geopolitics.
• Europe: The Netherlands plays a key role, but Europe as a whole remains marginalized in chip manufacturing and is trying to catch up through the European Chips Act.
• Southeast Asia: Malaysia, Vietnam, and other places are becoming emerging bases for packaging and testing and mature process nodes, but due to talent and infrastructure constraints, they will struggle to take on high-precision segments in the short term.

Investment Perspective: The Strategic Value Behind $30 Billion

Deloitte predicts that in 2026, at least $30 billion will be spent on EUV lithography equipment and HBM co-packaging tools affected by trade barriers. Compared with the approximately $300 billion AI chip market, this investment accounts for only 10%, yet it determines the capacity ceiling and innovation capability of the entire AI chip industry. From an investment perspective, these "bottleneck technologies" have extremely high strategic leverage.

For investors, supply chain security is becoming a core consideration that goes beyond purely financial metrics. The degree of independent controllability in equipment, materials, EDA, and advanced packaging will determine a company's long-term value and risk resilience. At the same time, export control policies themselves have created new investment themes—startups that can circumvent restrictions and achieve substitution may receive ultra-large funding.### Long-Term Outlook: The Changing Landscape Over the Next 3-10 Years

Next 3 years (2026-2028): The export control list may expand further, with advanced packaging and specialty chemicals becoming new focal points. The United States, Taiwan (China), and South Korea will continue to expand advanced process capacity, while mainland China builds advantages in mature processes and specialty technologies. The global semiconductor industry may see the emergence of a prototype of "two technology systems": one is an alliance-based system centered on GAA and EUV, and the other is a self-reliant system relying on multi-patterning and mature processes.

Next 5 years (2028-2030): China may reach a node equivalent to nearly 7nm through DUV multi-patterning technology, but the relative gap will remain. Advanced packaging will surpass traditional Moore's Law and become the main driver of AI chip performance improvements. Packaging technologies such as TSMC's CoWoS may become more important than some lithography steps. Meanwhile, emerging manufacturing bases in India and Southeast Asia will gradually take shape, but they will not be able to shake the existing landscape.

Next 10 years (2030-2035): The focus of the semiconductor supply chain will shift from "globalized division of labor" to "regional multipolarity." The United States, China, Europe, Japan, and Southeast Asia may each dominate specific segments. True "complete self-sufficiency" will be difficult to achieve in any country, but fragmentation of technical standards and IP systems will become the norm. In this context, companies with the ability to collaborate across systems will gain the greatest competitive advantage.

Industry Chain Analysis: The Complete Impact from Upstream to Downstream

Upstream: Equipment, Materials, and Key Minerals

• EUV lithography equipment: ASML is the sole supplier, and export controls have turned "advanced equipment" into a strategic weapon; China's investment in the DUV route will accelerate, but its capacity and yield still face enormous challenges.
• Etching equipment: US-dominated precision etching tools are critical to sub-5nm nodes; export controls may split the global etching equipment market, giving mainland China's domestic equipment a development window.
• Specialty gases and critical minerals: Silane, fluorides, gallium, germanium, antimony, and other materials are indispensable inputs for manufacturing advanced chips. China is the world's main exporter of gallium and germanium and may play a role in countermeasures, exacerbating supply chain frictions.
• Photomasks and optical components: The United States may further restrict the flow of these core components to non-allied countries, thereby affecting the maintenance and upgrade of global lithography machines.

Midstream: Design, Manufacturing, and Packaging- EDA/IP: The US-based Synopsys, Cadence, and Siemens EDA dominate the global market; if export controls are escalated, countries like China will be forced to develop domestic EDA, but substitution will be extremely difficult in the short term. - Wafer foundry: TSMC, Samsung, and Intel will further consolidate their advantages in advanced process nodes; mainland Chinese foundries such as SMIC may focus on mature process nodes plus specialty processes. - Packaging and testing: HBM co-packaging technology has become a key competitive factor for AI chips; the advanced packaging layouts of companies like TSMC, ASE, and Amkor will determine the bandwidth and power consumption of future AI chips.

Downstream: AI chip systems and cloud infrastructure

• AI accelerators: Designs from NVIDIA, AMD, Google TPU, etc., rely heavily on TSMC's advanced process nodes and CoWoS packaging; export controls may force some design companies to adjust product plans and even increase reliance on non-US EDA.
• Data center HPC: Capital expenditure for AI infrastructure construction is soaring, but supply chain uncertainty may push up project costs and prompt cloud vendors to stockpile key chips and equipment in advance.
• End-use applications: Downstream applications such as autonomous driving and generative AI services will face commercialization pressure due to rising chip costs—especially in regions that rely on advanced AI chips.

Conclusion: The most important industry judgment

The "bottleneck" of the semiconductor supply chain is expanding from lithography machines to more areas such as EDA, etching, and packaging. Deloitte's forecast clearly reveals a trend: export controls are no longer "point sanctions" targeting specific products, but a "systemic barrier" covering design, manufacturing, materials, and software.

For the global semiconductor industry, this is both a challenge and an opportunity. In the short term, supply chain fragmentation will drive up costs and reduce efficiency, especially in the field of advanced process nodes. In the long term, however, a multipolar supply chain system may give rise to more technological innovation and industrial pathway options. The huge end-user demand in the mainland China market will force EDA, equipment, and materials companies to seek new cooperation models, thereby accelerating the formation of a "de-Americanized" technology stack.

Ultimately, no country can achieve complete self-sufficiency in all key links of AI chips. In the next decade, cross-border operations, technology licensing, and compliant innovation will become core capabilities for semiconductor companies. And those companies that can build bridges in a divided world will define the next generation of the global semiconductor landscape.

---

*This article is based on the forecast report "New supply chain tech" published by Deloitte Insights. All facts and data are cited from this report and do not constitute investment advice.*1: Deloitte Insights, “New supply chain tech”, 2025. URL: https://www.deloitte.com/us/en/insights/industry/technology/technology-media-and-telecom-predictions/2026/new-supply-chain-tech.html