The next crack in the AI chip supply chain: How 2026 export controls will reshape the global semiconductor industry

Introduction

In 2025, the global semiconductor industry is once again overshadowed by the specter of geopolitics. Deloitte Insights' latest forecast points out that by 2026, a series of technologies that were previously not strictly controlled—from etching and GAA transistors in front-end chip manufacturing, to advanced packaging in the back-end, to EDA software tools and AI model weights—will become new bottlenecks in the supply chain. This means that the fragility of the global semiconductor supply chain has not only failed to ease, but is instead spreading into deeper and broader dimensions.

The core judgment of this forecast is that over the next two years, at least US$30 billion globally will be invested in key technology areas such as EUV lithography equipment and high-bandwidth memory (HBM) co-packaging tools, but these investments will be directly impacted by trade barriers. In contrast, the AI chip market supported by these technologies has an annual scale of up to US$300 billion. The stark disparity between investment and market size precisely illustrates the leverage effect of critical links in the supply chain—a single small valve can control the flow of blood for the entire industry.

This article will deconstruct the industrial logic behind Deloitte's forecast from four dimensions: the industry chain, technology routes, market competition, and geopolitics, and answer a core question: Is the global semiconductor industry entering a new era dominated by "bottleneck" technologies?

Background: From "Manufacturing Bottlenecks" to "Design Chokepoints"

Over the past few decades, the fragility of the semiconductor supply chain has been concentrated in the manufacturing stage—advanced lithography machines, high-purity silicon wafers, specialty gases, and so on. But Deloitte points out that between 2024 and 2025, the scope of export controls has expanded significantly, extending from physical equipment to intangible software and design methodologies. In December 2024, the United States further expanded export controls to cover software and tools that support the development of advanced computing nodes, directly reaching into EDA (Electronic Design Automation), the "operating system" of chip design.

Meanwhile, GAAFET (Gate-All-Around Field-Effect Transistor), a key transistor architecture for sub-5nm/sub-3nm nodes, became a target of export controls in 2025. This marks an important turning point: in the past, export controls mainly targeted manufacturing equipment, but now design methodology itself is being weaponized. AI model weights—the core parameters that determine the intelligence level of AI chip design tools—have also begun to enter the regulatory purview.

Deloitte expects that by 2026, EDA and logic design companies involved in AI accelerators will face stricter compliance reviews, including fine-grained disclosure requirements for entities, locations, and end uses. Chip companies' evaluation hardware, and even model weights used for testing, may become subject to tracing.

In-Depth Analysis

Technology Impact: How Are Technology Routes Being Redirected by Geopolitics?GAAFET is currently the most core technology node in the semiconductor industry. Compared with FinFET, GAAFET has clear advantages in power consumption and performance, making it the inevitable choice for AI computing chips. However, when GAAFET design tools and process kits are placed under export controls, fabs in non-U.S.-allied countries will face an awkward situation: either continue using the older-generation FinFET node, or switch to domestic EDA tools—but both paths mean longer product cycles and reduced competitiveness.

More subtle is the control over AI model weights themselves. Modern EDA tools are increasingly incorporating AI-assisted design, and the capability of AI-assisted design depends directly on the scale and precision of the model weights used for training. When these weights are restricted from flowing across borders, the pace of global chip design innovation will develop a regional scissors gap: regions with advanced model weights will have higher design efficiency, while restricted regions fall into "generational stagnation."

Supply Chain Impact: Who Benefits? Who Gets Hurt?

Deloitte explicitly lists several new "choke points":

• Etching equipment: The U.S. has imposed additional export restrictions on tools used for precision etching of advanced AI chips. At sub-5nm nodes, multi-patterning techniques (double, quadruple, spacer-based patterning) place extremely high demands on etching precision, and the U.S. controls the core IP of global etching equipment—even equipment manufactured overseas may face controls if it uses U.S. technology.
• EUV lithography equipment: ASML's EUV equipment from the Netherlands has long been embargoed, but Deloitte predicts that optical components (lenses, mirrors) and photomasks (reticles) could become the next batch of restricted parts.
• Specialty gases and critical minerals: Silane, fluorinated derivatives, and minerals such as gallium, germanium, and antimony are indispensable input materials for advanced-node manufacturing. The export concentration of these materials is extremely high, and countries such as China and Russia have begun implementing countermeasures.
• Advanced packaging tools: Back-end process equipment such as HBM co-packaging is explicitly listed in Deloitte's predicted list of technologies affected by trade barriers.

The beneficiaries of this round of controls are clearly those "vertically integrated" giants that can simultaneously master equipment, materials, software, and packaging technology—such as Samsung, Intel, and TSMC, which has supply-chain operations in the U.S. In contrast, small and medium-sized chip design companies that rely purely on global procurement, as well as fabs located in regions on the control list, will become the main risk bearers.

Competitive Landscape: The Game Among Three Countries and Four Regions

Deloitte predicts that in 2026, sub-5nm and sub-3nm capacity will accelerate its landing in the United States, Taiwan, and South Korea. This actually acknowledges a reality: the production of advanced process nodes will become increasingly concentrated in a few "trusted" regions.### Regional Implications: The Rise of Southeast Asia and the Marginalization of the EU?

In the broader context of supply chain relocation, Southeast Asian countries such as Malaysia, Vietnam, and Singapore are accelerating their uptake of packaging and testing capacity. Although Deloitte's research does not focus on this, the industry trend is already clear: multinational chip companies are dispersing packaging and testing operations to less regulated countries to avoid geopolitical risks. Meanwhile, although the EU has introduced its own CHIPS Act, its actual control over advanced manufacturing and design tools remains limited.

Investment Perspective: $30 Billion in Leverage to Unlock a $300 Billion Market

Deloitte's forecast is highly suggestive: in 2026, investment in "critical technologies" affected by trade barriers will be about $30 billion, while the AI chip market supported by these technologies will be about $300 billion. This means that although investment in supply chain segments is relatively small, its strategic leverage far exceeds the market itself. For investors, what matters is no longer just the valuation of chip design companies, but the "bottleneck index" of the entire supply chain—whoever controls the bottleneck holds the pricing power.

Long-Term Outlook: Strategic Possibilities for Three to Ten Years- 三年维度（2026-2028）：出口管制范围持续扩大，AI模型权重和EDA工具成为新战场。全球先进制程产能将进一步集中，中国在成熟制程上的产能占比继续上升。 - 五年维度（2026-2030）：新技术路线开始出现。除了GAAFET，CFET（互补场效应晶体管）、光互连等下一代技术将进入预备阶段。但多极化的管制环境可能导致半导体技术标准分裂——各地区发展彼此不兼容的设计工具和工艺生态。 - 十年维度（2026-2035）：全球半导体供应链可能从“全球化分工”转变为“区域化闭环”。美国、中国、欧洲、东南亚各自形成完整的产业链，但整体运行效率将低于全球分工时代。

Industry Chain Analysis: Chain Reactions from Upstream to Downstream

To more clearly illustrate the impact, we break the semiconductor industry chain into three tiers: upstream, midstream, and downstream.

Upstream: The Double Stranglehold of Equipment and Materials

• Lithography equipment: EUV is monopolized by ASML, but optical components and photomasks have become new points of restriction.
• Etching equipment: U.S. intellectual property covers the world's major etching equipment manufacturers, including Lam Research and Applied Materials.
• Materials: Exports of specialty gases (silane, fluorinated derivatives) and critical minerals (gallium, germanium, antimony) are mostly state-level actions and can easily be weaponized.
• EDA and IP: Synopsys, Cadence, and Siemens EDA dominate the market. If cut off, design would regress by a decade.

Midstream: Bottlenecks in Manufacturing and Packaging

• Front-end manufacturing: GAAFET process validation requires a complete process design kit (PDK) and EDA toolchain; restricted regions will find it difficult to access the latest design flows.
• Back-end packaging: Advanced packaging technologies such as HBM co-packaging and CoWoS are led by TSMC and Samsung, becoming a key bottleneck for AI chip performance. Deloitte has listed co-packaging tools as a technology affected by trade barriers, meaning this segment may also be brought under export controls.

Downstream: Structural Divergence in the AI Chip Market

• Beneficiaries: Ecosystems with proprietary EDA capabilities, advanced packaging capacity, and leading GPU vendors such as NVIDIA/AMD will hold an advantage in compliance cost control.
• Those at risk: Design companies highly dependent on the Chinese market, small and mid-sized ASIC vendors, and enterprises unable to enter the "trusted supplier" list.

ConclusionDeloitte's latest forecast reveals a stark industry reality: competition in the global semiconductor supply chain has surpassed traditional manufacturing capabilities and entered a phase of all-out confrontation across "design tools - transistor architecture - packaging processes - material supply." By 2026, we will see $30 billion in critical technology investments struggling within the cracks of regulatory controls, while the $300 billion AI chip market determines the enormous stakes of this game.

For industry decision-makers, the most important judgment is: supply chain security is no longer a supply chain issue, but a geopolitical issue. Any company's long-term strategy must treat "compliance capability" as a core competitiveness on par with R&D and manufacturing. As for the global semiconductor industry, an old era is coming to an end—a new era defined by export controls and supply chain fragmentation has fully begun.