ARTIFICIAL INTELLIGENCE

“The Chinese chip industry has done an amazing job of catching up. I think they’ve probably exceeded most people’s expectations in this.” – Matt Sheehan – Carnegie Endowment for International Peace

Matt Sheehan’s remark captures a central surprise of the last decade in geopolitics and technology: the speed and resilience of China’s semiconductor ascent under heavy external pressure.

At the heart of this story is China’s effort to close what used to look like an unbridgeable gap with the United States, Taiwan, South Korea, Japan, and Europe in advanced chips, tools, and know-how. National programs such as “Made in China 2025” explicitly targeted semiconductors as a strategic chokepoint, aiming to localize production and reduce dependence on foreign suppliers in logic chips, memory, and manufacturing equipment.² This was initially greeted with skepticism in many Western capitals and boardrooms, where the prevailing assumption was that export controls, restrictions on advanced tools, and China’s own technological lag would keep it permanently behind the frontier.

Sheehan’s observation points to where expectations proved wrong. Despite sweeping export controls on leading-edge lithography tools and high-end AI chips, Chinese firms have made faster-than-anticipated progress across the stack:

• In manufacturing equipment, domestic suppliers have rapidly increased their share in key process steps such as etching and thin-film deposition.^1,4 By 2025, the share of domestically developed semiconductor equipment in China’s fabs had risen to about 35%, overshooting Beijing’s 30% target for that year.¹ Local champions like Naura and AMEC have pushed into complex tools, delivering CVD, ALD, and other thin-film equipment for advanced memory and logic production lines used by major Chinese foundries such as SMIC and Huahong.^1,4
• In capital investment and ecosystem depth, mainland China has become the largest market in the world for semiconductor manufacturing equipment, with projected spending around $39 billion in 2026—more than Taiwan or South Korea.⁴ This spending fuels a dense local ecosystem of design houses, foundries, packaging firms, and toolmakers that did not exist at comparable scale a decade earlier.
• In AI and accelerator chips, Chinese firms have developed increasingly capable domestic alternatives even as they still seek access to high-end Nvidia GPUs. China’s AI sector drew global attention in 2025 with breakthroughs by firms such as DeepSeek, whose large models forced global competitors to reassess Chinese capabilities.⁵ At the same time, Beijing has leveraged its regulatory power to steer large platforms such as Alibaba and ByteDance toward a mix of imported and home-grown accelerators, explicitly tying access to Nvidia chips (like the H200) to parallel purchases of Chinese solutions.^3,5 This policy mix illustrates how industrial strategy and geopolitical bargaining are being fused to accelerate domestic chip progress while still tapping global technology where possible.³
• In memory and specialty devices, companies like Yangtze Memory Technologies (YMTC) have moved up the learning curve in 3D NAND and are investing heavily in further technology upgrades, DRAM development, and forward-looking R&D that demand increasingly sophisticated domestically supplied equipment.^1,4 These investments both absorb and shape the capabilities of the Chinese toolmakers that Sheehan has in mind.^1,4

Sheehan’s quote is also rooted in the broader geopolitical context he studies: the U.S.–China technology rivalry, where semiconductors are the most strategically sensitive terrain. Washington’s use of export controls on advanced lithography, EDA tools, and high-end AI chips was designed to “slow the pace” of Chinese military-relevant innovation. The expectation in many Western policy circles was that these controls would significantly impede Chinese progress. Instead, controls have:

• Reshaped China’s development path—from importing at the frontier to building domestically at one or two nodes behind it.
• Accelerated Beijing’s urgency to build local capability in areas once left to foreign suppliers, such as inspection and metrology tools, deposition, and etch.^1,4
• Incentivized enormous sunk investment and political attention to semiconductors in China’s five-year plans, where AI and chips now sit at the very center of national strategy.⁵

Although China still faces real bottlenecks—most notably in extreme ultraviolet (EUV) lithography, highly specialized tools, and some advanced process nodes—its system-level catch-up has been broader and quicker than many analysts predicted.^2,5 That is the gap between expectation and reality that Sheehan is highlighting.

Matt Sheehan: The voice behind the quote

Matt Sheehan is a leading analyst of the intersection between China, technology, and global politics. At the Carnegie Endowment for International Peace, he has focused on how AI, semiconductors, and data flows shape the strategic competition between the United States and China. His work sits at the frontier of what is often called “digital geopolitics”: the study of how code, chips, and compute influence power, security, and economic advantage.

Sheehan’s analysis is distinctive for three reasons:

• He combines on-the-ground understanding of Chinese policy and industry with close attention to U.S. regulatory moves, giving him a bilateral vantage point.
• He approaches policy not just through national security, but also through the innovation ecosystem—research labs, startups, open-source communities, and global supply chains.
• He emphasizes unexpected feedback loops: how U.S. restrictions can accelerate Chinese localization; how Chinese AI advances can reshape debates in Washington, Brussels, and Tokyo; and how commercial competition and security fears reinforce each other.

This background makes his judgment on the pace of Chinese semiconductor catch-up particularly salient: he is not an industry booster, but a policy analyst who has watched the interplay of strategy, regulation, and technology on both sides.

The broader intellectual backdrop: leading theorists of technology, catch-up, and geopolitics

Behind a seemingly simple observation about China’s chip industry lies a rich body of theory about how countries catch up technologically, how innovation moves across borders, and how geopolitics shapes advanced industries. Several intellectual traditions are especially relevant.

1. Late industrialization and the “catch-up” state

Key figures: Alexander Gerschenkron, Alice Amsden, Ha-Joon Chang

• Alexander Gerschenkron argued that “latecomer” countries industrialize differently from pioneers: they rely more heavily on state intervention, banks, and large industrial enterprises to compress decades of technological learning into a shorter period. China’s semiconductor push—state planning, giant national champions, directed finance, and targeted technology acquisition—is a textbook example of this latecomer pattern.
• Alice Amsden studied how economies like South Korea used targeted industrial policy, performance standards, and learning-by-doing to build globally competitive heavy and high-tech industries. Her emphasis on reciprocal control mechanisms—state support in exchange for performance—echoes in China’s mix of subsidies and hard metrics for chip firms (e.g., equipment localization targets, process-node milestones).
• Ha-Joon Chang brought this tradition into debates about globalization, arguing that today’s rich countries used aggressive industrial policies before later pushing “free-market” rules on latecomers. China’s semiconductor strategy—protecting and promoting domestic champions while acquiring foreign technology—is consistent with this “infant industry” logic, applied to the most complex manufacturing sector on earth.

These theorists provide the conceptual lens for understanding why China’s catch-up was plausible despite skepticism: latecomer states, given enough capital, policy focus, and market size, can leap across technological stages faster than many linear forecasts assume.

2. National innovation systems and technology policy

Key figures: Christopher Freeman, Bengt-Åke Lundvall, Richard Nelson, Mariana Mazzucato

• Christopher Freeman and Bengt-Åke Lundvall developed the idea of national innovation systems: webs of firms, universities, government agencies, and financial institutions that co-evolve to generate and diffuse innovation. China’s semiconductor rise reflects a deliberate effort to construct such a system around chips, combining universities, state labs, SOEs, private giants (like Alibaba and Huawei), and policy banks.
• Richard Nelson emphasized how governments shape technological trajectories through defense spending, procurement, and research funding. U.S. policies around semiconductors and AI mirrors this; China’s own national funds and state procurement echo similar mechanisms, but at enormous scale.
• Mariana Mazzucato introduced the idea of the “entrepreneurial state”, arguing that the public sector often takes the riskiest, most uncertain bets in breakthrough technologies. China’s massive and politically risky bets on semiconductor self-reliance—despite early policy failures and wasted capital—are a stark, real-time illustration of this concept.

These frameworks show why China’s chip gains are not just about firm-level success, but about system-level design: how policy, finance, and research infrastructure have been orchestrated to accelerate domestic capability.

3. Global value chains and “smile curves”

Key figures: Gary Gereffi, Timothy Sturgeon, Michael Porter

• Gary Gereffi and Timothy Sturgeon analyzed how industries fragment into global value chains, with design, manufacturing, and services allocated across countries according to capabilities and policy regimes. Semiconductors are the archetype: U.S. firms dominate GPUs and EDA tools; Taiwanese and Korean firms dominate advanced wafer fabrication and memory; Dutch and Japanese firms produce critical tools; Chinese firms historically concentrated on assembly, packaging, and lower-end fabrication.
• In this framework, export controls and industrial policies are attempts to reshape where in the chain China sits—from lower-value segments toward high-value design, advanced fabrication, and toolmaking.²
• The “smile curve” metaphor (popularized by Acer’s Stan Shih and linked to strategy thinkers like Michael Porter) suggests that value accrues at the edges: upstream in R&D and design, and downstream in brands, platforms, and services. For years, China captured more value in downstream device assembly and domestic platforms; Sheehan’s quote highlights China’s effort to climb the upstream side of the smile curve into high-value chip design and equipment.

4. Technology, geopolitics, and “weaponized interdependence”

Key figures: Henry Farrell, Abraham Newman, Michael Beckley, Graham Allison

• Henry Farrell and Abraham Newman advanced the concept of “weaponized interdependence”: states that control key hubs in global networks—financial, digital, or industrial—can use that position for coercive leverage. U.S. control over advanced lithography, chip design IP, and high-end AI hardware is one of the clearest real-world illustrations of this idea.
• The use of export controls and entity lists against Chinese tech firms is an application of this theory; China’s accelerated semiconductor localization is, in turn, a strategy to escape vulnerability to that leverage.
• Analysts such as Michael Beckley and Graham Allison focus on U.S.–China strategic competition, emphasizing how control of technologies like semiconductors shapes long-term power balances. For them, the pace of China’s chip catch-up is a central variable in the evolving balance of power.

Sheehan’s quote sits squarely in this intellectual conversation: it is an empirical judgment that bears directly on theories about whether technological chokepoints are sustainable and how quickly a targeted great power can adjust.

5. AI, compute, and the geopolitics of chips

Key figures: Jack Clark, Allan Dafoe, Daron Acemoglu, Ajay Agrawal

• Researchers of AI governance and economics increasingly treat compute and semiconductors as the strategic bottleneck for AI progress. Analysts like Jack Clark have emphasized how access to advanced accelerators shapes which countries can realistically train frontier models.
• Economists such as Daron Acemoglu and Ajay Agrawal highlight how AI and automation interact with productivity, inequality, and industrial structure. In China, AI and chips are now deeply intertwined: domestic AI labs both depend on and stimulate demand for advanced chips; chips, in turn, are justified politically as enablers of AI and digital sovereignty.^2,5
• The result is a feedback loop: AI breakthroughs (such as those highlighted by Xi Jinping in 2025) strengthen the case for aggressive semiconductor policy; semiconductor gains then enable more ambitious AI projects.⁵

This body of work provides the conceptual scaffolding for understanding why a statement about Chinese chip catch-up is not just about manufacturing, but about the future distribution of AI capability, economic power, and geopolitical influence.

Placed against this backdrop, Matt Sheehan’s line is more than a passing compliment to Chinese engineers. It crystallizes a broader reality: in one of the world’s most complex, capital-intensive, and tightly controlled industries, China has closed more of the gap, more quickly, under more adverse conditions than most experts anticipated. That surprise is now reshaping policy debates in Washington, Brussels, Tokyo, Seoul, and Taipei—and forcing a re-examination of many long-held assumptions about how fast latecomers can move at the technological frontier.

References

1. https://www.scmp.com/tech/big-tech/article/3339366/great-chip-leap-chinas-semiconductor-equipment-self-reliance-surges-past-targets

2. https://www.techinsights.com/chinese-semiconductor-developments

3. https://www.tomshardware.com/tech-industry/china-expected-to-approve-h200-imports-in-early-2026-report-claims-tech-giants-alibaba-and-bytedance-reportedly-ready-to-order-over-200-000-nvidia-chips-each-if-green-lit-by-beijing

4. https://eu.36kr.com/en/p/3634463429494016

5. https://dig.watch/updates/china-ai-breakthroughs-xi-jinping

6. https://expertnetworkcalls.com/93/semiconductor-market-outlook-key-trends-and-challenges-in-2026

7. https://sourceability.com/post/whats-ahead-in-2026-for-the-semiconductor-industry

8. https://www.pwc.com/gx/en/industries/technology/pwc-semiconductor-and-beyond-2026-full-report.pdf