This keynote address was delivered by Dr. Syaru Shirley Lin, CAPRI Founder and Chair, at the Morgan Stanley Asia AI Summit in Taipei on May 28, 2026.
Introduction
I am happy to be back at Morgan Stanley, speaking for the first time in my hometown. Since leaving the venture capital and private equity world, I’ve transitioned from identifying and investing in start-ups to creating my very own. At the World Economic Forum exactly 4 years ago, we announced the founding of the Center for Asia-Pacific Resilience and Innovation (CAPRI) as Taiwan’s first nonprofit, nongovernmental think tank conducting comparative public policy research across the Asia-Pacific. The think tank’s mission is to draw on the Asia Pacific experience to make the world more resilient and innovative.
A good example of our work is this: how do we replicate the Hsinchu industrial park in Arizona, including land acquisition, talent circulation, and tax policy design? We are supported by Taiwan’s leading semiconductor and financial companies and have started work with companies in Korea and Japan. At CAPRI, we carry out research across domains deeply relevant to the work you do as investors.
I do not need to tell you that AI is going to change our world. Yet we seem to live in two entirely different worlds. As people worldwide are concerned with the fracturing world order and cascading global conflicts, the world of AI investment and hardware manufacturing is charging full speed ahead, especially on this island. In the US, the five largest hyperscalers have committed about $690 billion in capital expenditures in 2026 alone, roughly $450 billion of it directly tied to AI infrastructure. This is the largest private technology investment cycle in history. By 2035, the AI industry market size itself is expected to exceed $5 trillion.
Being here in Taipei today, you probably know that Taiwan’s trade with the United States surged nearly 80% in 2025, and in Q1 2026, the US displaced China as Taiwan’s top trading partner for the first time in 25 years. Taiwan is now America’s fifth largest trading partner, trailing behind China, which fell to number four as US–China goods trade fell by 29%. The supply chain pivot and the security pivot are now pointing in exactly the same direction: in favor of Taiwan.
I think about how quickly this landscape has shifted when I consider companies like Alibaba, which Jack Ma presented to me in Hangzhou in 1999 as just an e-commerce platform. Today, Alibaba is one of the world’s leading AI companies, committing $53 billion to AI and cloud infrastructure over the next 3 years. This transformation from e-commerce start-up to global AI frontier captures what is happening across Asian technology broadly. The companies that built the internet economy in Asia are now building the AI economy.
While today the AI boom is taking root in Asia, this remarkable growth risks being overshadowed by great power competition. And so today I want to address what many investors quietly ask: Should we still be confident investing in Asia and in Taiwan? My answer is yes—not because the risks do not exist, but because they are often misunderstood, and many are manageable. I will start with three top risks, namely US–China technology rivalry, energy, and cross-Strait relations.
AI Competition Is About US–China Technology Rivalry
Artificial intelligence is no longer primarily a software phenomenon. It has become a physical system dependent on semiconductors, advanced manufacturing, electricity, cooling infrastructure, logistics networks, and industrial coordination at massive scale. AI is fundamentally constrained by infrastructure, not just algorithms. Once technology becomes infrastructure-dependent, it becomes geographically concentrated and geopolitical. This is why the intensifying strategic competition between the United States and China matters so deeply for investors. This competition is not simply about trade or tariffs. It is a structural contest over technological leadership, industrial capacity, military advantage, and ultimately the rules that will govern the global digital economy.
The most visible expression is US export controls on advanced semiconductors, which began in October 2022 and were significantly tightened in 2023, designed to maintain as large a technological lead as possible in foundational computing capabilities. As my book, Taiwan’s China Dilemma, shows, restrictions have long existed to keep China two generations behind, which the US and Taiwan have adhered to. However, while regulations kept China behind, now they are shutting the Chinese out completely.
It is worth being precise about what this strategic competition means in practice because the United States simultaneously holds two positions that are in genuine tension. Washington is deeply skeptical of China’s technological trajectory, resting on the premise that military–civil fusion means advanced commercial technology will be redirected toward military and surveillance applications, which Professor Wei Shyy, member of CAPRI’s International Advisory Council, has written extensively on. But US–China economic affinity remains structurally deep: bilateral goods trade exceeded $580 billion in 2024, and while it dropped by nearly 30% last year, Trump has indicated that he is seeking an economic detente and wishes to revive the US–China trading relations. Moreover, US capital markets have historically been a primary source of financing for Chinese technology companies.
This is not a contradiction that policy can easily resolve, but a permanent structural feature of the relationship. The US will continue to separate critical and strategic industries from China while seeking bilateral trade deals on items like soybeans, Boeing airplanes, and beef. The most recent Trump–Xi summit is emblematic of this tightrope: the two leaders agreed to create a Board of Trade and Board of Investment, intending to set a pathway for future trade. Yet, just before the summit, the US accused China of “distilling” or copying US AI technology, which China has obviously denied.
CAPRI Board Member Professor Harry Harding, who has advised five US administrations on US–China relations, spoke last year with Beijing’s government advisors at a conference hosted by CAPRI USA Foundation and Fudan University in Shanghai on the prospects of US–China relations. Professor Harding pointed out that the central challenge for the second Trump administration is not choosing between engagement and decoupling but constructing a framework that allows both competition and selective cooperation to coexist. The goal is not to reach a detente but to engage in a structured rivalry that will define US–China relations in the decades to come. At that time, consistent with past rhetoric, the Chinese advisor pushed back and re-emphasized “win-win cooperation.”
However, this month at the summit, Xi Jinping made it clear that “hiding its brightness and biding its time” (韜光養晦) is over for China. Xi Jinping put forth his “constructive strategic stability” doctrine, a framework Beijing pushed for more than a decade. The concept views Washington and Beijing as equal powers and Xi as able to define the rules of engagement. This has important ramifications for Taiwan, AI, export controls, and Iran.
We will see how Trump moves forward on Taiwan’s $14 billion arms sales package—especially its timing—and whether it is before the next Trump–Xi meeting on September 24. Trump does not want Xi to cancel the meeting and may give Xi this concession and postpone the arms sales. What Xi wants, however, is for Trump to ideally postpone the sales indefinitely, but at least until Taiwan’s next national elections in January 2028. Xi must signal to his domestic audience that he can manage US behavior.
So, while we may get future summits, hotlines, and targeted commercial deliverables on agriculture, energy, and maybe even some AI governance dialogue, the structural competition in semiconductors, AI, and critical minerals will continue. Investors who treat US policy as purely adversarial toward China—or purely accommodating—will consistently misread the signals.
China remains a major market in semiconductors. In 2024, China and Hong Kong absorbed 52% of Taiwan’s total semiconductor exports, roughly $85 billion out of $165 billion. South Korea’s semiconductor exposure to China has historically been in the 40%–45% range. These are not dependencies that unwind quickly, and China is not a passive participant. Professor Minxin Pei, renowned China expert who just published The Broken China Dream, argued this week at a CAPRI Conversation that Beijing is driven by insecurity and fear surrounding containment by other countries. Professor Pei says Xi believes that China and the US are already in the Thucydides Trap, which is why he raised it at the summit. This worldview has set in motion a long-term strategy to insulate China from both real and imagined threats from the outside.
We see this taking shape in the way China has pursued a self-reliant economy and is actively working towards its own complete AI infrastructure ecosystem. In May 2024, Beijing launched Big Fund III at $47.5 billion in state-backed semiconductor investment, the largest fund in Chinese history, signaling that chip self-sufficiency is a generational industrial strategy and not a short-term gesture. Most recently, just before the Trump–Xi summit, Beijing blocked Meta’s acquisition of Manus, a Chinese-founded AI start-up based in Singapore. The result is a gradual movement toward parallel technology ecosystems, where China and the United States may increasingly operate with different supply chains, different standards, and potentially different governance models for AI.
Decoupling presents separate opportunities for investors, especially for those who understand the two ecosystems well. A highly consequential dimension of this shift is the slowing movement of talent across borders. Visa tightening, export control compliance, and geopolitical mistrust are collectively reducing the ease with which engineers and researchers circulate, in both directions.
As a faculty member of the University of Virginia, I see this drastic change, since I have traditionally supervised many Chinese graduate students. The number of international scholars in the United States is down 18% against 2018 levels, and the overall student visa denial rate hit a record 41% in 2024. The consequences compound at the frontier: foreign-born professionals account for 43% of PhD-level scientists and engineers in the US, and a 2025 analysis of 800,000 AI workers across 115 countries found the US is losing its position as the default destination for top global AI talent.
China faces a different but equally structural constraint: despite now surpassing the US as the world’s largest producer of STEM doctoral graduates, it faces a projected AI talent shortfall of four million by 2030, driven not by foreign student flows but by a domestic quality and distribution gap that scale alone cannot close. The talent that is being displaced is not disappearing: it is being redirected to Malaysia, Thailand, and elsewhere in Asia. For the United States, this represents an erosion of the open talent system that has underpinned its technological leadership since 1945. For Asia, it is an emerging opportunity to build research depth and graduate ecosystems that did not previously exist. Over time, this friction in human capital circulation may prove as consequential as restrictions on physical technologies.
Despite Fragmentation, Asia Remains Indispensable
Despite these pressures, Asia remains structurally indispensable to the AI economy, and Taiwan sits at the center of that structure, not simply as a question of market share but of industrial density and accumulated expertise. Taiwan dominates advanced logic manufacturing, Japan supplies upstream materials and semiconductor equipment, and South Korea leads global memory. To put a number on Taiwan’s position, TSMC alone produces over 90% of the world’s advanced logic chips below 7 nanometers. There is no near-term substitute for that concentration of capability.
How long will this primacy last? TSMC’s own guidance projects that its fabs in Taiwan will continue to be the leading manufacturers for at least this decade, and likely well into the next. Another layer of the ecosystem that is important is chip design—the fabless companies that architect and engineer semiconductors. Taiwan has important fabless designers competing globally across consumer electronics, communications infrastructure, and increasingly AI applications.
The investment community is beginning to recognize this broader picture. Taiwan is the only geography outside the United States that has built a complete, end-to-end AI industrial ecosystem spanning chip design, advanced manufacturing, advanced packaging, AI system assembly, and an emerging software layer. No other geography has all five at scale.
So, Can the US Recreate Asia’s Semiconductor Ecosystems?
Discussions around reshoring and diversification must be approached with careful realism. The CHIPS and Science Act (enacted on August 9, 2022) has mobilized substantial investment to rebuild domestic semiconductor capacity. But it highlights the distinction between building physical infrastructure and recreating entire ecosystems.
I know this from direct experience. As a Goldman Sachs partner, I led the firm’s $1.6 billion first institutional funding round in 2000 for SMIC, China’s largest foundry, because I always thought China would need its own national champion. It is today a strategic national asset, but it took far longer than we envisioned and is still a work in progress. That experience highlights how capital is not the binding constraint. Building the human systems, the supplier density, and the institutional knowledge that make a fab competitive takes decades, not quarters.
TSMC’s competitive advantage is not primarily the machines. It is the human capital: thousands of highly skilled engineers and technicians embedded in a supplier ecosystem, sharing institutional knowledge that is tacit and extraordinarily hard to transfer. Hsinchu’s ecosystem took four decades to build in a specific cultural and economic context.
For the past year, CAPRI has been organizing delegations from Arizona State University and the University of Arizona to work with Taiwanese universities, companies, and legislators on talent circulation between Taiwan and the US. Together, we have discussed plans to develop talent pipelines and the importance of public–private partnerships in confronting workforce constraints. The talent challenges are enormous, so R&D remains critical in Taiwan. I would describe Arizona as a strategically valuable but structurally distinct complement to Taiwan—not a replica, but a meaningful node in a more geographically distributed network. The Japan fab, by contrast, is tracking better in part because the cultural and institutional alignment is closer.
Even with TSMC’s US investment, now totaling $65 billion across three planned fabs, Arizona is likely to represent roughly 10%–15% of global advanced logic output, enough to matter for US defense and national security applications and to provide a hedge against single-point geographic risk, but not nearly enough to replace Taiwan’s role. That is by design, and investors should evaluate Arizona on those terms, not as a substitute for Taiwan. Onshoring will not fully replicate the efficiency or density of existing Asian ecosystems in the near term. Investors should avoid conflating political ambition with industrial substitution.
Energy System Risk
Energy is the limiting condition for AI infrastructure. You probably heard that 2 days ago, Jensen Huang publicly called for more electricity in Taiwan: “Without energy, there is no economic growth.” This is exactly what CAPRI’s work is addressing: industrial policy and energy system planning must go hand in hand.
Taiwan sits at the center of both the demand and the vulnerability. Taiwan imports over 95% of its total energy. Today’s electricity mix is roughly 48% gas, 35% coal, and only about 15% clean energy, well short of the 20% renewables target the government set for 2025. LNG fuels more than half the grid, and Taiwan holds only 11 days of natural gas reserves, a dangerously thin cushion for an island running the world’s most advanced fabs. This vulnerability has a geopolitical dimension that is easy to overlook. Qatar alone supplies roughly one-third of Taiwan’s LNG, meaning that any prolonged disruption at the Strait of Hormuz would not merely raise costs: it could directly threaten production continuity at the world’s most critical semiconductor facilities.
At CAPRI, we are leading multiyear research in collaboration with national and international experts on energy resilience and the green transition across the Asia-Pacific. We look into the competing priorities within energy policy, as everyone wants energy that is cheap, secure, and sustainable. Attempting to meet this demand is particularly critical in Asia, where the bulk of the energy-intense semiconductor industry is housed and data centers are being built rapidly.
I want to be precise about what a “manageable” energy mix means here because it requires real policy choices, not wishful thinking. Taiwan has enacted a legally binding net-zero target for 2050, with an ambition for 60%–70% of electricity to come from renewables by mid-century. The nuclear policy debate, which was politically off limits for years, is genuinely reopening. LNG supply is being actively diversified away from Qatar toward Australia and the United States. The window for getting the energy policy design right is narrowing at exactly the moment AI infrastructure energy demands are accelerating most rapidly.
Are Cross-Strait Tensions the Biggest Risk?
The possibility of large-scale military conflict involving Taiwan is often discussed in binary terms, but the reality is more complex. For anyone who understands history, Taiwan’s centrality in US–China relations since 1949 predates semiconductors entirely—chips have only intensified what was already a foundational dispute. Taiwan mattered to Zhou Enlai and Mao Zedong in 1971 and 1972 when they met with Kissinger and Nixon, and it matters no less today. To assume that once the reshoring of semiconductor production is complete, Taiwan will recede as a strategic issue is misguided.
Realist scholars have long argued that Taiwan’s importance in military and geopolitical terms makes it an enduring imperative—too strategically located and too symbolically loaded to be resolved by supply chain adjustments alone. For Xi Jinping, Taiwan is explicitly framed as the centerpiece of China’s national rejuvenation, not a peripheral aspiration but a declared civilizational mission. Yet for all the rhetoric, this is one risk Xi will not take lightly. The costs of miscalculation in military, economic, and political terms would be of a magnitude that no plenum, no congress, and no succession timeline could survive intact.
A full-scale invasion would carry extremely high costs for any actor involved, including global economic disruption, financial sanctions, and profound instability in global technology supply chains. To put the economic stakes in context: even a 1-year disruption to Taiwan’s semiconductor production would cost global electronics, automotive, AI, and defense industries trillions of dollars. No adversary can be indifferent to that calculus.
The more immediate and persistent challenge is gray-zone pressure, such as military exercises, airspace incursions, cyber operations, and maritime signaling. For investors, gray-zone dynamics generate persistent uncertainty rather than discrete shocks: they affect shipping insurance premiums, risk assessments, and capital allocation decisions even in the absence of material disruption. The most important point is that uncertainty itself has become a structural feature of Taiwan’s operating environment. Despite this, Taiwan has continued to function as one of the most critical nodes in the global technology supply chain.
An Additional Emerging Risk: Governance Fragmentation
As the frontier technology race heads into two parallel systems, there is an emerging governance challenge. Companies and investors must understand how to navigate differing compliance regulations and technical standards. US export controls are tightening, expanding in scope, and increasingly extraterritorial in reach, requiring companies to enforce regulations that are technically and legally complex. Europe’s AI Act has added a second layer of complexity, creating regulations that companies must comply with for their technology to be deployed by European customers. China’s own technical standards, ranging from data localization to AI model certification, are creating a separate compliance environment for any company maintaining meaningful China operations.
These three competing environments are creating a compliance trilemma, pulling stakeholders in incompatible directions and forcing companies to make architectural choices that preference one environment over another. Managing this trilemma is becoming one of the defining operational challenges for the most sophisticated players in the AI supply chain. Companies that build governance agility into their operating model will carry a durable advantage over those that treat compliance as an afterthought.
A second dimension of the governance challenge is the sovereign AI paradox, which directly impacts Taiwan’s strategic position. Countries are now increasingly concerned over the power that the US or China may hold over AI governance, data, and cloud infrastructure, leading to an “AI sovereignty” movement. This refers to the desire of smaller states to build their own large language models, national cloud infrastructure, and AI governance frameworks. But countries pursuing AI sovereignty almost inevitably end up in long-term contracts with US hyperscalers and chip suppliers as the alternative supply chain does not exist yet.
These governance challenges highlight that there is a collision of technology policy, economic competitiveness, and regulatory frameworks that needs to be addressed. Most semiconductor analysis will treat these elements as separate problems. At CAPRI, we confront this policy challenge head on and acknowledge it must be addressed through an interdisciplinary lens. Our upcoming trilateral conference in Tokyo in collaboration with German partners will focus on AI resilience to bring together diverse stakeholders from Asia, Europe, and the US—including leading companies like SK Group, Microsoft, and Google—to discuss how AI governance and innovation need to go hand in hand. The three regions play distinct roles in the creation and regulation of AI, necessitating a dialogue platform to bring these perspectives together. As the AI ecosystem matures and evolves, our trilateral meeting underscores that sharing regional and cross-sector perspectives is integral to creating a resilient system that enables investors, governments, and users to effectively harness the power of AI.
Closing Thoughts
Taiwan is often framed through the lens of risk. But from an infrastructure perspective, Taiwan is a case study in resilience, showcasing a system that has operated under sustained geopolitical pressure while maintaining extraordinary industrial sophistication and global technological relevance. The future of AI infrastructure will emerge in environments capable of managing geopolitical tension while continuing to function as nodes of production, innovation, and capital deployment.
One thing I’d like you to take away is this: the United States’ asks of Taiwan are increasingly in tension with each other. Washington wants Taiwan to supply America with the most advanced chip maker, to restrict technology exports to China, to build fabs in Arizona, and to buy more weapons and maintain deterrence all simultaneously. These are not always compatible objectives. And Taiwan’s response to those demands is not made in a vacuum between Washington and Beijing, it is made by a democratic government accountable to Taiwanese voters who care about wages, jobs, and energy prices, not only geopolitics. The US faces the same challenge: voters want better jobs, lower prices, and to minimize spending on defending the world. Managing competing pressures will require more political sophistication on all sides than we have seen so far.
For investors, the implication is not to avoid risk, but to understand it. The structural risks I would describe as manageable, as distinct from catastrophic risks, include energy vulnerability, workforce constraints, technology supply chain concentration, and cross-Strait gray-zone pressure. These are real risks, but they are not, in and of themselves, existential ones. The catastrophic scenarios, whether it is a sustained blockade or direct military conflict, carry deterrence costs for any actor that make them less probable than headlines suggest, though not impossible.
It is precisely this intersection of technology, governance, and economic resilience that CAPRI was founded to study, because the investors who will price this region correctly are those who treat geopolitical complexity not as noise to be discounted, but as signal to be understood. Having watched this industry for over thirty years, from investing in Chinese technology companies to teaching geoeconomics at the University of Virginia and CUHK, what strikes me most is how consistently the outside world has underestimated Taiwan. Not just its technical capacity, but the depth and density of what has been quietly built here. For investors willing to do the work to correct it, that gap is where the opportunity lives.
Taiwan is not simply the world’s chip foundry. It is an emerging full-stack AI ecosystem, with record investment, global top ten rankings in AI infrastructure, and a rapidly expanding presence in robotics, AI agents, intelligent healthcare, and smart manufacturing. Jensen Huang just called Taiwan the “epicenter of the AI revolution,” with plans to spend $150 billion a year on this island. The narrative of Taiwan as a single-product geography has always been incomplete. It is becoming untenable.
