The "TAIWAN IS SAFE" MYTH — The Economic Numbers Nobody Wants You to See

Ajouté : 2026-05-28

[00:00:00]In 2022, the United States passed the CHIPS and Science Act, committing $52 billion to domestic semiconductor manufacturing.

[00:00:09]The stated rationale, drawn directly from the bill's finding section, was reducing dangerous dependence on a single geographic region for advanced chip production. That region was Taiwan.

[00:00:22]The same year, TSMC, Taiwan Semiconductor Manufacturing Company, broke ground on a fabrication plant in Arizona. The announcement was treated as a rebalancing, a managed redistribution of irreplaceable industrial capacity away from a flash point and towards stable ground. By 2024, TSMC's Taiwan operations still accounted for roughly 90% of the world's most advanced semiconductors. The chips below 3 nanometers that run AI systems, defense electronics, and the logic cores of virtually every high-end consumer device manufactured anywhere. The Arizona plant, when fully operational, was projected to handle a fraction of that volume at higher cost with a workforce that the company's own executives described in earnings calls as still being developed. The CHIPS Act was designed to reduce concentration. The concentration did not reduce. That gap between the policy intention and the documented outcome is not a failure of implementation. It is the subject of this video. Taiwan's economic relationship with China runs in a different direction than the security conversation suggests. China has been Taiwan's largest trading partner for over two decades. In 2023, cross-strait trade totaled approximately $158 billion.

[00:01:43]Taiwanese firms have maintained manufacturing operations across the mainland through periods of serious diplomatic tension, through the 1996 missile crisis, through Chen Shui-bian's presidency, through every iteration of the Taiwan Relations Act, the investment did not stop. It deepened. So, here are the two facts that will carry everything that follows. Western governments have formally identified Taiwan as a critical node in their own economic security and have committed public capital to reducing that dependency.

[00:02:16]At the same time, the dependency has not reduced. It is compounded because the technology has advanced faster than the rebalancing policy and because the companies doing the most advanced manufacturing have no short-term incentive to relocate capacity that is already profitable and operational.

[00:02:36]Those two facts are both on the record.

[00:02:39]They are not in dispute. What is in dispute, what nobody in the official commentary resolves, is what it means to call Taiwan safe when the economic architecture treating it as irreplaceable is the same architecture that makes the consequences of instability incalculable.

[00:02:57]The military deterrence argument and the supply chain argument are being run in parallel by the same governments and the same institutions. They are not the same argument. One says Taiwan status quo is defensible. The other says the status quo is so embedded that disruption would be catastrophic. A thing cannot be both defended and indispensable without the defense itself becoming a form of economic hostage taking. And the documents, read carefully, show that several governments have understood this for longer than the public conversation suggests. The CHIPS Act finding section used the word dangerous. That word was in the official record before the investment strategy changed anything. It remains accurate after. The mechanism that allows both facts to coexist has a name and it appears in the corporate filings more honestly than it does in the policy documents. It is called geographic concentration risk, and for the better part of a decade, it was disclosed, acknowledged, and then absorbed into the cost of doing business. TSMC's annual reports have carried geographic concentration language since at least 2018. The 2021 report, filed before the Chips Act passed, described Taiwan as subject to political and military risks that could materially affect operations. That sentence was in the risk disclosure section. It was not in the strategic outlook section. The distinction matters. Risk disclosure is what companies write when they need to inform investors without alarming customers. It is the institutional equivalent of knowing something and choosing where to file it. The customers were not alarmed.

[00:04:39]Apple, Nvidia, AMD, and Qualcomm, the four companies that collectively account for the majority of TSMC's advanced node revenue, continued to deepen their design dependency on TSMC's most Taiwan-concentrated processors through the same period the disclosures were being written.

[00:04:58]Nvidia's H100, the chip that became the infrastructure of the global AI build-out, was fabbed at TSMC's Taiwan facilities. There was no alternative.

[00:05:08]The process node it required did not exist anywhere else.

[00:05:12]This is the mechanism. The customer dependency creates a revenue base that funds the capital expenditure that extends the technology lead that makes the customer dependency more difficult to unwind. It is not a conspiracy. It is a compounding loop, and it was visible in the filings for anyone who read past the headline numbers.

[00:05:31]The policy layer added its own structure on top of this. The 2022 Chips Act was not the first attempt to address semiconductor concentration. The 2021 executive order on America's supply chains, signed in February of that year, identified semiconductors as a critical gap and called for a 100-day review.

[00:05:52]That review, published by the Department of Commerce in June 2021, concluded that the United States had lost its manufacturing edge through a combination of market incentives that favored offshore production and insufficient public investment in domestic capacity.

[00:06:08]The review was correct. It was also describing a condition that had been developing for 30 years. The 30-year timeline is worth holding. Taiwan's semiconductor dominance did not emerge from a single policy decision or a single corporate strategy. It was the cumulative result of sustained public investment through the Industrial Technology Research Institute beginning in the 1970s, deliberate talent development, and critically, a government willing to treat semiconductor manufacturing as strategic infrastructure rather than a market outcome. The United States during the same period treated it as a market outcome. The market sent it to Taiwan.

[00:06:50]What the CHIPS Act was attempting in 2022 was to reverse 30 years of compounding advantage in a 5-to-10-year window.

[00:06:59]The Commerce Department's own implementation documents projected that new domestic fabs would reach meaningful advanced node production no earlier than 2030. TSMC's next-generation process node was already in risk production in Taiwan by 2023.

[00:07:17]The irony, and it is a specific institutional irony, not a general one, is that the CHIPS Act's most significant near-term effect was to fund TSMC's Arizona expansion.

[00:07:29]The primary vehicle for reducing dependency on a Taiwanese company was a capital subsidy to that same company to build in America.

[00:07:38]The subsidy did not transfer the technology. It transferred one facility.

[00:07:42]The engineers, the process knowledge, the supply chain for the chemicals and equipment the fabs require, those remain concentrated where they have always been. Intel received Chips Act funding as well, targeting its domestic expansion. Intel's own filings through 2023 and 2024 documented repeated delays in its advanced node development. Delays significant enough that several customers it was meant to recapture had already redesigned products around TSMC's roadmap rather than wait. The mechanism then is this. The policy acknowledged the risk, funded a partial response, and the partial response was absorbed by the existing concentration rather than displacing it. Both facts remain true. The dependency was identified as dangerous, the dependency deepened anyway. What allows both to be true is that the institutions involved had different time horizons, and the market's time horizon was shorter than the risks. The conventional explanation for why this is manageable runs as follows.

[00:08:44]Taiwan has survived decades of cross-strait tension without economic disruption.

[00:08:49]The deterrence architecture, the United States security commitment, the Taiwan Relations Act, the presence of the Seventh Fleet, has held.

[00:08:57]Markets have priced the risk and found it acceptable.

[00:09:01]And the economic interdependence between Taiwan and China is itself a stabilizing force.

[00:09:07]Beijing's economy is too connected to global supply chains to absorb the disruption that a move against Taiwan would trigger.

[00:09:14]Mutual dependence, the argument goes, is its own form of deterrence.

[00:09:19]This argument is not wrong. It is incomplete. And the place where it stops short is visible in the primary record.

[00:09:27]The mutual dependence argument rests on the assumption that Beijing's calculus is primarily economic.

[00:09:34]The documents that complicate that assumption are not speculative. They are the Chinese government's own published policy texts. The 2005 anti-secession law passed by the National People's Congress established a legal framework obligating the state to employ non-peaceful means under specific conditions. Those conditions are defined broadly enough to include a declaration of independence or and this is the language that the commentary routinely alides, a determination that possibilities for peaceful reunification have been exhausted. That determination is made by Beijing on Beijing's timeline using criteria Beijing does not publish.

[00:10:14]The economic interdependence argument does not engage with that clause. It assumes the decision-making framework is cost-benefit. The legal framework Beijing has constructed is not purely cost-benefit. It contains a political tripwire that economic modeling cannot price. The second place the conventional explanation stops short is in its treatment of TSMC itself. The deterrence argument implies that the fab infrastructure would remain operational and accessible through a period of serious military tension. The operational reality is different. TSMC's advanced fabs require continuous inputs, ultra-pure water, specialty gases, photolithography equipment manufactured almost exclusively by ASML in the Netherlands, and a highly specialized workforce. A blockade that did not involve a single shot fired would interrupt those inputs within weeks. The fabs do not stockpile. The processes do not pause cleanly.

[00:11:11]Morris Chang, TSMC's founder, said in 2022 that in the event of a military conflict, TSMC would become unoperational. That is not an analyst projection. That is the assessment of the person who built the company on the record in a public forum. The deterrence framework had no answer for it because the deterrence framework is about preventing conflict, not about what the asset looks like during one. The detail worth returning to, the one introduced early and now carrying more weight, is the word the CHIPS Act used, dangerous.

[00:11:45]Not concentrated, not sub-optimal, dangerous.

[00:11:49]That word in a legislative finding section has a specific meaning.

[00:11:54]It means Congress assessed the current condition as posing a threat to national security or economic welfare, serious enough to justify emergency public expenditure.

[00:12:05]The $52 billion was the price of that assessment. What the record tells me is that the dangerous condition the CHIPS Act was responding to has not been resolved by the CHIPS Act, and the institutions that passed it understood at the time of passage that it would not be resolved within any politically relevant time horizon.

[00:12:24]The bill was not a solution, it was a signal directed at allies, at markets, and at Beijing that the United States recognized the vulnerability.

[00:12:33]Recognizing a vulnerability and closing it are not the same action. The documents make that distinction. The public [clears throat] summary of the policy did not. The strongest version of the conventional explanation also points to economic integration as a brake on escalation from the Taiwanese side. The argument that Taiwanese business interests, deeply embedded on the mainland, create a domestic political constituency for stability. This is historically accurate. The cross-strait business community has functioned as an informal back channel through multiple periods of tension. What changed is the technology export control regime.

[00:13:12]Beginning in October 2022, the United States imposed restrictions on the export of advanced semiconductor equipment and chips to China.

[00:13:21]Restrictions subsequently tightened in 2023. Those controls targeted the exact node of the supply chain where Taiwan sits. They were designed explicitly to prevent China from acquiring the technology needed to close the gap with TSMC. They also restructured the incentive calculation for Taiwanese firms operating on the mainland because the equipment those firms use now falls under American licensing jurisdiction regardless of where the factory is located. The informal stabilizing force, cross-strait business integration, was partially dismantled by an American policy decision. The conventional explanation has not updated to account for that. The tension from part one resolves this way. Western governments designated Taiwan as critical infrastructure, dangerous to depend on, irreplaceable in practice because the economic architecture they built over 30 years made it so.

[00:14:17]The safety argument and the dependency argument are not in contradiction. They are the same argument running in opposite directions at the same time.

[00:14:25]Taiwan is called safe because the alternative conclusion, that the world's most advanced semiconductor capacity sits on contested territory with no adequate backup, and a legal tripwire in the controlling power's domestic law, is a conclusion that markets, governments, and corporate planning cycles cannot operationalize.

[00:14:46]So, it is not stated. The documents state it, the summaries do not. The export control architecture introduced between 2022 and 2023 is where the resolution becomes concrete. The October 2022 Bureau of Industry and Security Rules, the most expansive unilateral technology export restriction the United States had imposed since the Cold War by the Commerce Department's own characterization, were designed to freeze China's semiconductor development at its current level. The logic was containment. If China cannot close the gap with TSMC, the leverage that Taiwan's manufacturing represents remains intact, and the cost of disrupting it remains prohibitive for Beijing. That logic has a structural problem.

[00:15:32]Containment strategies that depend on a single point of failure do not eliminate the risk. They concentrate it.

[00:15:40]Every additional year that TSMC's lead is preserved by export controls, rather than redistributed through genuine manufacturing diversification, is a year in which the gap between what the policy says and what the architecture requires grows wider. The CHIPS Act was meant to redistribute. The export controls were meant to contain.

[00:16:00]They are pulling in different directions, and the documents from both the Commerce Department and the National Security Council, the latter partially declassified in the 2022 National Security Strategy, treat them as complementary. They are not complementary. One reduces concentration. The other makes the concentrated asset more strategically valuable, which makes it a more attractive target, which increases the risk the first policy was designed to address. The Taiwanese government understands this geometry precisely. The Economic Development Commission's published industrial strategy documents describe TSMC's global centrality as a form of what they call the silicon shield. The idea that being indispensable to the global economy is itself a deterrent. That phrase entered official Taiwanese policy language because it is accurate as a description and useful as a posture. It is also a description of a hostage situation in which Taiwan is simultaneously the hostage and the negotiator. The forward-looking statement that the record supports is specific. The first genuine test of whether the CHIPS Act's diversification logic is working will not be a geopolitical event. It will be a technology event, the moment at which TSMC's two-nanometer process and its successors either do or do not get replicated at meaningful volume outside Taiwan. The Commerce Department's own implementation milestones, published in the CHIPS Program Office guidance documents, set 2030 as the earliest plausible date for that replication at scale. If that date slips, and the Intel filing history suggests institutional optimism about fab timelines should be discounted, the window in which the silicon shield argument remains credible extends accordingly. Beijing's planners can read those filings. They are public documents. My honest read on this, having worked through the legislative record, the corporate disclosures, the export control architecture, and the cross-strait trade data, is that the safety consensus around Taiwan is not a considered position. It is a necessary fiction maintained by institutions that have no good alternative to offer. The deterrence framework is real. The economic integration break is real. The silicon shield logic is real. None of them close the gap between what the CHIPS Act called dangerous in 2022 and what the concentration data shows in 2025.

[00:18:35]Naming a risk and funding a partial response to it is not the same as resolving it. The documents have been consistent on this. The public summary has been optimistic. If you read primary sources, policy texts, filings, export control registers, and you want that kind of analysis rather than the press release version, this channel is where that work gets done. The next video follows the export control architecture further than this one had room to go.