“As President Trump prepares for his summit with Xi Jinping, Taiwan is often described as one more negotiable item in a crowded U.S.-China agenda that also includes tariffs, rare earths, Iran, fentanyl and export controls. That is a mistake.
Taiwan isn't merely another file in the U.S.-China relationship. It has grown beyond even its traditional role as a lonely outpost for democracy and a vital test of American strategic credibility. Thanks to tech and market developments so recent that they were largely invisible during Mr. Trump's first term, the island is now also the factory floor for U.S. leadership in artificial intelligence.
Mr. Trump has rightly put AI dominance at the center of his national strategy. AI will shape economic growth, military advantage, medical innovation, intelligence collection and geopolitical influence. But America can't win the AI race if it treats Taiwan as a bargaining chip. For the foreseeable future, Taiwan's autonomy is a prerequisite for U.S. AI dominance.
AI doesn't live in the cloud; it lives in fabs, packaging plants, memory stacks, substrate lines, testing facilities and server factories. AI has a geographic map. Its most important point, other than America itself, is Taiwan.
Taiwan's importance to U.S. and allied economic strength may be without historical precedent: a relatively small island that anchors the world's most important technology stack. Yet neither markets nor policymakers seem to appreciate the risk fully.
Taiwan Semiconductor Manufacturing Co., or TSMC, fabricates roughly 90% of the world's most advanced semiconductor chips. Even that striking number vastly understates the U.S. and allied dependency. It can leave the impression that Taiwan's role is mainly a chip-fabrication problem -- and that a few advanced chip-making fabs on TSMC's new campus in Arizona can solve it. They can't.
Advanced AI systems require leading-edge logic chips but also advanced packaging, high-bandwidth-memory integration, chip-on-wafer bonding, substrate integration, testing, module assembly and server-level integration. Packaging technologies such as CoWoS (chip on wafer on substrate) are what combine advanced chips and high-bandwidth memory into the processors that train and run frontier AI models at scale. Packaging is no longer the back end of the semiconductor business but a front line.
This is why the current debate about "onshoring" can be dangerously misleading. TSMC's investment in Phoenix is a major achievement. So are new semiconductor investments elsewhere in America and in places like Japan. They will make the U.S. and its allies more resilient. This sort of building should be accelerated.
But facilities like the one in Phoenix are a down payment, not a solution. This point was underscored in October, when Nvidia and TSMC celebrated the first Nvidia Blackwell wafer produced at TSMC's Phoenix fab. It was a milestone. But before U.S.-made Blackwell wafers can become usable AI systems, they still must journey back across the Pacific for CoWoS advanced packaging and high-bandwidth-memory integration that can be done only in Taiwan.
The packaging firm Amkor's planned Arizona campus is a welcome step to complement TSMC's Phoenix fabs and bring more of the AI-chip manufacturing chain onto U.S. soil. But it also proves the continued importance of Taiwan. Production isn't expected until 2028, and one packaging campus won't erase Taiwan's overwhelming lead in CoWoS capacity, high-bandwidth-memory integration, substrates, testing specialists and yield engineering.
The Trump administration appears to recognize this reality, and has moved aggressively to integrate Taiwanese industry further into the American AI ecosystem. Through a combination of trade pressure, investment agreements and industrial policy, it has accelerated unprecedented Taiwanese commitments to build advanced semiconductor and AI infrastructure on U.S. soil.
Still, Arizona can't become Taiwan overnight -- or even over a decade. Taiwan's advantage isn't merely one company or one factory. It is the cluster: foundries, packaging houses, substrate suppliers, materials firms, equipment engineers, testing specialists, design-service providers and process experts operating in close proximity. That density shortens iteration cycles, improves yields, accelerates ramp-up and compounds tacit knowledge. Decades of operational learning can't be bought instantly, even with generous subsidies and political urgency.
True AI-hardware independence would require far more than fabs: comparable leading-edge manufacturing in the U.S. and allied countries, scaled advanced-packaging capacity, high-bandwidth-memory integration, advanced substrates, domestic or allied materials supply, testing and assembly ecosystems, equivalent yields and genuine design portability across foundries. That is a long-term project -- urgent, necessary and measured potentially in decades.
Mr. Xi understands that control over Taiwan would give Beijing both a military advantage and a chokehold on the physical infrastructure of the AI age. China wouldn't need to operate every plant flawlessly to throw the global AI ecosystem into crisis. The mere prospect of coercive control, disruption or selective access would upend the balance of technological power.
That is the backdrop of any Taiwan talk that comes up at the Trump-Xi summit. Until America can reproduce fabs as well as the packaging, bonding, testing, integration and supplier ecosystems around them, Taiwan's autonomy will remain indispensable to U.S. interests.
A country determined to win the defining technological race of the century can't allow its chief rival to control the industrial base on which that race depends.
---
Mr. Benard is a senior managing director at Cerberus and an adjunct fellow at the Hudson Institute. Mr. Feith is a senior fellow at Hudson. He served on the State Department policy-planning staff and as deputy assistant secretary of state for East Asia (2017-21) and White House senior director for technology and national security (2025).” [1]
When Lithuanian elite declared recently a war, attacking a sovereign Chinese territory, Lithuanian elite hoped to get trust and technology of Taiwanese. They just didn’t know that Taiwanese technology is so complicated and integrated. No way for Taiwanese to share it with naïve Lithuanians. The war with China must end.
As of early 2026, the diplomatic and economic situation between Lithuania, China, and Taiwan has undergone significant developments, with Lithuanian leadership, under Prime Minister Inga Ruginienė, signaling a desire to normalize relations with Beijing.
- Strategic Reevaluation: Lithuanian Prime Minister Inga Ruginienė stated in February 2026 that the 2021 decision to allow the "Taiwanese Representative Office" in Vilnius (using "Taiwanese" rather than "Taipei", hinting that Taiwan is not sovereign territory of China) was a "tactical mistake". She described the move as having "jumped in front of the train and lost".
- Renaming Talks: Reports in February 2026 indicated that Lithuania was considering renaming the office to align with naming conventions used by other European countries (using "Taipei"), in an attempt to restore relations with China.
- Trade Consequences: Following the 2021 decision, China imposed severe trade restrictions on Lithuanian products and those containing Lithuanian components, which caused significant issues for European companies and caused a significant decline in Lithuania-China trade.
- Technological Cooperation: Despite the diplomatic tensions with Beijing, Lithuania has continued to pursue technological and economic cooperation with Taiwan, including in the semiconductor and laser industries. As of April 2026, negotiations for a new Economic Cooperation Action Plan were underway, focusing on high-tech sectors. Talk is cheap. No technology moved to Lithuania.
- Future Outlook: Lithuania is seeking a balance where it can restore normal diplomatic ties with China while maintaining strong economic relations with Taiwan. The normalization of relations with China is a topic of internal political debate in Lithuania.
The situation remains a complex case of navigating economic dependence and technological partnership in a geopolitically charged environment.
1. Taiwan Is the Key to AI Dominance. Benard, Alexander; Feith, David. Wall Street Journal, Eastern edition; New York, N.Y.. 14 May 2026: A15.
Komentarų nėra:
Rašyti komentarą