US semiconductor gambit strains allies, spurs China's tech resilience

By Li Xuan

Lead: While Trump's semiconductor tariffs may yield short-term gains, China's long-term strategy of patience, scale and efficiency gains will allow it to remain competitive in the global AI race.

U.S. President Donald Trump signaled a shift toward aggressive unilateralism on March 5, 2025, rejecting the $52.7 billion CHIPS Act as a "handout to rich companies." Instead of subsidies for foreign firms, Trump favors punitive tariffs — up to 100% — to reshore chip production.

This tariff-first strategy pressured Taiwan Semiconductor Manufacturing Company (TSMC) into pledging $100 billion for five U.S. factories to avoid steep duties. Trump hailed this as proof of his approach's success. Yet, this shift — executed through direct negotiations with TSMC — raises questions about the long-term implications for both TSMC and Taiwan's economy, while offering insight into how the U.S. might handle emerging Chinese players like DeepSeek under tightened export controls.

"The American Trap"

TSMC's expansion into the U.S., driven by Trump's tariff threats, exemplifies Washington's willingness to flex economic muscle over its "allies." The company, which produces an estimated 90% of the world's most advanced chips, has been a cornerstone of Taiwan's economy. Yet, Taiwan has little choice but to accept the lopsided deal. Rejecting the deal would expose the island to punitive tariffs of up to 100%.

For TSMC, the shift poses operational challenges. Building and staffing advanced fabs in the U.S. is costlier than in Taiwan, where labor and infrastructure are optimized for efficiency. Industry analysts estimate that production costs could rise by up to 30%, potentially eroding TSMC's competitive edge against rivals like Samsung.

What TSMC experiences could be an updated version of Le piège Américain ("the American Trap"), where Alstom, the French energy and transport company, was forced to sell its energy division to General Electric in 2015, resulting in France's loss of its control over some essential industrial technologies.

Even worse, Taiwan's broader economic development hangs in the balance. Semiconductors account for around 35-40% of Taiwan's total export value, with TSMC alone contributing significantly to GDP growth. The forced migration of production capacity risks hollowing out this economic pillar, as jobs, innovation and supply chain ecosystems follow the fabs overseas. In fact, Taiwan faces a stark trade-off: short-term tariff relief at the expense of a long-term strategic economic pillar.

The semiconductor supply chain is highly fragmented, with no single economy able to produce an entire chip independently. The process spans three key areas: chip design, manufacturing and final assembly. U.S. firms like Nvidia lead in semiconductor architecture, particularly with software ecosystems such as CUDA. Taiwan island and South Korea dominate chip manufacturing, producing semiconductors with exceptionally low defect rates. Assembly, testing and packaging are predominantly carried out in Asia, where lower labor costs make large-scale operations more efficient.

The bipartisan CHIPS Act of 2022 primarily aims to strengthen U.S. chip foundry construction rather than advancing chip design capabilities, addressing a long-term decline in domestic semiconductor manufacturing. The U.S. share of global foundry capacity has plummeted from 37% in 1990 to just 12% in 2020.

Taiwan is currently a dominant force in chip manufacturing within the global supply chain, a competitive edge that America acknowledges as strategically significant in the AI era. Washington views semiconductor manufacturing as a critical national security asset and a latent competitive advantage yet to be fully realized.

DeepSeek and US export controls

Across the Strait, the Chinese mainland's semiconductor ambitions face a different U.S. strategy: unrelenting export controls. The Biden administration's January 2025 "Foundry Rule" tightened restrictions on advanced chip production, cutting Chinese firms like Huawei off from TSMC's cutting-edge capacity. Trump's administration seems poised to maintain, if not intensify, this approach, particularly toward emerging players like DeepSeek, a Chinese AI innovator under High-Flyer Capital Management.

DeepSeek's original innovations in algorithmic efficiency have rattled Western tech circles. Counterintuitively, the Center for Strategic and International Studies (CSIS) argues that DeepSeek's advancements in AI computational efficiency could disproportionately benefit U.S. firms over Chinese ones. With access to vastly superior computing resources, American companies can leverage these techniques to enhance AI performance and serve a broader customer base. Notably, it's far simpler for the U.S. to adopt DeepSeek's algorithmic innovations than for China to replicate America's computing infrastructure, including AI systems and cutting-edge chips. In fact, U.S. companies such as Microsoft and Meta have already adopted DeepSeek's techniques to enhance their AI capabilities.

The U.S. will likely persist — or even intensify — its export control measures, targeting DeepSeek with the same rigor applied to Huawei. This includes restricting access to Nvidia's advanced chips, such as the H100 and H800, as well as ASML's extreme ultraviolet (EUV) lithography machines critical for cutting-edge semiconductor production.

China's path forward

The U.S. views DeepSeek as a strategic challenge not merely through its emergence, but via its potential to bolster Huawei's CANN software ecosystem — a move that threatens NVIDIA's CUDA dominance, particularly as AI inference demand grows. A collaboration between DeepSeek and Huawei could bridge the software gap between CANN and CUDA, boosting the viability of Huawei's Ascend chips for inference tasks, which Barclays predicts will account for 70% of AI compute demand by 2026. If DeepSeek's expertise strengthens Huawei's ecosystem, China may establish a competitive foothold despite hardware constraints.

SMIC, Huawei's manufacturing partner, aims to match TSMC's role by scaling up 7-nanometer chip production. The company plans to produce 50,000 wafers monthly by late 2025, potentially yielding millions of Ascend 910C chips annually. Nikkei Asia reports that Huawei has invested $1.66 billion in advanced lithography machines, employing top talent in Shanghai and Shenzhen. While this push aims to overcome the 7 nm production barrier, significant breakthroughs might still be years away.

China continues to strongly support its semiconductor industry. The country's strategy hinges on patience and scale. While it may not match U.S. chip density soon, efficiency gains, domestic supply chains and a vast internal market could offset this gap. As Trump dismantles subsidies and pressures allies, China finds room to maneuver, turning adversity into opportunity. The global semiconductor race, far from decided, will test resilience as much as innovation — and here, China's resolve shines through.

Looking ahead, the AI semiconductor race will be shaped by more than just technological advancements. It will hinge on economic resilience, political strategy and adaptability in an increasingly divided global market. China's vast domestic market provides a unique scaling advantage, reinforcing its determination to remain competitive. Currently, Nvidia's dominance in AI chip design — anchored by its CUDA ecosystem — sets a formidable benchmark that China struggles to match, constrained by 7 nm manufacturing limits.

Yet breakthroughs by DeepSeek in computational efficiency suggest an alternative path — one where software optimization helps offset hardware constraints. If Huawei's Ascend chips, powered by its evolving CANN framework, integrate such innovations, China could carve out a niche in AI deployment, particularly in cost-sensitive markets and domestically driven applications.

Moreover, the Tiangong space station, now the only operational orbit platform, the Beidou navigation satellite system, the bullet train technology and many other achievements by China have been accomplished in spite of Western exclusions and restraints. This demonstrates that technology blockages have served to accelerate this nation's self-driven innovation and ecosystem development, contrary to the designs of the U.S and its allies. All things considered, China is poised to sustain progress in frontier areas in AI and quantum computing, further strengthening its global competitiveness in the coming months and years.

Li Xuan is an assistant professor at the Global Institute for Zhejiang Merchants Development at Zhejiang University of Technology and a researcher at Zhejiang Development and Security Research Think Tank Alliance. She has served as a policy consultant for the Organization for Economic Cooperation and Development.