To The Moon

Outlook for ASML and TSMC Amid Trump's Tariff Hike (30-35%)

The semiconductor industry faces a complex landscape shaped by surging AI demand and escalating trade tensions. Here's the outlook for ASML and TSMC:

1. Strong Earnings Driven by AI Boom

TSMC reported record Q2 2025 revenue of NT$933.8 billion ($31.9B), up 38.6% YoY, fueled by AI chip demand from clients like Nvidia and Apple .

ASML saw a rebound in Q4 2024 earnings (€3.94B orders), though Q1 2025 orders missed estimates due to tariff uncertainty .

2. Tariff Impacts and Industry Risks

Trump's 30-35% tariffs on semiconductors (expanding from earlier 25% proposals) aim to reshore manufacturing but face pushback:

- Supply chain disruptions: Auto giants (Tesla, GM) and tech firms oppose higher costs .

- Retaliation risks: China and ASEAN nations like Malaysia (24% U.S. tariffs paused for 90 days) may impose countermeasures .

- Cost inflation: TSMC acknowledges tariffs will raise production expenses, though AI demand offsets short-term risks .

3. Geopolitical and Market Uncertainty

- U.S.-China decoupling: Trump's policies prioritize “technological sovereignty,” pressuring ASML/TSMC to limit China exposure. ASML faces Dutch restrictions on servicing Chinese clients .

- Regional diversification: TSMC expands U.S. and EU fabs to mitigate risks. Malaysia and the Philippines face investment delays due to tariff uncertainty .

4. Long-Term Projections

- AI-driven growth: TSMC expects AI chip revenue to double in 2025, with 20% annual revenue growth despite tariffs .

- ASML's caution: While AI boosts demand for advanced lithography tools (e.g., High-NA EUV), tariffs and export controls cloud 2025–2026 forecasts .

The semiconductor sector remains bifurcated—AI growth offsets tariff risks in the near term, but prolonged trade wars could stifle innovation and raise costs. TSMC's global expansion and ASML's tech leadership position them to navigate volatility, though policy shifts remain a wild card.

AI Chips Most Impacted by U.S.-China Tariffs

The expanded U.S. export controls primarily target high-performance AI chips designed for data centers and advanced computing. Here are the most affected models:

1. Nvidia's Modified China-Specific Chips

- A800/H800: These were modified versions of Nvidia’s A100/H100 GPUs, created to comply with 2022 restrictions by reducing interconnect bandwidth. Despite being "weaker," they’re now banned under updated rules .

- A100/H100: Original flagship data-center GPUs, already prohibited for direct export to China since 2022.

2. AMD's MI Series

- MI200/MI300: AMD’s data-center GPUs face stricter licensing requirements. The MI200 (already impacted by 2022 rules) and upcoming MI300 chips risk being blocked if deemed too powerful .

3. Advanced Training/Inference Chips

New rules now restrict lower-tier chips that previously bypassed controls, including:

- Nvidia's L40S and RTX 4090 (dual-use risks for AI training)

- Intel's Gaudi2 accelerators .

Impact on Chinese AI Development

These tariffs aim to cripple China's ability to train large language models (LLMs) like Baidu’s Ernie or Alibaba’s Tongyi. While Chinese firms are pivoting to domestic alternatives (e.g., Huawei’s Ascend 910B), performance gaps remain significant .

Key Loophole: Cloud-based AI services (e.g., AWS, Azure) remain unrestricted, allowing Chinese firms to rent U.S. chip-powered computing .

AI Applications Most Impacted by Tariffs and Trade Restrictions

The latest U.S.-China trade measures disproportionately affect AI applications requiring advanced semiconductors, cross-border data flows, and global supply chain integration. Here are the most vulnerable sectors:

1. Generative AI and Large Language Models (LLMs)

Applications like ChatGPT and AI-powered content creation tools rely on Nvidia A100/H100 GPUs and similar high-end chips now restricted under expanded export controls. Training LLMs requires thousands of these chips, making them vulnerable to:

- Hardware shortages: Chinese AI firms face 6-12 month delays obtaining compliant chips like Huawei’s Ascend 910B .

- Cloud service limitations: While U.S. cloud providers (AWS, Azure) remain accessible, Beijing’s proposed "Great Firewall 2.0" could block foreign AI APIs .

2. Autonomous Systems and Robotics

Self-driving vehicles (projected 10% market penetration by 2030) and smart manufacturing robots depend on:

- Sensor fusion chips: NVIDIA's Orin and Qualcomm’s Ride Flex platforms, now subject to 35% tariffs when exported to China.

- Real-time data processing: Localization mandates under China’s Data Security Law disrupt training datasets for autonomous navigation .

3. AI-Powered Trade Optimization Tools

Tariff-sensitive applications like:

- Supply chain predictive analytics: Used by 78% of Fortune 500 companies to forecast disruptions. Training these models requires global logistics data now fragmented by trade walls.

- Dynamic pricing engines: Tools like X.me Digital’s telecom tariff optimizer lose accuracy without cross-border data flows .

4. Language and Localization Services

AI translation tools (e.g., eBay's machine translation, which boosted LATAM exports by 17.5%) face:

- Model degradation: Restricted access to multilingual training data from sanctioned regions.

- Hardware bottlenecks: Low-latency translation requires edge chips like Google's TPU v5e, now export-controlled .

Mitigation Strategies

Leading firms are adapting through:

- Chiplet-based designs: AMD and Intel's modular chips bypass export controls by combining approved components.

- Federated learning: Training AI models on decentralized data to comply with localization laws.

- Nearshoring compute: AWS and Alibaba now offer "AI sovereignty zones" with region-specific cloud infrastructure .

The AI applications most at risk are those combining dependency on sanctioned hardware, cross-border data, and just-in-time global coordination—key factors in 84% of tariff-impacted use cases .

AI Chips Impacting ASML and TSMC Amid Trade Restrictions

The U.S.-China tech war and AI-driven demand are reshaping semiconductor supply chains, with these key chips affecting ASML and TSMC:

1. Nvidia's A100/H100 and Modified A800/H800

- Impact on TSMC: TSMC manufactures these advanced AI GPUs, which face U.S. export bans to China. Despite tariffs, TSMC’s AI chip orders surged 30% YoY in Q2 2025 due to strong Nvidia demand .

- ASML's Role: These chips require ASML's EUV lithography machines. However, U.S. pressure to restrict ASML's sales to China threatens its market share in Chinese AI chip production .

2. AMD's MI300 Series

- TSMC's Production: TSMC fabricates AMD's MI300 accelerators, critical for AI data centers. Tariffs could raise costs, but AI demand offsets short-term risks .

- ASML's Exposure: MI300 chips rely on ASML's EUV systems. U.S. export controls on advanced chipmaking tools could slow ASML's order growth in non-China markets .

3. Huawei's Ascend 910B

- TSMC's Dilemma: Chinese firms like Huawei downgrade chip designs (e.g., Ascend 910B) to comply with U.S. rules, forcing TSMC to balance geopolitical risks and revenue from Chinese clients .

- ASML's Constraints: ASML cannot sell cutting-edge EUV tools to China, limiting its ability to support domestic Chinese alternatives to TSMC .

4. Intel's Gaudi3 and Future Chips

- TSMC's Competition: Intel’s Gaudi3 (produced by TSMC) competes with Nvidia but faces tariff-related cost inflation. TSMC’s Arizona fabs aim to mitigate this .

- ASML's High-NA EUV: Intel plans to adopt ASML’s next-gen High-NA EUV tools for 14A chips, while TSMC delays adoption, creating divergent demand .

Strategic Implications

- TSMC's AI Dominance: 62% of TSMC's 2025 revenue comes from AI chips, insulating it from tariff impacts in the short term .

- ASML's Uncertain Outlook: While AI drives demand for EUV tools, U.S. tariffs and China restrictions create volatility. Q1 2025 orders missed estimates by €1B, reflecting caution .

The chips most critical to both companies are advanced AI GPUs (Nvidia/AMD) and China-focused designs (Huawei), which hinge on geopolitical decisions and ASML’s equipment access.