[4/16] The Semiconductor War and AI Hegemony: The Crossroads for NVIDIA, ASML, and Meta

In April 2025, the U.S. government further tightened semiconductor export restrictions to China, forcing major U.S. chip giants like NVIDIA and AMD to take massive impairment charges, causing global tech stocks to plummet. Furthermore, ASML, the largest semiconductor manufacturing equipment maker, announced that it had missed its order targets, spreading uncertainty about the industry's future. Meanwhile, the second-largest ride-sharing company, Lyft, acquired the major European taxi-hailing service Free Now, signaling a move to seek growth in markets outside the U.S. and in the autonomous driving business. Additionally, the U.S. Federal Trade Commission (FTC) has proposed a narrow market definition of the "family-oriented social media market" in its antitrust lawsuit against Meta (formerly Facebook), leading to discussions about the possibility of breaking up the platform. At the same time, the U.S. government has promoted "national competitiveness in the AI era," outlining a new industrial policy that balances the efficiency of research and development with the prevention of technology leakage. This article provides an overview of these developments and analyzes the future industrial structure and investment environment across four areas: semiconductors, mobility, platforms, and AI policy.

1. The Depths of the U.S.-China Semiconductor Friction and the Mechanism of the Stock Market Crash

1-1. The "Visualization of Regulatory Risk" Indicated by NVIDIA and AMD's Impairment Charges

At the end of March 2025, the U.S. Department of Commerce mandated export licenses even for low-performance GPUs designed for China that had previously bypassed old regulations. In response, NVIDIA recorded a "$5.5 billion inventory valuation loss", and AMD recorded an "$800 million impairment charge." The ratio of sales to China for NVIDIA has halved from 26% in 2022 to 13% currently, and investors perceive this as "the process of stripping China out of the model is accelerating."

1-2. Advanced AI Development and Diffusion Rules

The purpose of the regulation is to "prevent China from obtaining the computing resources necessary for advanced AI training." The Diffusion Rule, scheduled to take effect in May, is expected to set an upper limit on the number of GPUs that can be purchased by country, and depending on its effectiveness, it could pose a risk of further downward revisions to earnings.

1-3. Demand Shifts Under Supply Constraints

NVIDIA's production capacity remains tight, and it can redirect the portion intended for China to advanced data centers or domestic government projects. As a result, a scenario where "volume impacts are offset by high unit prices" remains, but the fact that supply chain geopolitics have been incorporated into valuation metrics is significant.

2. The Plateau in the Capital Investment Cycle Suggested by ASML's Order Stagnation

2-1. The 1 Billion Euro "Mistake" and Customer Wait-and-See Approach

ASML's orders for January-March were about 1 billion euros below expectations. The company's CEO explained, "Uncertainty regarding tariffs is delaying customers' ordering timing." EUV lithography machines cost $380 million each, and if a 35% tariff were imposed, it would result in an "additional cost of $95 million," casting a shadow over the investment decisions of TSMC and Intel.

2-2. Pressure to Relocate Production Countries

The U.S. administration is strengthening its stance of requiring equipment manufacturers to perform final assembly domestically. While some processes, such as light source modules, have already been transferred to the U.S., it will take years of adjustment to move entire massive cleanrooms. Whether ASML can keep pace with the relocation of advanced fabs from "Taiwan to the U.S." will be a medium-term challenge.

3. Lyft's Acquisition of Free Now and the European Mobility Front

3-1. Market Entry Costs and the "Taxi Economic Zone"

Lyft acquired the Mercedes-BMW-affiliated Free Now, gaining a taxi network of over 150,000 vehicles across nine countries. CEO David Risher stated, "We are targeting untapped demand in the 4 billion euro European mobility market where 'phone booking is still mainstream.'"

3-2. Paving the Way for Autonomous Driving

Free Now has strong ties with regulators and fleet management expertise in various European cities. Lyft plans to "first launch in full scale from Atlanta" the autonomous driving pilot it is promoting in the U.S., which will serve as a stepping stone for future horizontal expansion into Europe.

4. Meta vs. FTC—The Future of the "Family SNS Monopoly" Debate

4-1. The Battle Over "Too Narrow Market Definition"

The FTC is continuing its lawsuit seeking the divestiture of Instagram and WhatsApp, arguing that their acquisitions (in 2012 and 2014) hindered competition. The point of contention is the validity of the niche market defined as 'Friends & Family Sharing SNS.' Meta is countering by arguing that 'TikTok, iMessage, and Snap are all actual competitors,' attempting to dilute the claim of monopoly by broadening the definition of the market.

4-2. Implications of the Instagram divestiture memo

A 2018 private message from Zuckerberg submitted as evidence noted that 'spinning off Instagram was an option.' While the FTC argues that 'competition would have been promoted if they had remained independent,' Cato researcher Jennifer Huddleston warns that 'punishing M&A deals after the fact when their success was uncertain will undermine the innovation induced by M&A.'

5. U.S. AI Roadmap: Policies pursuing both promotion and protection

5-1. R&D reform: 'Spending wisely, even if the amount is small'

Presidential Science and Technology Advisor Michael Kratsios explained that the administration is moving away from the 'practice of the same agencies continuing to spend federal research funds on the same topics,' and is expanding prize-based challenges and fast-track grant frameworks. He identified AI, quantum, and next-generation nuclear power as top priority areas.

5-2. Simplification and enforcement of export controls

Regarding the protection pillar, the policy is to 'streamline complex regulations and strictly enforce them in the event of violations.' On the subject of additional restrictions for NVIDIA, Kratsios emphasized that 'an environment where PRC companies can accelerate open-source models cannot be overlooked,' suggesting both institutionalization and strengthened enforcement.

5-3. Balancing academic research and national security

Amidst the intersection of arguments for cutting research funding to universities and the issue of antisemitism, he stated that it is possible to simultaneously pursue two goals: 'supporting fundamental academic research while not tolerating hate.' For the social implementation of research results, there is a plan to promote AI for Science in coordination with NVIDIA's Jensen Huang.

From semiconductors to platforms, mobility, and AI policy, we have reached a phase where 'regulatory shocks stemming from U.S.-China tensions' and 'actual demand underpinning long-term innovation' are clashing. While earnings adjustments will continue in the short term due to stagnant capital investment at ASML and reduced GPU exports, the growth in AI-related CAPEX and the return of manufacturing to the U.S. will serve as structural tailwinds. Companies and investors will need to map out medium- to long-term scenarios using three evaluation axes: (1) the ability to pass on regulatory costs, (2) geographic diversification of supply chains, and (3) R&D efficiency measures.

Related Articles