半导体世纪：2026 AI狂飙下的投资路线图

Key Data: Global semiconductor market size (2025) approximately $792 billion · Q1 2026 sales $298.5 billion · 2026 forecast approximately $975 billion · NVIDIA FY2026 revenue $215.9 billion · TSMC Q1 2026 net profit up 58% year-over-year

1. Why Semiconductors Are More Important Than Ever

Semiconductors are the physical foundation for artificial intelligence, cloud computing, smartphones, electric vehicles, and defense systems. Every time an AI model generates a response, chips perform billions of calculations in milliseconds. All of this runs on silicon.

Unlike previous cycles driven by a single device type (such as mobile phones or PCs), the current surge is supported by AI infrastructure spending. In 2026, the five major hyperscalers have committed over $600 billion to AI infrastructure, representing a 36% year-over-year increase.

This fundamental shift in demand structure is reflected in the fact that high-value AI chips contribute approximately half of the industry's revenue, yet account for less than 0.2% of total shipment volume. Semiconductors have evolved from consumer electronics components into strategic assets for companies with market capitalizations exceeding $10 trillion.

Educational Note: A modern AI chip contains billions of transistors etched onto a silicon die the size of a fingernail. The "nanometer" value of a chip represents the size of these features; a smaller nanometer number means more transistors can be integrated onto each chip, resulting in greater computational power. The more advanced the node, the more difficult the manufacturing process required.

2. Four Core Tracks: Who Controls the Silicon Blueprint?

Investors need to distinguish four key roles in the supply chain, rather than conflating them:

Designers (The Architects): These companies design chips but do not manufacture them. They own the intellectual property and hand the design blueprints to manufacturers. Since they don't operate factories, their gross margins are the highest in the tech industry, typically exceeding 70%. NVIDIA, AMD, Qualcomm, Apple, and Broadcom are all fabless companies.

Foundries (The Manufacturers): Foundries perform large-scale chip manufacturing in facilities known as fabs, with single plant construction costs reaching $20 billion or more. TSMC holds approximately 70% to 72% revenue share in the overall global foundry market and produces roughly 90% of the world's most advanced chips at 3nm and below. Every NVIDIA Blackwell GPU, every Apple A-series processor, and every advanced AI accelerator from the hyperscalers comes out of TSMC's fabs in Taiwan. This concentration means the world's most critical technology supply chain operates within a geographical area roughly the size of Belgium, located just 180 kilometers from mainland China.

Equipment Makers (The Tool Providers): Without the machines to make chips, you cannot make chips. ASML is the only company in the world capable of manufacturing extreme ultraviolet lithography machines, which are essential for patterning chip features at the 7nm node and below. Without ASML, the entire semiconductor technology roadmap would stall. Applied Materials, Lam Research, and KLA provide other critical tools needed for deposition, etching, and inspection processes.

Memory Makers (The Storage Layer): High Bandwidth Memory (HBM) sits right next to GPUs in data center servers, feeding data to the chips at speeds unattainable by any traditional memory. Without sufficient HBM, even the world's fastest GPUs would be idling and waiting. SK Hynix, Samsung, and Micron are the three major producers. HBM sales surpassed $30 billion in 2025, and total memory revenue is expected to reach approximately $200 billion in 2026.

3. Regional Dynamics: The Game and Restructuring of Global Supply Chains

The semiconductor industry has become central to global economic security. In the current complex international environment, investors need to focus on the deep adjustments in supply chain structure and the spillover effects of policies:

Reshoring and Localization: As multiple countries implement semiconductor incentive acts, the geographic concentration of advanced manufacturing processes is beginning to disperse moderately. The progress of TSMC's Arizona fab has become a yardstick for measuring "supply chain resilience." Early procurement agreements from giants like Apple signal a transition of advanced global capacity from a single region towards a multi-polar distribution.

Technology Access and Market Adaptation: Stringent export controls are forcing multinational chip giants to reassess their revenue structures. Companies like NVIDIA and ASML, operating within compliance frameworks, are maintaining global market share by developing customized products. This "compliance-driven innovation" is both a corporate survival strategy and a reflection of the rigid global demand for high-performance computing power.

Reallocation of Computing Resources: In regions with restricted access to computing power, the industry logic is shifting from "pursuing peak computing power" to "optimizing computing efficiency." Leading domestic players and model developers are attempting to alleviate the structural contradiction between computing power supply and demand through software optimization, architectural innovation (such as compute-in-memory), and deploying local alternatives in specific scenarios.

New Forms of Cross-Border Flow: Driven by the inertia of globalization, the cross-border flow of computing resources is taking on more covert and diverse forms. Policymakers are strengthening regulation by improving supply chain transparency and establishing chip traceability mechanisms. For investors, this means compliance risk has become a key dimension for assessing the premium on semiconductor assets.

 4. Key Companies to Watch

NVIDIA (NVDA)

NVIDIA is the most iconic company in the current semiconductor cycle. Its GPUs have become the default hardware for training AI models, and the CUDA software platform builds a software ecosystem moat more durable than any hardware advantage.

Key Financial Data:

• FY2026 Total Revenue: $215.9 billion, up 65% year-over-year (SEC Form 8-K, February 2026)
• Data Center Revenue: Approximately $193.7 billion to $194.0 billion, up 68% year-over-year
• FY2026 Q4 Revenue: $68.1 billion, up 73% year-over-year
• NVIDIA holds approximately 15.8% revenue share of the global semiconductor market
• Forward P/E Ratio: Approximately 32x

Core Issues for Investors:

• The Vera Rubin platform, built on TSMC's 3nm process with 336 billion transistors, offers up to 10x lower inference costs compared to Blackwell. AWS, Google Cloud, Microsoft Azure, and Oracle Cloud have all committed to deployment. NVIDIA has secured the majority of HBM4 supply from SK Hynix and Samsung.
• The depth of the CUDA moat is underestimated by most investors. Millions of developers have written AI software based on CUDA. Switching to a competitor's chip means rewriting years of accumulated code, creating enormous switching friction.
• Google, Amazon, and Microsoft building their own internal chips to reduce dependence on NVIDIA is the most significant long-term structural risk.
• Export controls to China represent one of the most significant implicit revenue pressures among current tech companies.

TSMC (TSM)

TSMC is both the world's most critical and geographically concentrated node in the technology supply chain.

Key Financial Data:

• 2025 Revenue: Approximately $122.5 billion to $122.9 billion, up roughly 31% to 36% year-over-year
• Q1 2026 Net Profit: Up 58% year-over-year, a record high for the fourth consecutive quarter
• Q2 2026 Revenue Guidance: $39.0 billion to $40.2 billion
• FY2026 Capital Expenditure: $52.0 billion to $56.0 billion
• In Q1 2026, 74% of wafer revenue came from advanced processes of 7nm and below
• Forward P/E Ratio: Approximately 24x

Core Issues for Investors:

• TSMC is the most direct beneficiary of AI chip spending regardless of who wins, making it a volume-oriented infrastructure bet on the entire AI theme, rather than a directional bet on any specific winner.
• The geopolitical risk premium explains TSMC's valuation discount relative to NVIDIA and Broadcom, despite having comparable or even stronger revenue growth. Investors must actively judge: does the 24x forward P/E ratio reasonably reflect the risk of a scenario that has never occurred?
• The Arizona diversification is real but currently limited in scale. A second fab is expected to begin 3nm production by the end of 2026, with Apple's chip procurement agreement providing early commercial validation.

ASML (ASML)

ASML is the only company in the world capable of manufacturing EUV lithography machines. Without these machines, chips below 7nm cannot be made; without these chips, there is no advanced AI.

Core Issues for Investors:

• ASML's EUV monopoly is the result of decades of accumulated expertise in physics, optics, and precision mechanical engineering. No other company is close to developing a comparable machine; this moat cannot be replicated in the short term.
• Every new fab constructed globally, whether supported by the CHIPS Act, Japan's semiconductor investment plan, or TSMC's expansion, represents demand for ASML equipment.
• Export restrictions to China have somewhat compressed its addressable market, and this limitation will persist as long as the current geopolitical environment remains.
• The long-term order backlog provides ASML with rare revenue visibility; customers must order years in advance, a situation highly unusual among most tech companies.

AMD (AMD)

AMD is NVIDIA's most substantial competitor in AI accelerators. It benefits from the same TSMC foundry relationship as NVIDIA and is attracting hyperscalers looking to diversify their supplier base.

Key Financial Data:

• Quarterly sales of the MI308 downgraded version (approved for export to China) reached $390 million
• Data Center GPU Revenue Guidance: 60% compound annual growth rate over the next five years

Core Issues for Investors:

• The bull case rests on hyperscalers' need for supplier diversification. No large tech company wants to rely entirely on a single chip supplier. NVIDIA's market dominance creates a structural incentive for introducing AMD as a second source.
• AMD's ROCm software platform is its most critical challenge. While it has made significant progress, it still lags behind CUDA in developer adoption. Closing the software gap is more important than closing the hardware gap.

Broadcom (AVGO)

Broadcom specializes in designing custom AI accelerators (ASICs) for hyperscalers—chips optimized for specific workloads rather than general-purpose GPUs. The TPUs used by Google across its entire AI product line are chips designed by Broadcom.

Key Financial Data:

• FY2026 AI Semiconductor Revenue Expected to Exceed $30 Billion
• Forward P/E Ratio: Approximately 41x, the highest among major semiconductor companies

Core Issues for Investors:

• As hyperscalers scale up AI deployments, custom chips optimized for specific workloads will become increasingly attractive. Broadcom's partnerships with Google and Meta are deep and stable, giving it a leading position in the custom chip space.
• The 41x forward P/E ratio requires Broadcom to maintain strong execution. Any slowdown in custom chip orders from hyperscalers will have a significant impact at this valuation level.

SK Hynix

SK Hynix leads the HBM market with approximately 53% to 62% market share. Its HBM3e is the memory standard for NVIDIA's Blackwell GPU, and HBM4 will be integrated into NVIDIA's Rubin platform, for which NVIDIA has secured most of the supply.

Core Issues for Investors:

• HBM is the real bottleneck for AI chip deployment. Even if NVIDIA delivers every GPU on time, without sufficient HBM these GPUs cannot operate at full capacity. This gives SK Hynix extraordinary pricing power in the current AI infrastructure build-out.
• SK Hynix is listed on the Korea Exchange. Exposure can be gained through Korean brokerage accounts, some international brokerages, or indirectly via semiconductor ETFs.
• Memory has historically been highly cyclical. Although HBM has some natural barriers to oversupply due to its specialized manufacturing process requirements, investors still need to understand the cyclical risks inherent in the memory sector.

5. Semiconductor ETFs

SMH — VanEck Semiconductor ETF

The most widely used semiconductor ETF, with assets under management of approximately $46 billion to $47 billion. It holds 26 companies, covering chip designers, foundries, equipment manufacturers, and memory producers. Top holdings: NVIDIA approximately 19.4%, TSMC approximately 11.6%, Broadcom approximately 7.7%. Expense ratio: 0.35%. Widely considered the most efficient single tool for covering the full supply chain of the AI semiconductor theme.

SOXX — iShares Semiconductor ETF

SMH's closest competitor, holding 30 companies. Long-term historical returns are roughly in line with SMH. Expense ratio: 0.35%. Five-year return as of 2025 was approximately 140%.

SOXQ — Invesco PHLX Semiconductor ETF

Provides roughly equivalent sector coverage to SMH and SOXX, but with a significantly lower expense ratio. Expense ratio: 0.19%, the lowest among major semiconductor ETFs, making it the best option for cost-conscious investors seeking similar sector exposure.

Educational Note: When comparing ETFs, pay attention to the weighting methodology. SMH uses a capped market-cap weighting, ensuring NVIDIA does not become overly concentrated. Understanding how an ETF is constructed helps you know what you actually own and how its performance might differ during sector rotations.

6. Key Risk Warnings for 2026

AI Concentration Risk. The entire industry has put all its eggs in the AI basket. If AI infrastructure spending slows due to disappointing monetization, geopolitical shocks, or efficiency breakthroughs, the impact on semiconductor revenue will be direct and immediate. Deloitte, against the backdrop of record industry revenues, explicitly lists this as a core risk.

Geopolitical and Supply Chain Risk. TSMC produces approximately 90% of the world's most advanced chips in Taiwan. Any disruption to manufacturing operations in Taiwan would have an impact on the entire global tech industry that is difficult to overstate. The Arizona diversification is progressing, but truly shifting the manufacturing center of gravity away from Taiwan will take years.

Export Control Policy Uncertainty. U.S. semiconductor export controls are subject to political influence and policy change risks. The current administration has both maintained some controls and eased others, including revoking the Biden-era AI diffusion rule. Future policy decisions could open new markets for U.S. chip companies or close existing channels.

Memory Cyclicality Risk. Driven by AI demand, consumer memory prices surged approximately 4x between September and November 2025, with expectations of a further increase of up to 50% in early 2026. Deloitte warns that memory capacity expansion could trigger oversupply and a price collapse in late 2026 or 2027. Markets that overshoot on the upside also tend to overshoot on the downside.

Valuation Risk. NVIDIA's forward P/E of roughly 32x and Broadcom's ~41x embed extremely high growth expectations. A single quarter of revenue below expectations, guidance downgrade, or shift in market sentiment could trigger a sharp stock price decline, even if the underlying business remains sound.

7. Key Catalysts to Watch

The Trillion-Dollar Milestone. Q1 2026 semiconductor sales reached $298.5 billion, making the full-year target of $975 billion to $1 trillion realistically achievable. Whether momentum continues in the second half of the year, or AI spending slows leading to a weaker end to the year, is the most watched question for the entire sector.

TSMC Arizona Fab Ramp-Up. The second Arizona fab is scheduled to start 3nm chip production in late 2026. Yields and output volume will determine the speed at which the U.S. reduces its reliance on Taiwan for manufacturing; Apple's chip procurement agreement provides the first meaningful commercial validation.

NVIDIA Vera Rubin Platform Deployment. The promise of 10x lower inference costs is NVIDIA's most important product milestone. Successful deployment by hyperscalers will significantly extend NVIDIA's data center revenue growth curve; any delays or performance shortfalls would be a major negative catalyst.

AMD Market Share Progress. AMD's MI350 and MI400 products, expected in 2026, will test whether its ROCm software improvements are sufficient to attract large-scale deployments by hyperscalers, moving beyond the current pilot project stage.

Memory Pricing and HBM4 Supply. The integration of HBM4 with NVIDIA's Rubin platform creates new demand pull. Tracking SK Hynix's HBM4 production yields, along with the progress of Samsung and Micron in HBM4 product certification, will be key signals for determining memory pricing dynamics in 2027.

Framework for Researching This Sector:

• Investors seeking the highest conviction exposure to AI chips will focus on NVIDIA, accepting the risks posed by export control revenue constraints and current valuation levels
• Investors seeking AI infrastructure exposure while reducing single-stock concentration risk will research SMH or SOXX, covering the complete supply chain
• Investors who believe TSMC's geopolitical discount is excessive relative to its ongoing diversification efforts may find its lower valuation multiple compared to its growth rate worthy of deeper investigation
• Investors seeking exposure to the most defensible segment of the supply chain will focus on ASML, as every new fab built anywhere in the world creates demand for its equipment

The demand is real. The growth is extraordinary. The risks—including geopolitical concentration, AI demand dependency,