How Do You Invest in the Semiconductor Industry?

Semiconductors are the most strategically important manufactured goods in the modern economy — the essential components that enable every digital device, from smartphones to satellites to surgical robots. Semiconductor investing offers extraordinary return potential during supply-constrained or demand-surge cycles, as demonstrated by the AI-driven revenue explosion of 2023–2024, but also carries severe cyclicality that can produce 40–60% sector drawdowns when inventory cycles turn negative. Understanding the semiconductor industry's structure, competitive dynamics, and cycle patterns is prerequisite knowledge for any investor drawn to this high-stakes corner of the IT sector.

Quick definition: The semiconductor industry encompasses companies that design, manufacture, and sell integrated circuits and the equipment used to produce them, with distinct business models spanning fabless design firms, integrated device manufacturers, foundries, and equipment makers.

Key takeaways

• Semiconductors divide into fabless designers, integrated device manufacturers (IDMs), pure-play foundries, and equipment makers — each with different risk/return profiles
• TSMC in Taiwan produces the world's most advanced chips, creating geopolitical concentration risk
• Semiconductor revenue cycles are severe: -30% to -50% declines in down-cycles are not unusual
• AI infrastructure demand created an unprecedented GPU revenue cycle in 2023–2024
• SOXX and other semiconductor ETFs provide focused exposure to this high-volatility sub-sector

The semiconductor value chain: four types of companies

The semiconductor industry is organized around four distinct business models, each occupying a different role in the value chain:

Fabless semiconductor companies design chips but do not own fabrication facilities. They outsource all manufacturing to foundries. This asset-light model allows fabless companies to invest heavily in design talent and intellectual property without the enormous capital costs of building and maintaining fabs. Fabless companies typically generate the highest margins in the industry — gross margins of 50–70% are common. Nvidia, AMD, and Qualcomm are examples of fabless companies.

Integrated Device Manufacturers (IDMs) design and manufacture their own chips. Intel is the archetype: it has historically designed microprocessors and manufactured them in its own fabs. IDMs carry higher capital intensity but maintain tighter control over manufacturing processes and product roadmaps. The IDM model has come under pressure as the capital requirements for leading-edge manufacturing have escalated beyond what most companies can sustain independently.

Pure-play foundries manufacture chips designed by others. They own enormous fabrication facilities but do not design chips themselves. TSMC (Taiwan Semiconductor Manufacturing Company) is the dominant global foundry, manufacturing chips for Apple, Nvidia, AMD, Qualcomm, and hundreds of other companies. Samsung and GlobalFoundries are other major foundries. The foundry business is extraordinarily capital intensive — TSMC spent approximately $36 billion in capital expenditures in 2023 alone.

Semiconductor equipment companies make the machines that build chips. Companies like ASML (which makes the extreme ultraviolet lithography machines required for the most advanced nodes), Applied Materials, and Lam Research are essential suppliers whose equipment determines what chips can be manufactured. Equipment companies have natural oligopoly characteristics because the technology barriers to entry are extreme.

The semiconductor cycle: boom, bust, and repeat

Semiconductor revenues follow a pronounced inventory cycle driven by the mismatch between long manufacturing lead times and rapidly changing end-market demand. When demand for electronic products rises, chip buyers increase orders aggressively and build inventory to ensure supply security. This surge in orders drives utilization rates at foundries to 100%, encourages capacity expansion, and produces euphoric revenue growth at chip companies.

When demand moderates — even slightly — the inventory correction can be sharp and prolonged. Chip buyers who built inventory during the shortage phase stop ordering, sometimes for quarters, while they work through excess stock. Foundry utilization rates fall from 100% to 70–75%, directly compressing chip company revenues and margins.

A concrete example: Semiconductor revenues broadly fell roughly 15–20% in 2023 following the COVID-era demand surge and subsequent inventory buildup. Memory chip prices fell more severely — DRAM spot prices dropped 50–60% from their 2022 peaks. Investors who did not anticipate this cycle suffered significant losses; those who recognized the inventory overhang early and reduced exposure avoided the worst.

How it flows

AI and the semiconductor revolution

The artificial intelligence investment cycle that accelerated in 2023 created the most significant sustained demand surge in semiconductor history for specific chip categories. Graphics processing units (GPUs), originally designed for video games, proved extraordinarily well-suited to AI training workloads because their massively parallel architecture can perform the matrix multiplication operations required for neural network training far faster than traditional CPUs.

Nvidia's H100 and H200 GPUs became the essential hardware of AI data center buildout. Cloud hyperscalers — Amazon Web Services, Microsoft Azure, Google Cloud, Meta — spent collectively hundreds of billions of dollars purchasing GPU clusters to train large language models and serve AI inference workloads. This concentrated demand created a supply shortage for leading-edge GPUs, allowing Nvidia to charge premium prices and generate operating margins exceeding 60%.

The AI demand cycle also benefited high-bandwidth memory (HBM) suppliers (SK Hynix, Micron), advanced packaging companies, and power semiconductors, while partially offsetting the weakness in smartphone and PC chips from cyclical consumer demand moderation.

Geopolitical concentration risk

The semiconductor supply chain carries acute geopolitical concentration risk. TSMC manufactures approximately 90% of the world's most advanced (sub-5nm) chips from facilities in Taiwan, an island over which China claims sovereignty and periodically threatens military action. A disruption to TSMC's Taiwan operations would cause an immediate global economic crisis, as virtually every advanced electronic device — smartphones, data center servers, automotive electronics, military systems — depends on TSMC-manufactured chips.

This concentration risk has driven significant government policy responses. The US CHIPS Act of 2022 provided approximately $52 billion in subsidies and tax credits to encourage semiconductor manufacturing investment in the United States. TSMC is building fabs in Arizona. Intel is receiving billions in CHIPS Act funding. European governments are offering similar incentives to attract semiconductor investment. This reshoring effort will take a decade or more to meaningfully diversify supply chain geography, but it represents a structural shift in semiconductor industry capital allocation.

Export controls are another geopolitical dimension. The US has implemented progressively tighter export controls on advanced semiconductor technology to China, restricting sales of the most capable AI chips and semiconductor manufacturing equipment to Chinese companies. These controls aim to slow China's AI and military semiconductor capabilities but also constrain revenue opportunities for US chip companies.

Real-world examples

Nvidia's transformation from a gaming graphics chip company to an AI infrastructure company illustrates how AI demand can reshape a semiconductor business in a remarkably short period. In fiscal year 2022 (ending January 2022), Nvidia generated roughly $16.7 billion in total revenues, with gaming and data center approximately evenly split. By fiscal year 2024, total revenues had risen to approximately $60.9 billion — with data center (primarily AI GPUs) representing roughly 78% of revenue. This is one of the fastest revenue transformations of a large-cap company in stock market history.

The DRAM memory cycle of 2021–2023 illustrates the semiconductor inventory cycle at textbook scale. Memory chip prices surged in 2021–2022 as pandemic-driven electronics demand and supply chain constraints created shortages. Companies like Micron and SK Hynix reported record revenues and margins. When PC and smartphone demand normalized in 2022 and customers began drawing down inventory, memory chip prices collapsed. Micron's quarterly revenues fell roughly 50% from peak to trough between 2022 and 2023, demonstrating the commodity-like cyclicality of memory semiconductors.

Common mistakes

Treating semiconductors as a single homogeneous industry. Memory chips (DRAM, NAND flash) are commodity products whose prices are set by global supply/demand balance. Logic chips (CPUs, GPUs, custom ASICs) carry high intellectual property value and pricing power. Analog chips serve industrial and automotive markets with long design-in cycles and sticky customer relationships. Each sub-segment has different competitive dynamics and investment characteristics.

Ignoring the equipment cycle. Semiconductor equipment orders lead the chip revenue cycle. When chip companies cut capital spending during down cycles, equipment company revenues fall before chip company revenues recover. Investors who track equipment company orders and guidance can gain early insight into the direction of the semiconductor cycle.

Underestimating the impact of geopolitics. Export controls, tariffs, and potential Taiwan conflict scenarios represent tail risks that do not appear prominently in financial models but can have devastating impact on semiconductor supply chains and company revenues. Position sizing should reflect this geopolitical risk premium.

FAQ

What is the difference between a chip designer and a chip manufacturer?

A chip designer (fabless company) creates the circuit design — the intellectual property that defines what the chip does. A chip manufacturer (foundry or IDM) physically produces the chip in a fabrication plant using the designer's blueprints. The fabless model separates these functions, allowing designers to focus on intellectual property while foundries focus on manufacturing excellence.

How do investors track the semiconductor cycle?

Key indicators include the Philadelphia Semiconductor Index (SOXX) as a market signal, semiconductor company gross margin trends, foundry utilization rate disclosures, memory spot prices, and the Semiconductor Industry Association (SIA) monthly shipment data. The book-to-bill ratio — the ratio of new orders received to chips shipped — is a classic leading indicator of cycle direction.

What are the best ETFs for semiconductor exposure?

SOXX (iShares PHLX Semiconductor ETF) and SMH (VanEck Semiconductor ETF) are the two most liquid pure-play semiconductor ETFs. Both concentrate holdings in the 25–30 largest semiconductor companies. SOXX is slightly more diversified; SMH has historically had higher weighting to Nvidia and TSMC ADRs. Both have expense ratios of approximately 0.35%.

Does the CHIPS Act guarantee US semiconductor competitiveness?

No. The CHIPS Act provides financial incentives that reduce the cost of building US fabs, but the technology gap between US-manufactured chips and TSMC's most advanced processes will take years to close. TSMC brings not only capital but decades of accumulated process engineering knowledge that cannot be replicated quickly. The Act improves the long-run trajectory of US semiconductor manufacturing but is not a near-term solution to TSMC concentration risk. Confirm current CHIPS Act program details at commerce.gov.

Related concepts

• IT Sector Overview
• Software Business Models
• IT and Interest Rates
• IT Global Supply Chains
• AI Impact on IT Sector

Summary

The semiconductor industry is one of the most rewarding and most challenging sectors in public equity investing — combining world-class intellectual property with severe inventory cycles, geopolitical concentration risk, and extraordinary capital requirements for manufacturing leadership. Understanding the four business model types (fabless, IDM, foundry, equipment), the inventory cycle dynamics, and the AI demand revolution reshaping GPU economics is essential for investors in this sub-sector. Semiconductor stocks can double or fall by half within a single year depending on cycle positioning, making entry point and position sizing decisions critical. The sector's strategic importance to national economies and military capabilities adds a geopolitical dimension that requires ongoing monitoring beyond traditional financial analysis.

Next

→ Software Business Models in the IT Sector