Semiconductor Cycle Check: A Balanced View on AI Shortages and the Sustainability of a Supercycle

1. Introduction: How Should We Read the Current Cycle?

The semiconductor market has historically moved through repeated cycles of demand expansion, supply shortages, inventory adjustment, capital spending increases, and price declines. The 2025–2026 cycle, however, appears to have a somewhat different structure. AI infrastructure investment is rapidly lifting demand for high-performance GPUs, HBM, server DRAM, leading-edge foundry capacity, and advanced packaging. At the same time, it is difficult to say that all end markets are recovering with the same intensity.

As a result, the current cycle can be interpreted through three broad paths. The first is an asymmetric expansion centered on AI infrastructure. The second is a broad-based supercycle in which demand recovery spreads across the semiconductor industry. The third is an earlier slowdown, in which price increases and supply-chain uncertainty begin to weaken end demand.

Based on currently available public data, the first path — an AI-centered asymmetric expansion — appears relatively more likely. This is not a definitive conclusion, but a cautious reading of the demand, supply, pricing, and policy indicators currently visible. A broader supercycle and an earlier slowdown remain possible and should not be dismissed.

2. Market Size: The Expansion Is Clear, but Interpretation Requires Caution

Multiple sources indicate that the global semiconductor market entered a strong expansion phase in 2025. Gartner reported that worldwide semiconductor revenue reached approximately $793 billion in 2025, up 21% year over year.^[1] WSTS also estimated the 2025 global semiconductor market at $795.6 billion, representing 26.2% year-over-year growth.^[2] Although the two organizations may use different detailed methodologies, both point to the same broad conclusion: the 2025 market reached a historically high level.

The chart below shows global semiconductor revenue on a consistent Gartner basis. After a downturn in 2023, the market recovered in 2024, reaccelerated strongly in 2025, and is expected by market research firms to show a major upside move in 2026. However, much of this increase appears to reflect AI demand and sharp memory price increases. It should therefore be distinguished from a broad-based expansion in unit shipments across all end markets.^[6]

The 2026 outlook is also broadly positive. Gartner forecasts global semiconductor revenue to exceed $1 trillion in 2026, while IDC identifies AI infrastructure and data-center investment as key growth drivers.^[3] These forecasts support the view that the semiconductor market may retain strong near-term growth momentum.

As the figure shows, major institutions generally point to a high level of semiconductor revenue in 2026, but the dispersion across forecasts is also meaningful. This suggests that the market is in a strong growth phase, while also requiring caution when assessing the intensity and durability of that growth.

Importantly, market-size expansion does not automatically imply uniform improvement across all end markets. The 2026 market expansion appears to include not only AI-related volume growth but also significant DRAM and NAND pricing effects.^{[4] [5]} Higher prices can lift supplier revenue, but they can also create pressure in price-sensitive markets such as PCs and smartphones.

3. Demand Structure: The AI-Centered Expansion Looks More Visible

Based on currently available evidence, the most visible growth axis in semiconductor demand is AI infrastructure, data centers, servers, and high-performance computing. WSTS highlighted strong growth in the computer application segment in 2025, while IDC points to data-center semiconductors and AI infrastructure spending as central drivers of future market expansion.^{[2] [3]}

Company-level data point in the same direction. According to TSMC’s first-quarter 2026 materials, HPC accounted for 61% of revenue, while 7nm and below accounted for 74% of wafer revenue. ^[7] This suggests that demand for leading-edge nodes has shifted from a smartphone-centered structure toward one led by AI and high-performance computing.

In contrast, the smartphone market may remain relatively weaker. IDC has indicated that memory shortages and higher average selling prices could weigh on smartphone and PC shipments. ^[8] Automotive and industrial semiconductors still have long-term growth drivers, including electrification, automation, and energy efficiency. In the near term, however, these segments appear to be shaped by a combination of inventory normalization and gradual recovery.

4. Supply Bottlenecks: More Likely Specific Than Universal

In this cycle, supply constraints appear to be concentrated in specific high-value areas rather than across front-end wafer capacity as a whole. Key examples include HBM, CoWoS, 2.5D/3D advanced packaging, EUV equipment, and selected specialty gases and substrates.

TrendForce has argued that shortages in 2.5D/3D advanced packaging may persist through 2026. ^[9] Packaging capacity that combines AI GPUs with HBM is widely viewed as a critical bottleneck in AI semiconductor supply. This means that adding wafer capacity alone may not immediately resolve the supply constraints affecting AI chips.

ASML’s orders and backlog also point to continued demand for leading-edge capacity investment. ASML reported €13.2 billion in net bookings for the fourth quarter of 2025 and a year-end backlog of €38.8 billion.^[10] SEMI has also suggested that semiconductor equipment and 300mm fab equipment spending could remain at high levels beyond 2026.^[11]

Still, these bottlenecks should not be assumed to be permanent. As equipment investment and packaging expansion are converted into actual production capacity, some bottlenecks may begin to ease after 2027. At present, the more reasonable interpretation is that supply tightness is concentrated in AI-related areas, while future capacity additions and price adjustments should also be considered.

5. Price Increases: Both a Growth Driver and a Risk Factor

A key variable in the 2026 semiconductor market is the increase in memory prices. Gartner has noted sharp DRAM and NAND price increases in its 2026 market outlook, while TrendForce has forecast contract price increases for both DRAM and NAND in the second and third quarters of 2026. ^{[4] [5]} These price increases can support revenue and profitability for memory suppliers.

However, price increases are double-edged. For suppliers, they can improve revenue and margins. For device manufacturers and end consumers, they raise costs. In price-sensitive markets such as smartphones and PCs, higher memory prices can contribute to lower shipments or longer replacement cycles. ^[8]

Memory price inflation can therefore strengthen the current cycle, but it can also weaken it if prices rise beyond what end demand can absorb. If demand remains strong enough, higher prices favor suppliers. If demand cannot withstand the price pressure, higher prices can become a source of demand destruction.

6. Policy and Geopolitics: Opportunity and Uncertainty in Supply-Chain Reconfiguration

Semiconductors have become deeply exposed to policy and geopolitical forces. The U.S. CHIPS for America program supports domestic semiconductor manufacturing investment and supply-chain diversification, while the EU Chips Act aims to strengthen Europe’s semiconductor ecosystem. ^{[12] [13]}

Policy, however, does not only mean supply expansion. Subsidies can help build production capacity, but export controls and regionalization can reduce supply-chain efficiency and raise costs. In January 2026, BIS announced a policy change related to license review for advanced computing commodities, and in May 2025 it announced the rescission of the AI Diffusion Rule along with strengthened chip-related export controls.^{[14] [15]}

These policy shifts show that AI semiconductor supply chains are not governed by economics alone. Policy may improve long-term supply-chain resilience, but in the short term it can add uncertainty to corporate investment decisions and customer allocation plans.

7. Possible Scenarios

8. Overall Assessment

The current semiconductor cycle is strong, but it is not uniform. AI infrastructure and data-center investment are central drivers of market growth, while HBM, leading-edge process nodes, and advanced packaging appear to combine supply tightness with pricing power. By contrast, consumer-facing IT markets and some non-AI industrial segments may continue to face pricing pressure and inventory normalization.

Based on the evidence currently available, an AI-centered asymmetric expansion appears to be the most likely path. A broad-based supercycle remains possible, but it would require additional evidence that recovery is spreading into smartphones, PCs, automotive, industrial, and analog semiconductors. An earlier slowdown appears less likely at this stage, but remains a risk if sharp price increases and supply additions converge.

In short, the most cautious interpretation is that the semiconductor market is in a strong but uneven expansion. The key question is not whether one should be optimistic or pessimistic about semiconductors as a whole, but which areas of demand and supply are genuinely tight.

9. Limitations

This analysis is based on public materials from 2025–2026, including company investor relations materials, industry association data, market research firm publications, and government announcements. Many 2026 figures are forecasts rather than full-year actual results. Some sources also have limitations, including restricted access, dynamic pages, or company materials whose exact locations may be difficult to verify in detail.

This analysis does not assign numerical probabilities to the scenarios. Some earlier materials included probability-like framing, but those figures were not independently verifiable through public sources or an explicit quantitative model. For that reason, this post focuses on relative likelihood, necessary conditions, and key indicators to monitor rather than fixed probability estimates.