Global semiconductor shortage and the AI-era semiconductor cycle: Nomura’s Chip Shortage Index signals a deep, structurally persistent semiconductor shortage

Overview

This macro-focused semiconductor report introduces Nomura’s Chip Shortage Index (CSi) to track the balance between chip demand and deliverable supply. Nomura concludes that the market remains in a boom-stage shortage regime and that the current AI-led cycle has more durable price support than prior industry-led cycles.

Core views

Nomura argues that chip availability has become a macroeconomic constraint rather than simply a response to the global cycle. AI capital spending is now large enough that physical chip availability limits how much planned investment can be executed. The report highlights Korea and Taiwan because semiconductor scarcity affects exports, income, liquidity and broader business-cycle momentum there. It notes Bloomberg estimates that four US big-tech firms’ capital spending could reach USD1.1trn in 2027, above estimated US defence spending of USD901bn, while hyperscaler capital spending should exceed USD700bn this year. The report attributes the present tightness to multi-year AI demand meeting a supply curve that has become slow and uneven. Inference use rises with utilisation of infrastructure built under multi-year plans, making demand less sensitive to higher chip prices than a conventional consumer-electronics replacement cycle. On the supply side, further miniaturisation requires costly EUV tools, low initial yields and longer yield-improvement periods. The memory bottleneck is especially acute: HBM consumes several times as much silicon area per functional bit as commodity DRAM, and its stacking and through-silicon-via processes have lower yields. Converting fab lines to HBM therefore removes conventional-memory capacity. Nomura reports that memory-chip export prices rose by an average 135% year on year in H1 2026. New capacity takes two to three years from commitment and then requires a yield ramp, arriving in large increments rather than smoothly. Nomura stresses that structural scarcity raises the average price level and the amplitude of cycles but does not eliminate boom-bust dynamics. Long lead times can induce producers to expand together after high prices, creating future oversupply once capacity arrives; customers can also overorder during shortages and cancel demand as availability improves. The report contrasts this structural trend with the current cyclical phase, noting that past upcycles still ended when capacity, logistics or inventories normalised. The Nomura CSi is designed as a monthly real-time measure of chip scarcity. It is centred on 100: readings above 100 indicate shortage and readings below 100 indicate supply slack. Nomura derives separate demand and supply factors from roughly 20–25 macroeconomic and industry series using a dynamic factor model, then uses a three-variable vector autoregression to estimate how those factors affect semiconductor export prices. Those price responses determine the index weights, following the logic of a financial-conditions index. Semiconductor export prices are the model’s anchor because they transmit the chip cycle into Korea’s terms of trade, trade surplus, corporate cash flow and liquidity. The index spans 2012–2026; its lowest sample reading was 98.45 in June 2023. The current July reading of 103.8 is close to the sample high and places the CSi in the boom stage. Small recent monthly moves do not confirm a turn. Nomura considers a near-term move below 100 highly unlikely given multi-year AI demand and long capacity lead times. A decline while the index remains above 100 would mean shortage pressure is easing, not that the market has entered supply slack. The report’s CSi clock defines boom as an above-100 and rising trend, slowdown as above-100 and falling, recovery as below-100 and rising, and recession as below-100 and falling. Nomura’s historical decomposition separates chip-price moves into CSi shocks—changes in the demand-versus-deliverable-supply balance—and own-price shocks reflecting industry dynamics such as inventories, production cuts, allocation-related double-ordering, pricing and product-specific developments. The 2012–14 and 2016–18 upcycles were mainly industry-led; the 2020–21 episode was shortage-led but reversed as logistics normalised, demand softened and inventories built. The 2024–25 price recovery was again driven by production discipline, inventory normalisation and producer pricing while broad-market CSi conditions remained slack. Since Q4 2025, Nomura finds that AI and HBM demand have become sufficiently wafer-intensive to tighten deliverable supply across a broader market share. Both decomposition components are positive, but the shortage contribution has risen more sharply and is now the larger driver; supply-chain inventories are near historical lows. Unlike the pandemic episode, relief now depends on construction and successful yield ramps rather than normalising logistics. The report’s impulse responses show that a CSi shock persists longer than an own-price shock and can sustain higher chip prices for more than the average 24-month upcycle. Because the current price upturn began in Q4 2025, Nomura’s model points to continued gains through 2027 and beyond, provided the shortage remains the principal driver.

Analysis framework

Nomura first explains the AI-demand and supply-bottleneck mechanisms, then builds a monthly shortage indicator from common demand and supply factors. It uses export-price responses in VAR models to weight the index and to compare the current cycle with earlier semiconductor upcycles, separating broad scarcity effects from industry-specific price momentum.

Methodology notes

• Industry AnalysisSupply-demand framework

  Demand versus deliverable semiconductor supply

  The report frames scarcity as the gap between AI-driven demand and supply that can actually be delivered after capacity and yield constraints.

• Other

  Dynamic factor model and vector autoregression (VAR)

  Nomura extracts common demand and supply signals from roughly 20–25 series, then estimates their relationships with semiconductor export prices to weight the CSi and assess shock persistence.

• Cycle and Business ConditionsBusiness-Cycle Inflection Analysis

  Nomura CSi clock

  The index level relative to 100 and its monthly direction classify the chip market into recovery, boom, slowdown and recession stages.

Key data

• Nomura CSi, July 2026103.8Near the sample high; above 100 signals a chip shortage and Nomura classifies the reading as boom stage.
• Memory-chip export prices135% y-o-y on average in H1 2026Cited as evidence of the memory-wall bottleneck and tight supply.
• Current price upturn startQ4 2025Nomura’s model projects gains through 2027 and beyond if shortage remains the main driver.
• New capacity lead timeTwo to three years from commitment, plus yield rampExplains why supply relief is slow.
• Average chip upcycle24 monthsNomura finds a CSi shock can sustain higher prices for longer than the average upcycle.
• CSi sample low98.45 in June 2023Lowest reading in the 2012–2026 sample, indicating greatest supply slack.

Impact & implications

Nomura sees semiconductor scarcity as an increasingly important driver of Korean and Taiwanese business cycles and a source of sustained price pressure in the current AI-led chip cycle. The report distinguishes persistent structural tightness from a permanent end to cyclicality: eventual capacity additions and inventory adjustment can still reverse the market.

Risks

• A persistent bottleneck does not prevent an eventual downturn; high prices can trigger capacity expansion that later creates oversupply.
• Customer overordering during shortages can create phantom demand that is cancelled when scarcity eases.
• The 2020–21 shortage reversed as logistics normalised, end-demand weakened and inventories accumulated.
• The projection through 2027 and beyond depends on the shortage remaining the principal driver of the cycle.

What to watch

• Monthly Nomura CSi updates, especially whether the index remains above the 100 shortage threshold and whether its direction changes persistently.
• Demand and supply components of the CSi to identify whether AI demand or capacity conditions are driving changes.
• Capacity construction and yield-ramp progress, which determine durable supply relief.
• Supply-chain inventory levels and semiconductor export prices.
• AI and HBM demand, including big-tech capital-spending expectations and inference usage.