Sector deep dive: semiconductors

Definition

Semiconductor analysis covers the industry’s distinct layers — fabless designers, foundries, integrated device makers, equipment suppliers, EDA/IP vendors, and memory — each with different economics riding one of the most CYCLICAL demand patterns in technology. The sector’s analytical signature is the inventory cycle: long capacity lead times meet fast-moving demand, producing recurring boom-shortage-glut-bust sequences (documented across decades of SIA data), all overlaid since the 2020s by AI-driven capex waves and export-control geopolitics. Chips are simultaneously deep-cyclical (memory) and near-monopoly toll roads (EDA, lithography) — layer identification comes before any metric.

How it works / structure

• The value-chain layers (engine-relevant): FABLESS (design margin, TSMC-dependent), FOUNDRY (capex-intensive scale monopolies — leading-edge concentration into effectively one node supplier is the sector’s defining structural fact), EQUIPMENT (sells the cycle’s capex — orders lead the cycle), MEMORY (commodity pricing, the purest boom-bust), EDA/IP (subscription toll roads — fa-moat-analysis switching costs at their strongest), ANALOG (long product lives, autos/ industrial exposure, the “slow” corner).
• The cycle’s anatomy: demand signal → double- ordering through the supply chain → capacity additions on 2-3 year lead times → demand cools as supply lands → inventory correction (documented recurring sequence; the 2021 shortage → 2023 glut is the latest full cycle); INVENTORY DAYS across the chain and equipment BOOK-TO-BILL are the cycle’s gauges.
• The metrics stack: utilization rates (foundry margins swing on it), wafer starts, ASP trends (memory spot vs contract pricing), design-win pipelines (fabless revenue visibility), capex-to- revenue ratios (cycle-phase tell), and fa-guidance-estimates mechanics — semis guidance misses cluster at cycle turns, and the stocks historically BOTTOM while estimates still fall (documented pattern, the sector’s classic trap for estimate-followers).
• The structural overlays: AI capex concentration (a demand vertical large enough to decouple leaders from the classic cycle — open question, labeled), export controls and CHIPS-Act reshoring (policy as a direct revenue variable), and Taiwan concentration (a genuine tail risk priced nowhere continuously).

When it applies

Cycle positioning (semis lead broad tech and often the macro-business-cycle — the SOX/SMH relative strength is a standard risk-appetite gauge); growth allocation (style-growth-investing concentration in the sector’s compounders); equipment names as capex-cycle expressions; memory as the deep-cyclical trade (book-value anchored entries, practitioner convention labeled).

Risk profile & failure modes

• Cycle mistiming via estimates (the signature trap): buying on “cheap” P/Es at peak earnings and selling on “expensive” P/Es at troughs inverts the sector’s documented pattern — cyclical valuation requires normalized earnings, not spot multiples.
• Double-order illusion: shortage-era demand contains phantom orders that evaporate at the turn — backlog quality is unknowable in real time (documented in every cycle post-mortem).
• Concentration risks stacked: index-level AI concentration + single-foundry dependence + geographic tail — the sector’s diversification is thinner than its ticker count suggests (risk-correlation-exposure).
• Policy discontinuities: export-control announcements have repriced individual names 10-20% in a session — a standing event class with no calendar (event-litigation-regulatory lineage).

Evidence & limits

SIA annual reports document industry structure and the cycle record; the layers’ economics are public financial history. The AI-decoupling question and policy trajectory are open — entries carry them as labeled uncertainties with both-sided scenarios.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Memory names bought below 1.2x book during inventory corrections outperform the SOX over the following 2 years (deep-cycle entry convention)” — falsified by the cohort spread.
• “Equipment book-to-bill falling below 0.9 precedes foundry-capex guidance cuts within 2 quarters (cycle-sequence check)” — falsified by the guidance record.

Cross-references

• The framework entries: fa-sector-metrics, fa-moat-analysis, fa-guidance-estimates
• The cycle frame: macro-business-cycle; the style consumer: style-growth-investing
• The rotation gauge: strategy-sector-rotation