US Semiconductor Value Chain · Investment Map

US Semiconductors in One Map:
An Investment Decision Map from EDA to the Cloud

Mapping 20 leaders along "upstream tools → midstream manufacturing → downstream demand": each company's core products, moat, investment thesis, key metrics to track, and main risks, alongside ETF tools and a signal calendar — meant to support, not replace, your own decisions.

Data as of: early June 2026 · Market figures are approximate; verify live data before placing any order
Top-5 hyperscaler 2026E capex
≈$700B
Analyst consensus; the funding source for the whole chain
TSMC 2026 capex guidance
$52–56B
Record high; 9 new fabs this year
NVIDIA visible order revenue
>$500B
Blackwell+Rubin over the next five quarters
Supply shortage expected to persist until
Beyond 2027
Per TSMC management; chain-wide shortage
2030 global semiconductor market
$1.5T
TSMC forecast; AI/HPC ~55%
01

Value Chain Map: Product Flows Right, Money Flows Left

The key to reading this chain is two directions: the product flow moves from upstream to downstream (tools → wafers → chips → compute); the money flow runs the other way — the capex of downstream hyperscalers becomes midstream orders, which become upstream equipment and software revenue. So to gauge the cycle, look at the money on the far right first. Click any node and the company list below filters automatically.

Upstream · Tools & Capacity SupplyMidstream · Design & ManufacturingDownstream · Demand & Monetization
UPSTREAM UPSTREAM
→
MIDSTREAM MIDSTREAM
→
DOWNSTREAM DOWNSTREAM
Product flow → tools / equipment / wafers / chips / compute ← Money flow: capex / orders / procurement (leading indicator of the cycle)
02

Leading Company Profiles

Moat ratings are qualitative judgments (5-point scale: a composite of monopoly/duopoly structure, switching costs, technology generation gap, customer lock-in, and margin durability). Click a card for the full profile: investment thesis, key metrics, main risks, and corresponding ETF exposure.
03

ETF Tools: How to Get Sector Beta Without Stock-Picking

Figure out your total exposure before adding more: semiconductors already make up ~33% of the Nasdaq 100, so holding QQQ already implies heavy semi exposure. The core difference among these four tools is "concentration" — SMH bets on the leaders, equal-weight XSD bets on breadth.

Ticker Tracked Index Expense Ratio Structure Best For
SMH MarketVector US Listed Semiconductor 25 Index 0.35% Highly concentrated: NVIDIA + TSMC often exceed 30% combined; includes ADRs Bullish on "winner-take-all" leadership; willing to bear concentration risk
SOXX NYSE/ICE Semiconductor Index (Philadelphia lineage) 0.35% ~30 holdings, single-name cap ~8%, more diversified than SMH Want sector beta but with single-company risk controlled
XSD S&P Semiconductor Select Industry Index (equal weight) 0.35% Equal-weight, higher small/mid-cap share, dilutes NVIDIA's impact Betting the rally broadens from leaders to second-tier names
SOXLLeveraged 3x daily leverage on the ICE Semiconductor Index ≈0.75% Daily-compounded rebalancing; significant volatility drag in choppy markets Short-term trading only; not suitable for long-term holding
04

Decision Signal Calendar: What to Watch, and When

On this chain, "someone else's earnings report is your research note." Ordered by importance — changes in the first two signals get priced across the entire value chain at once.

Quarterly · end of Jan/Apr/Jul/Oct
Capex guidance from the four big hyperscalers
The capex guidance in earnings from Microsoft, Google, Meta, and Amazon is the chain's single most important leading indicator — raise it and the whole chain rises; mention "discipline/slowdown" and the whole chain falls. Also watch AI monetization progress, which determines how sustainable the spending is.
Quarterly · late Feb/May/Aug/Nov
NVIDIA earnings and conference call
The market's AI bellwether. The point isn't whether revenue beats (it likely does) but order visibility, margin trajectory, and the hyperscaler share of revenue — the February report beat yet the stock still fell 5.46% the next day, because the market trades on spending sustainability.
Around the 10th monthly + quarterly call
TSMC monthly revenue and earnings call
The only node that reports monthly — the highest-frequency window into industry-wide health. Capex and AI-business CAGR guidance on the call (now raised to the mid-to-high 50% range) directly price equipment stocks.
Monthly · start/mid-month
Memory contract and spot prices
DRAM/NAND/HBM prices are the lifeline of the memory segment. Currently 60–70% of server-grade DDR5 is locked under 3–5 year hyperscaler contracts; watch whether the long-contract share can truly smooth out the cycle.
Ad hoc · policy windows
US export controls and tariffs on China
Directly affects equipment stocks (China revenue share) and NVIDIA's China-business assumptions. NVIDIA's current guidance already excludes China data-center revenue — any easing would be an upside surprise.
Daily
SOX relative strength vs Nasdaq + rate path
A weakening SOX/Nasdaq ratio often leads fundamental turning points; high-valuation segments (EDA, design) are sensitive to Treasury yields, and a rate-cut cycle is a tailwind for the valuation denominator.
05

Risk Matrix: Systemic vs. Segment-Specific

Systemic risks hit every segment and can't be diversified away within the chain; segment-specific risks can be hedged through position structure (e.g., holding both the GPU and ASIC chains, or pairing design with equipment, or controlling cycle exposure via memory).

Systemic Risks (Chain-Wide Resonance)

  1. AI capex peaking or monetization falling short — Wall Street's biggest current debate. If AI spending fails to convert into revenue, hyperscalers cutting budgets would amplify the shock step by step down "orders → capacity → equipment."
  2. Valuation and rates — the sector's overall valuation is at historic highs, doubly sensitive to growth downgrades and rising rates; the SOX dropping over 5% in a single day on May 12 is a sample of risk appetite reversing under stretched valuations.
  3. Geopolitics — the Taiwan Strait situation is a tail risk for TSMC and the entire supply chain; escalating export controls would hit China revenue in both equipment and design at once.
  4. Power and infrastructure bottlenecks — US data-center power demand is expected to double within two years, and power supply may in turn constrain the pace of AI compute deployment.

Segment-Specific Risks (Structurally Hedgeable)

  1. Memory price reversal — the most cyclical segment, with its "tech commodity" nature intact; long contracts cushion but can't eliminate the cycle, and downside margin leverage is just as large when prices turn.
  2. GPU vs. ASIC share battle — hyperscaler in-house chips (TPU/Trainium/MTIA) eroding general-purpose GPU share is bad for NVIDIA and good for Broadcom/Marvell, and vice versa.
  3. Foundry overcapacity (post-2027) — TSMC's 9 new fabs plus Intel and Samsung expansions are concentrated in post-2027 ramp; if demand growth cools, utilization and price pressure emerge.
  4. Individual company execution risk — Intel 18A and foundry customer qualification, AMD's software ecosystem catch-up, Arm's long-term RISC-V substitution, etc., are stock-level rather than industry-level issues.