Why the Mag 7 Lost $950B in One Week

Alphabet disclosed $13.9 billion in Q4 capex and guided 2026 capex to $175–185 billion (up from $91 billion in 2025), and Amazon raised 2026 capex to about $200 billion, contributing to roughly $710 billion of projected hyperscaler spending in 2026 (~$2 billion/day). The guidance shock sent S&P 500 software & services down (about $1 trillion erased) and the Nasdaq down roughly 4% before a Friday rebound, while large-cap AI names collectively lost as much as ~$950 billion in market value at one point. The note argues markets are transitioning from Stage 1 (hyperscaler winners) to Stage 2 (AI infrastructure and supply-chain beneficiaries such as networking, power, memory and metals), while warning legacy “KIDS” software/data businesses face durable margin and valuation pressure.

Analysis

Market structure: The hyperscaler capex shock reallocates value from Stage‑1 mega caps into Stage‑2 “picks and shovels” — servers, networking, memory, power, and raw materials. Expect sustained 2026 demand: $710B+ capex implies multi‑year revenue tailwinds for ANET, AVGO, ETN and select metal suppliers, while legacy KIDS software (CRM, NOW, MORN, EFX) faces margin compression as AI replaces high‑margin knowledge work.

Risk assessment: Key tail risks are ROI failure (hyperscalers cut new builds if utilization <60% vs models), regulatory export controls on advanced chips, and an oversupply cycle in datacenter components that could depress pricing within 12–24 months. Short horizon (days–weeks) volatility will remain high around earnings and capex updates; structurally (quarters–years) winners are those with sticky enterprise contracts and differentiated hardware/software stacks.

Trade implications: Favor long exposure to infrastructure suppliers with 6–18 month time horizons and conservative sizing (1–3% per name), financed with option collars or call spreads to manage volatility. Implement pair trades that long ANET/AVGO vs short KIDS names (CRM/NOW) to express thematic rotation while neutralizing beta. Commodities (copper, silver) and utilities (VST, ETN) are tactical hedges for rising power/metal demand; bonds may underperform if fiscal/capex drives higher corporate borrowing and medium‑term yields rise.

Contrarian angles: Consensus worries about hyperscaler ROI may be overdone — if utilization and monetization improve, mega caps can re‑rate, creating squeeze risk for Stage‑2 longs that are already priced for perfection. Conversely, some Stage‑2 names may be priced ahead of contracted revenue; watch orderbooks and backlog disclosures. Historical parallel: 2016–18 cloud buildout rewarded infrastructure suppliers but punished undifferentiated software; outcomes hinge on multi‑quarter visibility and supply constraints.

Key Decisions for Investors

• Establish a 2–3% net long position split 50/50 in Arista Networks (ANET) and Broadcom (AVGO) using 9–15 month call spreads (buy 12‑month ATM call, sell 24‑month +20% strike) to limit cost; target 40–80% upside, trim at +50%, stop‑loss at −30%.
• Add a 1–2% tactical utility/power exposure in Eaton (ETN) and Vistra (VST) (equal weight) via outright shares or 12‑month call spreads to capture energy demand from data centers; hold 6–18 months and reassess after next two quarterly earnings/utility demand reports.
• Initiate a dollar‑neutral pair trade: long ANET (1.5% notional) vs short Salesforce (CRM) (1.5% notional) — prefer buying ANET calls and CRM 6–9 month puts for asymmetric payoff; close if spread narrows/widens by 15% or after CRM/ANET next earnings.
• Reduce direct exposure to KIDS names by 30–50% in software-heavy portfolios and implement hedges: buy 6–9 month puts on NOW/CRM sized to cover reduced positions (0.5–1% portfolio risk). If hyperscaler capex guidance is reduced >10% in a quarter, tighten stops and rotate 50% of realized proceeds into undervalued Stage‑2 names.
• Monitor regulatory/capex catalysts over next 30–60 days: track GOOG/AMZN capex revisions, DOJ/FTC AI investigations, and export‑control headlines. If export controls expand or capex revisions fall >10%, close longs in semiconductor equipment and increase cash/hedges by 2–4% within 5 trading days.