Verdict: KEEP | Mega-cap anchor #4 (4.53%) | Informational only — no active bet

Factor Profile

iev regress AMZN (250d trailing)

SPY     β = +1.71    65.6% of variance
MTUM    β = -0.36   (anti-momentum)
Idio                 48.5% of variance

α = -11.2% annualized    σ_idio = 24.3%    R² = 51.5%

This is half a market bet. 48.5% idio is well below the 75% target — SPY alone explains two-thirds of variance. β=1.71 means every dollar of AMZN weight delivers ≈$1.70 of effective market exposure. Anti-momentum loading confirms the stock is in a downtrend. Factor profile is nearly identical to TSLA (β=1.97, idio=52%). Both are high-beta QQQ proxies, not stock picks.

Trailing α = -11.2%. Negative. But driven by capex-narrative punishment, not fundamental deterioration. The $200B 2026 capex guide (Feb 5 earnings call) spooked investors — stock fell ≈8% post-earnings. Q4 EPS essentially flat ($1.95 vs consensus) despite revenue +10% and AWS +19.7%. The market is punishing investment intensity across hyperscalers.

Three Engines

Source: 10-K Note 10 (filed Feb 6, 2026).

The Depreciation Story

This is the margin story for the next 2-3 years.

FCF collapsed 71% in FY2025. At $200B capex, FCF goes deeply negative in 2026 — which is why they raised $52.6B in new debt in March ($37.0B USD on 3/13, EUR 14.5B on 3/16). Total debt heading to ≈$118.6B from $68.8B at year-end. Interest doubles from $2.3B to ≈$4.7B.

Management SHORTENED server useful lives from 6 to 5 years effective Jan 1, 2025, adding $1.4B D&A and reducing net income by $1.0B ($0.10/share). This is conservative accounting — the opposite of extending lives to flatter margins. It signals management prioritizes economic reality, but means the depreciation headwind is structurally steeper than peers.

AWS operating margin already compressed from 37.0% to 35.4% (FY2024 → FY2025). At $200B capex hitting 5-year lives, expect $90-100B annual D&A by FY2027.

Tax offset the market underweights: The 2025 Tax Act gives 100% accelerated depreciation + immediate R&D expensing. For $200B capex, the cash tax shield is ≈$10-17B annually (timing benefit at 21% on the difference between full-year tax deduction and GAAP year-1 D&A). GAAP P/E of 25.1x overstates valuation — cash P/E is closer to 21-22x. This is a timing benefit (reverses over years 2-5), not permanent value, but it means the bearish "25x on 11% growth" framing is modestly misleading.

Where Advertising Hides the Truth

Amazon doesn't report advertising as a segment. At industry retail media margins (50-60% operating), advertising generates $34-41B of operating income — nearly ALL of the $34.4B non-AWS operating income (NA $29.6B + International $4.8B). This means everything else in retail (online stores, physical stores, subscriptions, fulfillment) is collectively break-even or loss-making.

Two businesses generate essentially all profit: AWS (35.4% margin, $45.6B OI) + Advertising (≈55% margin, ≈$38B OI). The rest is customer acquisition infrastructure. The quality of earnings is higher than segment reporting shows.

Consensus

Estimates (as of March 28, 2026):

FY2026 EPS range: $6.34-$11.34 (78% spread). The Street doesn't know the margin trajectory any better than we do.

Ratings: 96% Buy. Zero sells. One downgrade — DA Davidson to Neutral/$175 on "AWS losing its lead" (growth rate comparison that ignores absolute dollar growth). Two holds crept in post-Q4 miss. Mean PT $280-287, implying +40% upside.

Beat cadence collapsed: +$0.34 → +$0.21 → +$0.37 → +$0.38 → -$0.02 over 5 quarters. Street recalibrated upward and overshot. Consensus Q1 revenue at $180.6B sits $2.1B above company guide ceiling ($178.5B) — pricing in a beat. If AMZN merely meets its own guide, that's a revenue miss.

Valuation (corrected):

25x on a decelerating year is not cheap. The value proposition is 21x on FY2027's acceleration — a call option on the capex cycle delivering operating leverage. PEG 0.96x on FY2027 is genuinely attractive IF it materializes.

Mispricing Analysis

Five candidates examined. None actionable.

1. GAAP vs cash divergence (tax shields). Cash P/E ≈21-22x vs GAAP 25.1x. Real but known to institutional investors. Timing benefit, not permanent. ≈10-16% P/E discount.

2. Advertising profit quality. SOTP suggests ≈6% undervaluation ($2,273B adjusted vs $2,140B market cap). Known structural argument — every sell-side model runs SOTP.

3. AWS absolute dollar growth. AWS adds more incremental dollars ($21.1B) than Google Cloud ($16B). The "losing its lead" narrative is a growth rate fallacy on the largest base. Modest impact (≈1.9%) if market reframes. Known to careful analysts.

4. Consensus above guide — Q1 miss setup. Tactically relevant but works AGAINST us (we're long QQQ). Consensus revenue $2.1B above guide high. Combined with collapsed beat cadence, the setup favors disappointment. Can't act — mega-cap anchor constraint.

5. Negative alpha mean reversion. -11.2% trailing alpha is either temporary (capex narrative, reverses when cycle delivers) or permanent (structural deceleration, 31% → 11% EPS growth). Evidence is mixed — genuinely uncertain. The market is pricing this uncertainty correctly at 25.1x/20.7x.

Net: No mispricing we have edge on. $2.14T market cap, 50+ analysts, R² = 51.5%. Every datapoint in this memo is public and priced. The stock is in a genuine doorway state between "capex builds a moat" and "capex destroys returns." Market is pricing ≈55-60% bull / 40-45% bear. That seems approximately right.

Why Not Remove

1. Weight. 4.53% = ≈$45K short. At β=1.71, that's ≈$77K effective SPY exposure removed. The filtration strategy can't absorb this much single-name beta risk.
2. Mega-cap anchor. Top 10 QQQ weight. Structural hold by design.
3. No idio thesis for underperformance. Negative trailing alpha is factor-driven (capex narrative), not company-specific. Fundamentals accelerating (AWS backlog $244B +40%, advertising +22%, OI +17% ex-charges).
4. Edge: zero. High conviction ≠ edge. We know everything the market knows and nothing more.

Catalysts in Window

Q1 financials pre-date the April 2 tariff deadline — the Q1 numbers should be clean. The real risk is Q2 guidance with tariff uncertainty baked in. The $5B-wide OI guidance range ($16.5-21.5B) for Q1 already signals management uncertainty.

Balance Sheet Snapshot (Post-March 2026 Issuance)

Interest at ≈$4.7B is still <6% of operating income. No financial covenants. Maturity schedule is long-dated (14.1yr weighted average on existing, new tranches out to 2076). The balance sheet is leveraging but not stressed.

Summary

AMZN is a $2.14T market-beta amplifier that happens to run three excellent businesses. The stock is in a genuine doorway state: 25.1x FY2026 (decelerating) with a call option on 20.7x FY2027 (accelerating). No mispricing identified that we have edge on. The most underappreciated feature is the GAAP vs cash divergence from tax shields (cash P/E ≈21-22x vs headline 25.1x), but this is a timing benefit known to institutional investors.

For the basket: benchmark weight is correct. Zero edge = market roughly right = hold at index weight. Monitor Q1 earnings April 30 for QQQ downside exposure.

Sources: AMZN 10-K FY2025 (filed Feb 6, 2026), 8-Ks (Feb 27, Mar 13, Mar 16), Q4 2025 earnings transcript (Feb 5), worldview (24 evidence items), StockAnalysis.com, MarketBeat, yfinance.