Wide moat where it matters — but the stock already prices the bull case.

• Two profit pools, one storefront. AWS earns 35.4% segment margin on a $364B contractual backlog; Amazon Ads owns 79.7% of US retail-media spend at ~80% gross margin. Roughly two-thirds of operating income comes from a quarter of revenue.
• Price already pays for it. Bull SOTP tops out at $2.77T using AWS at Microsoft-comparable 11× and ads at Meta-style 7–8×. Current EV is $3.07T — above the bull case before anything goes wrong.
• FCF margin is the worst in the peer set. 1.6% in FY25 — below Walmart at 2.1% and Costco at 2.9%. Tape sits at fresh all-time highs (last close $275.25, intraday high $276.26) on RSI 81 after a 24% one-month rally.

AWS reaccelerated to 28% — and was still the slowest of the three hyperscalers.

• Azure 40%, Google Cloud 63%, AWS 28%. Q1 FY26 is the first reporting cycle on the public record where AWS lags both rivals. Synergy Research describes AWS share as flat-to-eroding while peers grow 1.4–2.2× faster.
• The bull SOTP rests on the wrong comp. Multiple-convergence to Microsoft assumes AWS is the leader. If AWS converges with the hyperscaler median, the largest mechanical re-rate in the SOTP reverses — AMZN earns a laggard multiple, not a Microsoft one.
• The backlog and the AI-startup mark are one bet, not two. The April round was a contemporaneous $5B Amazon → AI startup + $100B AI startup → AWS, priced at the same valuation as the carrying mark. RPO concentration is being read as diversified hyperscale demand.

Roughly $40B of FY25 GAAP and CFO is non-recurring, non-cash, or working-capital-dependent.

• $15.2B AI-investment mark in Other Income. Non-cash Level-3 fair-value gain on preferred stock; carrying value $60.6B on $8B invested. Q1 FY26 booked another $16.8B. A 30% markdown reverses about $5B of GAAP earnings.
• $15–18B working-capital lift. Days payable outstanding stretched from 100 to 111 in one year. Reversion to 100 erases roughly the same dollars from cash flow in a single year.
• $10–12B cash-tax holiday. The 2025 tax act reinstated 100% bonus depreciation and immediate R&D expensing — cash taxes paid fell from $12.3B to $8.3B even as pretax rose 42%. "Specials" — severance, FTC, impairments — recurred in three of the last four years.

Free cash flow collapsed 77% as capex tripled the prior over-build peak.

Each dollar above depreciation bets that AWS revenue follows within 24 months. The 2014 and 2022 over-build cycles each produced 18-month margin troughs at a fraction of today's scale. The mechanical question is whether $300B+ of cumulative AI capex earns AWS's 30%+ margin before the depreciation lands. FY26 FCF likely turns negative; the balance sheet ($57B net cash) absorbs the build. The optionality cost of $443B of PP&E is the real risk.

Three dated events resolve 60–70% of the SOTP debate inside six months.

• July 30 — Q2 FY26 print. First post-guide-raise check on whether AWS 28% is durable. Q2 guide already runs 3–5% above pre-print Street; the bar is no longer low. AWS at 26%+ with margin at 35%+ keeps the multiple-convergence trade live; sub-25% with margin slipping forces a multi-quarter de-rate.
• Q2 / Q3 10-Q — Note 5 Level-3 rollforward. A $10B+ markdown to the AI-investment mark would reverse a chunk of GAAP earnings and undermine the cited AWS backlog growth. One filing decides three of the four variant-perception calls.
• October — FTC v. Amazon antitrust trial. Structural remedies (FBA separation, Prime unbundling) are on the table; 17 state AGs co-plaintiffs. Settlement signaling lifts the overhang; trial proceeding with structural remedies forces a different SOTP for the marketplace.