HScreen · AMZN

AMZN · Amazon.com, Inc. 7d old

conviction3

coverage tierT1-FULL

horizon12 months

Return distribution probability-weighted scenarios, never collapsed to a point

-25%24%

+14%50%

+40%26%

EV +11.40%

why these probabilities: BASE (+14%, ~299): AWS grows low-20s%, retail/advertising margin expansion continues on the 2026Q1 North America trajectory, EV/EBITDA holds ~15x on a larger EBITDA base. Derived from earnings-power growth at a flat multiple, not from any broker mark — no dated street target for AMZN exists in this pack. BULL (+40%, ~368): AWS reaccelerates through mid-20s%, the Bedrock/Anthropic mix shows the operating leverage the third-party work argues for, capex duration questions quiet as AI revenue visibly covers depreciation, and the multiple re-rates toward 18x. WOULD RAISE the bull probability ABOVE 26%: a disclosed AWS backlog/RPO acceleration, a Trainium volume disclosure that lowers unit cost of serving, or revenue revision breadth turning positive from -0.67 — none of which are in the five-year history the base rate is drawn from. WOULD CUT it: continued negative revenue revisions alongside up-EPS revisions (margin-led, volume-poor earnings are lower quality and re-rate less), or evidence that hyperscaler capex commitments are being met without incremental cloud revenue. BEAR (-25%, ~197, near the 52-week low of 196): the falsifier fires — AWS decelerates below the mid-teens while depreciation from the committed capex lands in the P&L, so EPS is cut on both lines at once and the multiple compresses to ~12x on a payback re-rating rather than a cyclical one. This is not a generic de

not built — options-implied distribution on the same axis, with the KL number and where the disagreement sits
needs: desk-options implied bands (v1.4 Part 4a) — activates at 60% coverage; currently 33%

Visibility how far ahead this name can be seen

not built — visibility score
needs: visibility_score.py

Trust how much weight this name's own record can carry (credibility Z)

not built — credibility
needs: credibility.py

Sizing survival-first size (half-Kelly) vs the enforced ladder

kelly not computable — no data

Payout durability does the distribution survive a downturn?

PASS

DPS trajectoryinsufficient_history

interest cover now35.17x

interest cover at FORWARD rates11.17x

Falsifiers what would prove this wrong

AWS revenue growth holds at or above the high-teens percentage year over year
AWS segment operating margin does not fall below the low-30s on a sustained basis as depreciation from the current capex cycle lands
The AI capex is funded by rising cloud revenue rather than by multiple-supported spending
Upstream token economics continue to support tier-2 capex — token spend rising with cost per token falling
The revenue revision tape stops deteriorating
Memory and energy cost inflation do not compress AWS unit economics

Conflicts surfaced, never netted out (T4)

Our base case rests on AWS volume growth in the low-20s, but the revenue revision tape is negative and broad: breadth -0.67, 10 estimate cuts against 2 raises over 365 days, median move -1.10%. The street is cutting the	high
The two tapes disagree with each other, which is itself the finding. Up-EPS with down-revenue is margin-led earnings quality — cost discipline and mix rather than demand. If that is what is happening, the earnings deserv	high
No options-implied bear probability exists for AMZN in this pack (null, no chain entry), so the strongest available anchor — a dated daily market price for exactly this question — is missing. Our p_bear is reconciled onl	medium
The data provider classifies AMZN as Consumer Cyclical / Specialty Retail while our thesis is an AI-infrastructure thesis. Any screen, factor model or peer comparison built on the provider taxonomy will place this name a	medium
The pack's upstream demand evidence (model-lab ARR, token economics, hyperscaler capex commitments) establishes that the money exists. It does NOT establish that it comes to AWS rather than Azure or GCP, and the pack ret	high

The case against independent adversary, different model family (P6) broken

Version history from the research record

archived versions1

archived	hash	bytes
2026-08-17T14:07:30	64794ae6e36a	25,841

One version only — the archive was created today. The diff view fills in on the next re-run.

Rendered from canonical artifacts only (T1) — no terminal database exists. Every figure is read from the system's own files at render time; where an artifact is missing the panel names the dependency rather than showing an empty box. No execution path exists anywhere in this interface (T5), not even a link.