HScreen · MSFT

MSFT · Microsoft Corporation 9d old

conviction3 → 3.00 trust-weighted

coverage tierT1-FULL

horizon12 months

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

-23%24%

+12%48%

+45%28%

EV +12.84%

why these probabilities: PRICE BASIS USD 495.40 at 2026-08-17 (fmp:/stable/profile?symbol=MSFT); 12-month horizon; total returns include the ~0.73% indicated dividend yield (USD 0.91 quarterly). All three legs are built forward - growth rate, then EPS, then the multiple that growth earns - and are defined RELATIVE TO WHAT THE MARKET EXPECTS. BASE (48%) = consensus DELIVERED, and it is anchored TO consensus, not competed with: FY2027 revenue growth +18.2% and EPS growth +15.4% (fmp:/stable/analyst-estimates?symbol=MSFT&period=annual, as_of 2026-08-17, 36 revenue / 30 EPS contributors, revenue dispersion only 3.6%), FY2028 EPS growth +18.5%. Rolling the multiple forward one year on NTM EPS of ~24.6 and allowing MODEST compression from 23.8x to ~22.5x for rising capex intensity and the depreciation step-up gives ~USD 553, about +12% total. Cross-check: the street's own consensus target of 547.74 implies +10.6% price, +11.3% total - so the base leg sits within 1pp of the market's own mark, which is the point of the leg. BULL (28%) = revenue growth ~24% versus consensus 18.2%, on named evidence: management stating demand exceeds available supply in a relatively extreme moment, the CY2026 investment plan UNCHANGED at ~USD 220bn once the ~USD 15bn guidance cut is correctly read as lease classification and presentation rather than reduced data-centre construction or server procurement, Copilot sales tracking a

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

0.70 scored on 3/4 components

horizon1.000

dispersion_tightness0.550

delivery_reliabilityabsent — guidance_vs_delivery returned nothing

tape_stability0.540

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

guidance_deliveryunavailable

estimates_depth0.333

evidence_depth0.286

composite0.309

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

ladder (enforced)3%

kelly-derived (report-only)0.07%

divergence2.93pp

inputsEV(ann) 12.84% · σ 24.46 · Z 0.3095

Payout durability does the distribution survive a downturn?

PASS

DPS trajectoryprogressive

interest cover now50.88x

interest cover at FORWARD rates25.7x

Falsifiers what would prove this wrong

Azure growth remains above 20% constant currency through FY2027 as capacity lands
The ~USD 15bn FY guidance reduction is lease classification and presentation only, with the CY2026 investment plan unchanged at ~USD 220bn
Microsoft Cloud gross margin percentage compression stays contained as AI infrastructure scales
Copilot/M365 commercial cloud retains pricing on default enterprise deployment
The EPS revision tape stops deteriorating
Enterprise token/inference spend keeps growing, so tier-2 capex has a payer behind it
Hyperscaler rivals do not take share of the same enterprise AI budget via own silicon

Conflicts surfaced, never netted out (T4)

I score growth 8 and weight the bull leg at 28%, above the pooled base rate, while the EPS revision tape is DOWN: breadth -0.294, 11 estimates down against 6 up over 365d, median move -1.38%, velocity 1.4 moves/30d. Reve	HIGH - this is the single strongest piece of evidence against my bull weighting and it is the reason management_delivery is capped at 8 and conviction at 3 rather than 4.
My probability-weighted expectation is +12.84% against the street's +10.57% implied by a 547.74 consensus target. The direction agrees but the shape does not: I hold a 24% bear leg at -23% while 82 covering analysts publ	MEDIUM - agreement on level, disagreement on distribution. Wide targets over tight revenue dispersion means the street disagrees about the MULTIPLE, not the forecast, which is exactly where my bear leg sits.
I score quality 8 on scale and enterprise distribution, but the pack's own expert evidence states Copilot's productivity-layer moat comes from default deployment, enterprise IT approval and LOW SWITCHING COSTS, and that	HIGH - if the application layer is genuinely low-switching-cost then quality 8 is too high and the bull re-rating leg is weaker. Unresolved, and it is why quality is not 9-10.
The bear leg carries 24% while the only market-priced view available is an ATM 96d chain_iv_median of 31.09 with no strike ladder, so P(-23%) cannot be read from a price. The strongest available anchor class is therefore	MEDIUM - not a disagreement but an unfalsifiable position: the anchor that would discipline the bear leg most is unavailable, so the leg rests on a disclosed mechanism rather than a market price.
The pack retrieved NO substitution evidence at all on GOOGL/AMZN/ORCL/CRWV taking share of the same enterprise AI budget, including own-silicon insourcing funded from the very capex that appears as demand evidence. Absen	HIGH - this is the main unhedged risk to the growth score. Recorded as a gap rather than resolved.

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

Version history from the research record

archived versions5

archived	hash	bytes
2026-08-17T10:34:17	303811d1bb7c	33,954
2026-08-17T14:07:30	faefb6eb2144	38,613
2026-08-17T14:25:34	53f69b1fe963	31,404
2026-08-17T15:40:00	4ae029f0c550	31,613
2026-08-17T15:54:35	d6dc7e69b462	32,396

What changed in the latest version

-	as_of: 2026-08-17
+	as_of: '2026-08-17'
-	provenance:
-	produced_by: analyst-aisemi
-	produced_at: 2026-08-17
-	inputs:
-	- system/evidence-packs/MSFT.md (generated 2026-08-17T13:14:22Z, build_evidence_pack.py)
-	- system/state-files/MSFT.json (state_file_build.py, 2026-08-16T06:55:14Z)
-	- estimates_query.py drift/outliers as_of 2026-08-17T13:15:09Z
-	rubric: rubric_v1.yaml
-	schema: dossier.v1
-	value: 'Microsoft Corporation (NASDAQ, Software - Infrastructure). Three segments: Productivity and
-	Business Processes (M365 commercial cloud, revenue $37.8bn +14% in FY26Q4), Intelligent Cloud (Azure),
-	More Personal Computing. Tier-2 of the AI demand cascade — it owns the capex decision and writes the
…	diff truncated

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.