MSFT › analyze
MSFT · Analyze from before
This updates: Output/Stocks/Technology/MSFT/analyze-2026-07-15.md (2026-07-15, 🟢 BUY/Add, conviction 8.0, FV $420-500, entry $380-400, trim $520+ fixed dollar).
Price then → now: $384.93 (7/14 close) → $505.06 (2026-08-27), +31.2% in 43 days. 52-week range $349.20-553.72; spot is 91.2% of the 52w high, up from 70% at baseline.
Event list since baseline:
1. 2026-07-29 — Q4 FY2026 print (fiscal year ended 2026-06-30): revenue $90.0B (+18% YoY), net income $35.8B (+32% YoY), diluted EPS $4.81. Azure grew 40-43% cc for the quarter (sources vary slightly) and crossed $100B in trailing annual revenue for the first time. Stock added ~$260B of market value in one session.
2. Q1 FY2027 guide issued same call: Azure ~45% cc growth guided, beating the ~41.4% Street consensus — an acceleration, not the deceleration the baseline flagged as a live risk.
3. FY2027 capex guide: $255-260B (calendar-year basis), a ~35% step up from the ~$190B CY2026 figure. Effective FY2027, Microsoft extends datacenter/office useful life from 15 to 25 years, which reclassifies a slice of future datacenter leases from finance to operating — this mechanically trims the reported capex/CY2026 figure (~$190B → ~$175B) without changing real investment.
4. FY2027 operating-margin guide: down <1 point — materially milder than the ~4.5pt-to-63% gross-margin compression the baseline flagged as a named risk.
5. Analyst target raises across the board post-print (Wells Fargo $650→$700, Citigroup $570→$600, Cantor $502→$522, Tigress $680→$690, Bernstein $646→$647), consensus mean target now $569.45 (was $559.86), median $550, average rating 1.36 "Strong Buy" (52 analysts). Only Stifel sits at Hold ($450 PT); no Sell/Underweight in the set.
6. FY2026 (fiscal year, now closed) results confirm the capex-vs-FCF story continued, not stabilized: capex $115.95B (+79.6% YoY) vs FCF $66.99B (-6.5% YoY) — the second consecutive annual FCF decline.
7. A new distortion surfaced on the earnings side: Microsoft now marks its OpenAI stake to fair value each quarter under the OpenAI recapitalization terms. Q2 FY2026 (Dec quarter, already inside the baseline's TTM window but not flagged in the July report) carried a $7.6B / $1.02 EPS non-operating gain from this mark. This is the same mechanism the repo's pattern-ai-build-inflates-earnings-while-destroying-fcf note documents on GOOGL/AMZN/NOW/MU — MSFT is directly in scope and had not previously been tested against it.
8. No AI-run-rate refresh — the $37B/+123% YoY figure (as of Q3 FY26) was not updated on the Q4 call; it remains the latest official number.
The delta ledger
Lead items — SUPERSEDED / new findings:
| # | Claim (baseline) | Type/Decay | Old → New | Status | What carried it |
|---|---|---|---|---|---|
| 21 | Forward P/E ~20x | Price/fast | 20x → vendor forwardPE 21.43x is FY2028, not FY2027 |
🔄 SUPERSEDED — pitfall fired | epsCurrentYear $19.75 / priceEpsCurrentYear 25.57x is the true current-year (FY2027) multiple, per [[pitfall-vendor-forward-eps-is-the-wrong-fiscal-year]]. Implied epsForward $23.57 is one year further out. Cross-checked via Yahoo's own dual fields, consistent with the pitfall's documented AMD/ANET mechanism. |
| 20 | Fair value $420-500, bear floor $360 | Price/fast | → $460-560, bear floor ~$400 | 🔄 SUPERSEDED | Re-derived from scratch (see Valuation below) on the corrected forward EPS and the confirmed FY2027 guide set. |
| 23 | Entry zone $380-400 | Price/fast | → $420-460 | 🔄 SUPERSEDED — stale entry zone | Per [[pitfall-stale-entry-zone-suppresses-a-name]]: the old zone is a derivative of the old FV and never fills against the new one. Watchlist had already marked this "ZONE EXITED." |
| 24 | Trim $520+ (fixed dollar) | Price/fast | → 30x ttm (≈$539) | 🔄 SUPERSEDED — converted per CLAUDE.md rule | Fixed-dollar trims go stale as earnings grow; re-based to a multiple of ttm EPS ($17.97) rather than the contaminated forwardPE field. |
| 22 | Bogle expected return ~14%/yr | Price/fast | → ~10-11%/yr | 🔄 SUPERSEDED | Lower dividend yield (0.73% vs 0.91%) and a materially higher entry multiple compress the forward-return estimate even with growth intact. |
| 25 | Verdict BUY/Add, conviction 8.0 | Judgment | → HOLD, conviction 7.5 | 🔄 SUPERSEDED | Adjudicated last, from the settled ledger below. |
| 3 | "Evergreen assessment... near-term margin compression... reasonable risk/reward" — implicit assumption the compression would be steep | Trend/slow | Guided ~63% GM → actual FY26 GM 67.94% (-88bps only); FY27 op-margin guide <1pt decline | 🔄 PARTIALLY RETRACTED — the feared severity did not appear | Direct financials (fin.py, roic.ai cross-check) plus the FY2027 guide reported across multiple outlets. A predicted risk that substantially failed to materialize is a force for the moat, per the Phase 4b instruction — recorded as such, not dropped. |
| — | Azure faces a "genuine" competitive front (Google Cloud) that could decelerate the thesis (named risk #3) | Structural/cycle | Feared deceleration → 45% cc guide, beating 41.4% consensus | 🔄 SUPERSEDED (risk resolved favorably, this pass) | Company's own guide plus consensus-beat framing from two independent outlets. Google Cloud's own growth rate was not independently re-checked this pass — flagged as untested below; the claim is about MSFT's trajectory, which is unambiguously up. |
| 🆕 | Earnings-quality caveat: FY26 NI carries a ~$7.6B / $1.02 EPS non-operating OpenAI-stake mark, recurring quarter to quarter under the new recapitalization accounting | State/fast | n/a → new finding | 🆕 NEW | MSFT's own 8-K/press disclosure (Q2 FY26), corroborated by CNBC's Q2 FY26 earnings coverage. Net income growth (31.3% FY26) outran operating income growth (20.8%) — exactly the gap this mark explains. |
| 🆕 | Useful-life extension (15→25yr) reclassifies future datacenter leases from finance to operating category starting FY2027 | State/fast | n/a → new finding, live pitfall instance | 🆕 NEW | This is the mechanism [[pitfall-lease-financed-capex-hides-from-the-fcf-line]] describes, caught in real time rather than after the fact. Action for the next pass: track Δ gross PP&E and the finance-lease balance in FY2027 10-Qs, not the capex line alone — the capex line will understate real investment intensity going forward. |
Trend claim that drifted further against the thesis:
| # | Claim | Old → New | Status |
|---|---|---|---|
| 7 | FCF CAGR "lags badly," ~9.6% (5yr), capex-drag story | FY26 FCF $66.99B, -6.5% YoY (2nd straight annual decline: $74.07B FY24 → $71.61B FY25 → $66.99B FY26); fin.py 3yr FCF CAGR now 4.0%, while OCF 3yr CAGR is 27.8% — the gap is widening, not stabilizing. Capex now 62.9% of OCF (was 47.4% in FY25). FY2027 CY capex guide of $255-260B (+35%) confirms this continues at least one more year. | 📉 DRIFTED — against the thesis, not yet breaking it. Breaks if FY2027 FCF turns negative or capex growth doesn't decelerate by FY2028 guidance. |
Carried / refreshed — compact list (all re-tested against fresh Q4 FY2026 data, fin.py + yahoo-finance MCP + roic.ai cross-check):
- ✅ Moat sources intact: switching costs, network effects, cost advantage/scale, intangibles — all still present and reinforcing. Reinforced, not just carried: Azure crossed $100B and re-accelerated.
- ✅ Evergreen / forever-business rating — unchanged.
- 🔁 Revenue CAGR 14.5%(5y)/14.3%(8y) → FY26 revenue +17.7% YoY, fin.py 3yr CAGR 16.1% — accelerating.
- 🔁 Net income CAGR 18.1%(5y) → FY26 NI +31.3% YoY, 3yr CAGR 22.7% — up, but see the OpenAI-mark caveat above; the clean (operating-driven) growth rate is closer to the 20.8% operating-income growth.
- ✅ Dividend CAGR ~9.7-10% — held; FCF payout ratio actually improved to 19.8% (was 23.6% FY25) despite the FCF decline, because dividends grew slower than the (still-large) cash base.
- 🔁 ROIC/ROE/ROA: ROIC 28.2%→27.5% (roic.ai cross-check, mild dip, still elite), ROE 32.8%→34.0%, ROA 18.0%→19.4% (roic.ai) / 14.1% (Yahoo, different asset-base convention — both directionally consistent, up).
- ✅ Fortress balance sheet: Debt/Assets improved to 7.5% (fin.py, was 9.8%); net-cash position and low leverage intact.
- 🔁 Shares outstanding: 7,465M → 7.43B, essentially flat (fin.py 3yr CAGR -0.1%); gross buybacks up to $22.27B (was $18.42B) but still mostly offsetting SBC issuance rather than net-retiring shares at scale.
- ✅ Institutional ownership 75.7% → 75.9%, unchanged.
- ✅ AI run-rate $37B/+123% YoY — not refreshed by the company since Q3 FY26, remains the latest official figure, carried as-is.
- ✅ Q3 FY2026 FCF $15.8B / OCF $46.7B (capex 66% of OCF) — re-verified against yahoo-finance quarterly cash flow: matches to the dollar ($15,803M). This was correctly labeled a single quarter, not TTM, in the baseline —
[[pitfall-single-quarter-fcf-read-as-ttm]]does not fire here; flagged as a discipline the baseline got right. - ✅ Graham IV ~$155 (fin.py: $155.19) — still correctly ignored as meaningless for an asset-light, 8.5x-book compounder.
How the close calls were decided
Was the Azure/gross-margin risk resolution a real reversal, or cherry-picked good news? Both named risks (#2 gross margin, #3 Azure deceleration) were tested against the company's own guidance and against independent press coverage (CNBC, Morningstar, TradingKey, GeekWire), not against a single bullish source. Gross margin fell 88bps (68.82%→67.94%), nowhere near the baseline's feared 63% floor; op margin actually rose 116bps YoY (45.62%→46.78%) for the year that just closed, and the company's own forward guide for FY2027 is a sub-1-point decline. Azure's guide (45% cc for Q1 FY27) explicitly beat consensus (41.4%). Two independent forces, same direction, on primary company disclosure — this clears the "no single source" bar. What was NOT independently re-verified: Google Cloud's own growth rate this pass, so the comparative "genuine competitive pressure" framing is carried forward on the baseline's word, not re-tested — see Untested below.
Was the FCF/capex deterioration weighed correctly against the margin/Azure good news? Yes — deliberately kept as a separate DRIFTED row rather than netted against the two resolved risks. The capex trajectory is not just continuing, it is guided to accelerate further (+35% CY2027 vs CY2026) — this is a distinct claim from the margin-compression fear the baseline named, and the two should not be allowed to cancel out. A business can hold its margins and moat while still spending itself into a multi-year FCF trough; that is exactly the shape here.
Was the OpenAI equity-mark finding treated as a retraction of the "clean bill of health" verdict, or a caveat? A caveat. The mark is real, disclosed, and recurring (mark-to-market each quarter under the new recapitalization terms), but it explains roughly 5 points of the ~6-point pretax-margin expansion — it does not turn a good quarter into a bad one, and operating income still grew a real 20.8%. Filed as a data-quality caveat on the earnings line, parallel to (not overriding) the strong operating trend.
Was the vendor forward-P/E error treated as invalidating the whole valuation, or just the one field? Just the field, and it was corrected using Yahoo's own epsCurrentYear/priceEpsCurrentYear pair (the documented fix in [[pitfall-vendor-forward-eps-is-the-wrong-fiscal-year]]), not discarded. The corrected 25.6x current-year multiple, not the vendor's 21.4x, is what feeds the trim/entry math below.
Thesis persistence and conviction delta
Structural claims: 4/4 carried (100%). Moat sources, AI/Copilot distribution advantage, evergreen rating, and the Azure competitive-position claim all held or improved.
Trend claims: 5/6 carried or refreshed (83%). Revenue, net income, dividend, ROIC/ROE/ROA, and gross-margin-stability all held. Only the FCF-CAGR claim drifted further against the thesis.
Combined Structural + Trend thesis persistence: 9/10 = 90%.
Conviction: 8.0 → 7.5. High persistence against a big price move (+31% in 43 days) is, per the framework, "the classic re-rating setup: the business held and the multiple moved" — and that is exactly what happened here. The downgrade is not a business-quality call; it is driven entirely by the Price rows (#20-24), all SUPERSEDED in the direction of a closed discount: fair value midpoint moved from ~$465 to ~$510 while price moved from $385 to $505, so the ~30% discount the baseline was buying is gone. The two named risks that resolved favorably (#2, #3) argue for not cutting conviction further; the FCF/capex claim drifting argues against raising it back to 8.0. Net: -0.5, verdict moves from BUY/Add to HOLD.
What is genuinely new
- The OpenAI mark-to-market mechanism (see ledger) — MSFT joins GOOGL/AMZN/NOW/MU in
[[pattern-ai-build-inflates-earnings-while-destroying-fcf]]'s scope with a concrete, sourced instance. Worth adding MSFT to that note's ticker list with this session's figures. - The useful-life/lease-reclassification mechanism — a live, forward-looking instance of
[[pitfall-lease-financed-capex-hides-from-the-fcf-line]], caught before it distorts a quarter rather than after. - Analyst target compression relative to price: at baseline, spot sat 45% below the mean target; now it sits 12.8% below ($569.45 mean) and 8.9% below the median ($550). The Street has both raised targets and watched price close most of the gap — consistent with the re-rating read above, not with a broken thesis.
Updated verdict — HOLD, conviction 7.5
Fair value: $460-560 (midpoint ~$510). Built on FY2027 current-year consensus EPS $19.75 (Yahoo epsCurrentYear, corrected per the forward-EPS pitfall) at a 24-29x band — the high end of the baseline's 22-28x band, widened modestly to reflect the resolved Azure/margin risks, not widened aggressively given the FCF/capex claim is still drifting.
| Model | Output | Weight | Note |
|---|---|---|---|
| True current-year P/E (corrected) | 25.6x on $19.75 FY27 EPS | High | priceEpsCurrentYear; vendor forwardPE (21.4x) is one fiscal year too far out — do not quote it. |
| Bogle expected return | ~10-11%/yr | High | 0.73% yield + ~10-13% earnings growth, modest multiple-reversion drag. Down from ~14%/yr — the higher entry multiple is doing the work. |
| FCF yield | 1.79% trailing / 4.9% on OCF | Med | Trailing FCF yield worse than baseline's 2.5% — capex outpaced the OCF gain again. |
| PEG | 1.61 (vendor) / ~1.6 corrected | Med | Up from ~1.2 — less margin of safety in the growth-adjusted multiple. |
| Graham √(22.5·EPS·BVPS) | $155 | Ignore | Meaningless for an asset-light 8.5x-book compounder — unchanged verdict from baseline. |
| DYT | 0.73% vs 0.79% 5yr avg | Low | Flipped from mild-positive (baseline) to mild-negative — price-driven per [[pitfall-dyt-inverts-when-price-caused-the-yield]], correctly too small to lean on either direction. |
Bear floor ~$400: if FY2027 FCF turns negative for a full year, capex growth doesn't decelerate into the FY2028 guide, and the multiple compresses to ~20x on flattish EPS ($19.75), that arithmetic lands here — one year of earnings growth above the baseline's $360 floor.
Entry zone $420-460 (supersedes the stale $380-400): roughly a 9-17% pullback from spot, and close to the baseline's own former fair-value floor, which is a plausible technical/valuation confluence given the stock round-tripped a similar range (~$385 to ~$554) inside the last twelve months.
Trim: 30x ttm (≈$539 at current ttm EPS $17.97 — recomputes as EPS rises). Basis is ttm, not fwd, deliberately: the vendor forwardPE field is contaminated (wrong fiscal year) and the site's mechanical dollar = mult × EPS computation should not inherit that error.
Key risks, updated: 1. FCF/capex trajectory — escalated, not resolved. FY2027 CY capex guide of $255-260B (+35%) means the FCF trough continues at least one more year; this is the risk that most directly threatens the thesis if it runs a third or fourth year without visible AI monetization. 2. Earnings-quality drag from OpenAI marks — new. Recurring mark-to-market on the OpenAI stake will keep adding noise (in either direction) to headline EPS; treat GAAP net-income growth with a discount until the operating-income line is checked alongside it. 3. Valuation — the discount is gone. Spot ($505) sits inside the new FV range, near the top of what the baseline itself would have called fair. This is a HOLD, not an AVOID, but it is no longer the "closest to value in years" setup the baseline described. 4. ~~Gross margin compression~~ and ~~Azure deceleration~~ — downgraded from named risks to watch items, per the evidence above; re-test at the Q1 FY2027 print rather than treating as live threats.
Upgrade condition: a pullback into the $420-460 entry zone with the FY2027 fundamentals still intact, or FCF re-accelerating (positive YoY growth) while capex growth decelerates toward the FY2028 guide — either would restore the "great company, temporarily cheap" setup.
Trim condition: a sustained move through ~$539 (30x ttm) without a corresponding step-up in FY2027 EPS delivery, or a second consecutive year of FCF decline into FY2027 alongside multiple expansion rather than compression.
A portfolio-specific passage was removed from the public build.
What this pass did NOT test
A portfolio-specific passage was removed from the public build.