Microsoft Corporation (MSFT) — Coverage File
Microsoft Corporation (MSFT) — Coverage File
Data as of: FY2026 (year ended June 30, 2026); FY2026 10-K filed Jul 29, 2026; latest source Goldman Sachs conference, Sep 9, 2026 File last updated: October 10, 2026 (initial build)
0. 30-second reopen
- Business: An enterprise software and cloud platform with $331.8B of FY2026 revenue at a 46.8% operating margin; Microsoft Cloud is $214.4B of it and Azure passed $100B, growing 41%.
- Core debate: Will the AI build-out ($115.9B of cash capex in FY2026, $743.8B of contractual obligations) earn returns like the old capital-light Microsoft, and how much of the demand behind it rests on OpenAI?
- Base thesis: Microsoft sells AI at every layer (infrastructure, platform, seats plus consumption) into an installed base it already owns, so a contracted $678B backlog and Azure supply constraints should turn the capex into revenue while seat-based ARPU keeps the high-margin half compounding.
- Main bear case: Depreciation arrives before revenue, Microsoft Cloud gross margin keeps sliding from 69% to 66% and lower, and an OpenAI-heavy backlog proves less durable than it looks, with no second engine in consumer to absorb it.
- What to watch: Microsoft Cloud gross margin, Azure growth versus capex, the RPO mix and 12-month conversion, OpenAI revenue and receivables, M365 ARPU spread over seats, and what the new FY2027 segments still show.
- Valuation: ~$3.97T market cap, ~30x FY2026 GAAP EPS, ~1.7% FCF yield (Oct 9, 2026 web price): priced on earnings power with free cash flow depressed by the build.
1. Business snapshot
Microsoft reports three segments in FY2026, changing to two in FY2027. Productivity and Business Processes (M365 commercial and consumer, LinkedIn, Dynamics) had $140.0B of revenue at a 59.9% operating margin; Intelligent Cloud (Azure, server products, enterprise services) $137.8B at 41.3%; More Personal Computing (Windows OEM and devices, Xbox, search advertising) $54.1B at 26.6% [FY2026 10-K]. By product, server products and cloud services were $129.4B, M365 commercial $102.0B, Xbox $21.8B, LinkedIn $19.8B, Windows and devices $17.1B and search advertising $15.2B. No single customer exceeds 10% of revenue [FY2026 10-K].
Revenue is mostly contracted: multi-year enterprise agreements invoiced annually in advance (unearned revenue $75.7B), per-seat subscriptions, and consumption-based Azure. Commercial remaining performance obligation (RPO) reached $678B, up 84%, with ~30% recognisable in 12 months [FY2026 10-K]. From FY2027 Microsoft will report two segments, Agents and Infra and Devices and Consumer [8-K Sep 2, 2026]; the presentation that defines them is not in sources/.
2. Core debate
The first question is whether the AI capital cycle earns its keep. Cash additions to property and equipment nearly doubled to $115.9B (35% of revenue, against 12% in FY2022), depreciation rose from $22.0B to $34.3B, and leases signed but not yet commenced, mostly datacenters, jumped from $92.7B to $329.1B [FY2026 10-K]. Management guides calendar 2026 capex to about $175B after reclassifying some leases, and says FY2027 capex will grow again [Q4 FY2026 call]. The bull reads an 84% jump in backlog landing in the same year as the spend; the bear reads a utility-like cost base being bolted onto a software company.
The second is how much of the demand is one customer. About 45% of commercial RPO came from OpenAI at December 2025 [Q2 FY2026 call]; OpenAI produced $24.1B of FY2026 revenue (7.3%) while Microsoft holds ~25% of it and has funded $11.9B of a $13.0B commitment [FY2026 10-K]. Management says all sequential RPO growth in Q4 came from customers outside frontier-model companies and that nearly 90% of Microsoft Cloud revenue is from non-frontier customers [Q4 FY2026 call]. Whether that broadening continues decides how much of the backlog should be discounted.
3. Base thesis
Microsoft is the incumbent distribution layer for enterprise AI. It already owns the identity, productivity, developer (GitHub) and security seats, and it can charge for the same compute three ways: as Azure infrastructure, as platform services (Foundry, Fabric, databases), and as finished products sold per seat plus consumption (M365 Copilot, GitHub Copilot, Dynamics agents). That is why FY2026 revenue growth accelerated to 18% and operating income grew 21% with the margin rising to 46.8% even as gross margin fell [FY2026 10-K]. Demand still exceeds supply [Q4 FY2026 call], so capacity added is capacity sold, and the per-seat franchise keeps adding high-margin ARPU on top.
Assumptions this rests on:
- Capacity converts to revenue on schedule: Azure growth stays around 40% or more while new capacity lands (Q1 FY2027 guide ~45% constant currency) [Q4 FY2026 call].
- Microsoft Cloud gross margin stabilises in the mid-60s as efficiency (own silicon, MAI models, usage-based pricing) offsets depreciation; management guided it "relatively stable" quarter on quarter for Q1 FY2027 [Q4 FY2026 call].
- M365 commercial cloud revenue keeps outgrowing seats by around 10 points through Copilot, E5 and E7, and accelerates through FY2027 as consumption billing is added [FY2026 10-K; Q4 FY2026 call].
- Backlog growth broadens beyond OpenAI and Anthropic, so concentration falls as RPO converts.
- Operating cash flow keeps covering capex and the dividend without structural leverage; management expects FY2027 free cash flow to stay positive [Q4 FY2026 call].
4. Main bear case
The bear case is a margin reset, not a demand collapse. Intelligent Cloud cost of revenue grew 44% against 30% revenue growth, taking segment gross margin from 62.2% to 58.0%, and Microsoft Cloud gross margin has fallen from 72% (FY2023) to 71%, 69% and now 66% [FY2023–FY2026 10-Ks]. Servers and network equipment at cost rose from $132.8B to $215.9B in one year and depreciate over two to six years, so the FY2026 depreciation charge reflects only part of the new fleet [FY2026 10-K]. Free cash flow (operating cash flow less cash capex) has fallen three years running, from $74.1B to $71.6B to $67.0B, while net income rose from $88.1B to $133.7B; a further $19.9B of "other investing" outflow went mainly to prepay for components [FY2026 10-K].
How it breaks: AI demand normalises or reprices before the fixed costs do. Contractual obligations reached $743.8B, nearly equal to total assets of $758.4B, with $241.9B due in FY2027 [FY2026 10-K], so spending cannot simply stop. OpenAI, which Microsoft both funds and sells to, renegotiates or under-consumes; frontier labs move up into the application layer Microsoft monetises; and More Personal Computing, down 1% with Xbox impairments and a guided high-teens decline in Windows OEM and devices for FY2027, offers no offset [FY2026 10-K; Q4 FY2026 call]. Reported earnings meanwhile flatter the picture: $5.0B of FY2026 net income came from OpenAI dilution gains, and a further $3.2B Q4 gain on Anthropic is not excluded from adjusted EPS [FY2026 10-K; Q4 FY2026 call].
5. Business quality
Very high in the franchise, and changing in character. Productivity and Business Processes runs an 82% gross margin and a 60% operating margin; its gross margin rose in FY2026 even as Copilot usage grew [FY2026 10-K]. The moat is distribution and integration: M365, Entra identity, Teams, GitHub (225 million users) and the enterprise agreement, into which new SKUs are sold without new customer acquisition [Q4 FY2026 call]. Cash arrives ahead of revenue (the change in unearned revenue added $9.4B to FY2026 operating cash flow), and the top line is diversified by customer and geography [FY2026 10-K].
What has changed is capital intensity. Net property and equipment rose from $205.0B to $313.1B in a year and is now the largest asset on the balance sheet [FY2026 10-K]. Microsoft is becoming partly a capacity business, where returns depend on utilisation, component prices and power, not only on software economics. Management's answer is that about two-thirds of capex is short-lived CPUs and GPUs that can be slowed if demand turns, and that one fleet serves Azure, first-party apps and internal R&D, so capacity is fungible [Q4 FY2026 call; Deutsche Bank conf. Aug 2026].
6. Key value drivers
Azure: capacity online × utilisation × price, now with Foundry (100,000 customers, revenue more than doubled), PostgreSQL (+55%) and Fabric (40,000 paid customers) layered on [Q4 FY2026 call]. M365 commercial: seats (+6%, mostly SMB and frontline, the lowest-ARPU cohorts) × ARPU (Copilot over 30 million paid seats, E5, the new E7 suite) plus usage billing added in July 2026 [FY2026 10-K; Q4 FY2026 call]. GitHub Copilot moved to usage-based billing in June 2026, which lifted revenue and segment margin [Q4 FY2026 call]. The cost side is the Microsoft Cloud gross margin, driven by depreciation, component prices (about $25B of calendar-2026 capex is component inflation [Q3 FY2026 call]), own silicon (Maia 200, Cobalt 200) and cheaper first-party models. Below the line, OpenAI marks and other equity stakes swing reported EPS. Consumer is a drag in FY2027: Windows OEM and devices guided down high teens, Xbox under a reset management says returns to growth in FY2027 [Q4 FY2026 call].
7. Financial profile
| Fiscal year (June) | FY2022 | FY2023 | FY2024 | FY2025 | FY2026 |
|---|---|---|---|---|---|
| Revenue | $198.3B | $211.9B | $245.1B | $281.7B | $331.8B |
| Operating income | $83.4B | $88.5B | $109.4B | $128.5B | $155.2B |
| Operating margin | 42.1% | 41.8% | 44.6% | 45.6% | 46.8% |
| Operating cash flow | $89.0B | $87.6B | $118.5B | $136.2B | $182.9B |
| Cash capex (PP&E) | $23.9B | $28.1B | $44.5B | $64.6B | $115.9B |
| Capex / revenue | 12.0% | 13.3% | 18.1% | 22.9% | 34.9% |
| Free cash flow (OCF − capex) | $65.1B | $59.5B | $74.1B | $71.6B | $67.0B |
| Diluted EPS (GAAP) | $9.65 | $9.68 | $11.80 | $13.64 | $17.95 |
Sources: FY2022–FY2026 10-Ks; margins, capex ratio and FCF computed by this file.
FY2026 adjusted EPS (excluding OpenAI) was $17.28, +22% [FY2026 10-K]. Q4 FY2026 revenue was $90B (+18%), Microsoft Cloud gross margin 65%, capex including finance leases $41B and free cash flow $19.6B [Q4 FY2026 call]. Balance sheet at June 30, 2026: cash and short-term investments $76.8B (from $94.6B), debt $40.3B, finance lease liabilities $66.6B (from $46.2B), equity investments $36.3B [FY2026 10-K]. Q1 FY2027 guide: revenue $89.85–90.95B (+16–17%), flat operating margin, capex over $50B; FY2027: double-digit revenue and operating income growth, operating margin down less than a point, tax rate ~20% [Q4 FY2026 call].
Two reporting changes affect comparability from FY2027. Datacenter and office building useful lives extend from 15 to 25 years, which management says barely helps FY2027 operating income but shifts future datacenter leases from finance to operating leases, which are excluded from reported capex; that alone cut the calendar-2026 capex figure from roughly $190B to roughly $175B with spending unchanged [Q3 FY2026 call; Q4 FY2026 call]. And the segment change removes the Productivity / Intelligent Cloud split this file uses to track AI margins [8-K Sep 2, 2026].
8. Valuation
Microsoft gives no long-term financial targets in sources; the operative frame is the FY2027 guide of double-digit revenue and operating income growth, a slightly lower operating margin and positive free cash flow [Q4 FY2026 call]. At ~30x GAAP and ~31x adjusted FY2026 EPS, the shares price Microsoft on its earnings power. On free cash flow the yield is ~1.7%, or ~1.2% after the $19.9B of component prepayments (inferred), so the valuation assumes capex intensity falls back from 35% of revenue or that the assets earn far more than today's free cash flow shows. Earnings themselves include non-operating gains (OpenAI dilution, Anthropic) that should be stripped out when judging the multiple.
Market data (outside sources/, web, dated) Share price 3.97T (stockanalysis.com, retrieved Oct 10, 2026). Derived by this file from filings: 7,427M shares outstanding at June 30, 2026; net cash excluding leases ~76.8B cash and investments less 30B net debt including 17.95), ~31.0x adjusted EPS (3.94T, ~25.4x FY2026 operating income, FCF yield ~1.7% (0.91 quarterly). Next earnings date Oct 28, 2026 (same web source).
9. Management & capital allocation
Satya Nadella has been CEO since 2014 and chairman; Amy Hood is CFO. Both have run the cloud transition and now the AI build, and both frame capex in terms of short-lived assets, fungible capacity and efficiency "grind" [Q4 FY2026 call]. Hood says Microsoft will borrow if needed and "the ROIC is there" [Goldman Sachs conf. Sep 2026]. Board turnover in 2026: Carmine Di Sibio joined in May; Reid Hoffman will not stand for re-election [8-K May 14, 2026; 8-K Jun 5, 2026].
Capital allocation has shifted from returns to reinvestment. FY2026 repurchases were $16.7B (36M shares), largely offsetting $12.4B of stock-based compensation: diluted shares moved only from 7,469M (FY2024) to 7,453M (FY2026). Dividends declared were $27.0B, and $40.6B of the $60B buyback authorisation remains [FY2026 10-K]. Long-term debt fell from $40.2B to $31.1B, but finance leases are now the larger fixed claim. The biggest past acquisition, Activision (closed October 2023, $69.1B net cash in FY2024 acquisitions), sits inside an Xbox business whose revenue fell 7% in FY2026 and took impairment charges [FY2024 10-K; FY2026 10-K]. The OpenAI stake has been the best-returning bet on paper; the Anthropic stake produced a $3.2B gain in Q4 [Q4 FY2026 call].
Pay: CEO total compensation was 84.2M in stock awards), up from $79.1M [2025 proxy]. The CEO's $50M target equity is 100% performance shares, scored on revenue growth only (Azure 35%, Microsoft Cloud ex-Azure 35%, consumer services 15%, Xbox 15%) with a relative-TSR modifier; the cash bonus adds operating income [2025 proxy]. No metric captures capital intensity, free cash flow or return on invested capital in the years capex tripled. The FY2023 performance shares paid 161.52% of target after a 150% TSR modifier [2025 proxy]. Nadella held 900,572 shares at the 2025 record date and received 93.46% support for re-election [2025 proxy; 8-K Dec 8, 2025]. Insider transaction data is not in sources.
Claims to verify over time:
- Demand exceeds supply through the rest of calendar 2026 [Deutsche Bank conf. Aug 2026].
- Short-lived assets let Microsoft slow spending quickly if demand changes [Q4 FY2026 call].
- Xbox returns to growth in FY2027 [Q4 FY2026 call].
- M365 commercial cloud growth accelerates through FY2027 [Q4 FY2026 call].
10. Risks ranked by damage
1. AI capex returns and the depreciation staircase Damage rank: High · Permanence: High · Fixability: Medium Mechanism: Assets bought at peak component prices depreciate over two to six years whether or not they are fully utilised, compressing Microsoft Cloud margins if revenue per unit of capacity falls. Evidence today: depreciation +56% to $34.3B; Intelligent Cloud gross margin 62.2% → 58.0%; Microsoft Cloud gross margin 69% → 66%; capex 35% of revenue; FCF down three years running [FY2026 10-K]. The 10-K itself warns that overestimating demand may cause underutilisation and asset impairment. What would confirm it: Microsoft Cloud gross margin below ~63% while Azure growth decelerates below ~35%. Impact: lower through-cycle returns on a much larger asset base; the largest single driver of value.
2. OpenAI concentration and circularity Damage rank: High · Permanence: Medium · Fixability: Medium Mechanism: One counterparty, in which Microsoft is a ~25% owner and funder, supplies a large share of backlog and 7.3% of revenue; its finances, strategy or contract terms could change. Evidence today: ~45% of commercial RPO at December 2025; $250B of incremental Azure commitments signed October 2025; $24.1B FY2026 revenue and $6.0B receivable (~91 days) [Q2 FY2026 call; Q1 FY2026 call; FY2026 10-K]. Offsetting: RPO ex-OpenAI +25% and all Q4 sequential RPO growth from non-frontier customers [Q4 FY2026 call]. What would confirm it: OpenAI receivables growing faster than OpenAI revenue, a disclosed renegotiation, or RPO shrinking. Impact: lost revenue plus capacity built for one customer; the equity stake would also be marked down.
3. Frontier labs and model commoditisation move up the stack Damage rank: Medium · Permanence: High · Fixability: Medium Mechanism: Model providers that are also Azure customers sell directly into enterprise workflows, reducing Copilot to one option among many and pressuring seat pricing. Evidence today: the 10-K names frontier model providers, "some of which are also current or potential partners", as competitors; management concedes "areas of overlap competitively" [FY2026 10-K; Deutsche Bank conf. Aug 2026]. Microsoft's response is model choice (11,000 models, own MAI family) and governance (Agent 365) [Q4 FY2026 call]. What would confirm it: M365 ARPU spread over seats narrowing; Copilot net adds stalling. Impact: hits the highest-margin segment.
4. Supply, power and component cost Damage rank: Medium · Permanence: Low · Fixability: Medium Mechanism: Land, power, GPUs and memory constrain how fast capex turns into revenue and raise its cost. Evidence today: ~$25B of calendar-2026 capex is component inflation; Azure has been supply-constrained for several quarters [Q3 FY2026 call; Q4 FY2026 call]. What would confirm it: further capex raises without matching Azure acceleration. Impact: mostly timing and margin, not franchise.
5. Consumer drag (Xbox, Windows OEM) Damage rank: Low · Permanence: Medium · Fixability: Medium Mechanism: Gaming and PC demand weaken while Activision assets underperform. Evidence today: Xbox revenue −7%, hardware −29%, impairments; Windows OEM and devices guided down high teens in FY2027 [FY2026 10-K; Q4 FY2026 call]. What would confirm it: Xbox content and services still declining in H2 FY2027. Impact: small relative to the commercial business, but it removes any offset to a cloud margin reset.
6. Reporting visibility Damage rank: Low · Permanence: Medium · Fixability: High Mechanism: New segments, longer useful lives, the finance-to-operating lease shift and large equity marks make AI economics harder to track and capex and free cash flow look better without changing cash spent. Evidence today: 8-K Sep 2, 2026; useful-life change effective FY2027 [Q4 FY2026 call]. What would confirm it: Microsoft Cloud gross margin or Azure growth disclosures dropped in the new structure. Impact: raises the risk of misreading the case rather than changing it.
11. Metrics to monitor
| Metric | Why it matters | Prior | Latest | As of |
|---|---|---|---|---|
| Azure growth | Capacity converting to revenue (§3 assumption 1) | 34% (FY2025) | 43% (41% FY2026) | Q4 FY2026 |
| Microsoft Cloud gross margin | Bear case driver (§10 risk 1) | 69% (FY2025) | 65% (66% FY2026) | Q4 FY2026 |
| Intelligent Cloud cost of revenue vs revenue growth | Margin reset in the AI half | not tracked | +44% vs +30% | FY2026 |
| Commercial RPO; share due in 12 months | Backlog size and quality | $368B; ~40% | $678B; ~30% | Jun 30, 2026 |
| RPO growth ex-OpenAI / bookings ex-OpenAI | Concentration (§10 risk 2) | not disclosed | +25% / +18% | Q4 FY2026 |
| OpenAI revenue / receivable | Circularity check | not in sources | $24.1B / $6.0B | FY2026 |
| M365 commercial cloud growth less seat growth | ARPU from Copilot, E5, E7 | 15% − 6% (FY2025) | 17% − 6% | FY2026 |
| Cash capex / operating cash flow | Self-funding (§3 assumption 5) | 47% (FY2025) | 63% | FY2026 |
| Free cash flow (OCF − cash capex) | Cash return capacity | $71.6B (FY2025) | $67.0B | FY2026 |
| Depreciation expense | Staircase | $22.0B (FY2025) | $34.3B | FY2026 |
| Leases not yet commenced | Committed future capacity | $92.7B | $329.1B | Jun 30, 2026 |
| Diluted shares | Buyback vs dilution | 7,465M (FY2025) | 7,453M | FY2026 |
12. Thesis tripwires
| Direction | Tripwire |
|---|---|
| Stronger | Microsoft Cloud gross margin holds 65%+ for four quarters while Azure stays near 40% |
| Stronger | OpenAI's share of RPO falls below a third through non-frontier growth, disclosed by management |
| Stronger | Capex growth slows below revenue growth in FY2028 with Azure still above 30% |
| Weaker | Microsoft Cloud gross margin below 63% |
| Weaker | Free cash flow falls for a fourth year (FY2027), or Microsoft issues debt to fund buybacks |
| Weaker | M365 commercial cloud growth minus seat growth falls below ~6 points |
| Breaks | OpenAI materially reduces or renegotiates its Azure commitments |
| Breaks | An asset impairment of datacenter or GPU capacity |
13. Open questions
- What exactly do the FY2027 segments, Agents and Infra and Devices and Consumer, disclose, and are Microsoft Cloud gross margin and Azure growth still reported? The "FY27 Segments and Investor Metrics" presentation (8-K Sep 2, 2026, Exhibit 99.1) is not in sources/ and should be collected.
- What is OpenAI's share of RPO at June 2026 (last disclosed ~45% at December 2025), and how large is the Anthropic Azure commitment signed in November 2025?
- How much future capacity moves from finance to operating leases after the useful-life change, and how should capex and free cash flow be restated for comparability?
- Are the $19.9B of component prepayments recurring, and should they be treated as capex?
- What are the economics of the October 2025 OpenAI agreement (revenue share, IP rights, duration)? Sources state the rights exist but not the terms.
- Q1 FY2027 results (Oct 28, 2026 per web): first print under the new segments and first test of the stable Microsoft Cloud margin guide.
14. Update log
| Date | Trigger | What changed | Impact on thesis |
|---|---|---|---|
| 2026-10-10 | Initial build | Initial Coverage File created from the FY2022–FY2026 10-Ks, FY2026 10-Qs, the FY2026 earnings calls, the Deutsche Bank (Aug 2026) and Goldman Sachs (Sep 2026) conference transcripts, the 2025 proxy and 2025–26 8-Ks. The Bull & Bear memo (built on the FY2026 filings before transcripts were collected) was used as the backbone for §3–§4; this file adds transcript evidence the memo lacked (OpenAI ~45% of RPO, RPO ex-OpenAI +25%, CY2026 capex ~$175B, useful-life change, FY2027 segment change, Anthropic gain). Only the §8 share price and earnings date come from the web, labelled. Insider transaction data and the FY2027 segment presentation are not in sources/. | Baseline established. |
15. Source map
| Document | Period / date | Used for |
|---|---|---|
| Form 10-K | FY2026 (filed 2026-07-29) | Segments, revenue mix, RPO, OpenAI related-party data, PP&E, leases, contractual obligations, cash flow, buybacks |
| Forms 10-K | FY2022–FY2025 | Five-year history, Microsoft Cloud gross margin, FY2025 baselines, Activision cash outlay |
| Forms 10-Q | Q1–Q3 FY2026 | OpenAI recapitalisation, interim RPO |
| Earnings call transcripts | Q1–Q4 FY2026 (2025-10-29 to 2026-07-29) | Azure, Copilot, RPO mix, capex guidance, useful-life change, FY2027 outlook |
| Deutsche Bank conference transcript | 2026-08-27 | Demand vs supply, backlog breadth, frontier-lab competition |
| Goldman Sachs Communacopia transcript (CFO) | 2026-09-09 | Re-segmentation rationale, debt stance, capex flexibility |
| Proxy statement (DEF 14A) | 2025-10-21 | CEO pay, PSA metrics and payout, ownership |
| Current reports (8-K) | 2025-12-08, 2026-05-14, 2026-06-05, 2026-09-02 | Vote results, board changes, FY2027 segment change |
| Web market data (stockanalysis.com) | 2026-10-09 close | §8 market-data callout and earnings date only |
| Microsoft — Bull & Bear Memo; Microsoft-Business-Overview | 2026 (prior work) | Orientation only (not evidence) |