MSFT — Research Report (2026-07-22)

TL;DR

• What it is: the world's largest enterprise software company — Windows, Office/M365, Azure cloud, LinkedIn, Xbox, plus a huge OpenAI partnership.
• Overall take: excellent business, priced for near-perfection — Azure's AI growth is real but the stock already assumes it continues for years.
• Educational decision summary: Watch, Medium conviction — own it if you believe in a 5-10yr AI/cloud compounder; the entry price offers little margin of safety today.
• Market cap: ~$2.96T (2nd/3rd largest company globally — scale means every new dollar of growth is hard-won).
• FY25 revenue: $281.7B, +14.9% YoY (growth is still double-digit at massive scale — rare).
• FY25 free cash flow: ~$71.6B, but capex nearly doubled YoY to $64.6B (AI buildout is eating the cash machine).
• P/E: ~29x on FY25 earnings (paying a premium vs. its own 10-year average).
• Strongest bull point: Azure growing ~40% YoY with an $80B+ order backlog it can't even fulfill due to power/chip constraints — demand is not the problem.
• Strongest bear point: capex is compounding faster than free cash flow, and if AI monetization (Copilot, Azure AI margins) disappoints, today's price has far to fall (Morgan Stanley's own bear case: $250).
• Watch next: FY2026 Q4 earnings (expected ~2026-07-29) — Azure growth rate and capex guidance for FY2027.

1. Business Overview

So what: Microsoft sells software and cloud infrastructure to almost every company on earth, and increasingly rents out AI compute — three legs, one balance sheet.

Microsoft reports three segments:

Customers span nearly every industry and geography — no single customer concentration risk (the opposite problem most companies worry about). Revenue model is overwhelmingly recurring: subscription (M365, Azure consumption commitments) rather than one-time license sales, which is why Microsoft's revenue is unusually predictable for its size.

The company's biggest strategic bet of this decade is its cumulative multi-billion-dollar investment in and commercial partnership with OpenAI, which routes heavily through Azure ("Azure AI" revenue) — this is the single biggest swing factor in the stock's valuation today.

2. Financials (5-Year Trend, FY2020-FY2025, fiscal year ends June 30)

So what: revenue and profit both roughly doubled since 2021, but the newest twist is capital expenditure (capex) growing even faster than cash flow — the AI buildout is genuinely capital-intensive.

Note on data continuity: Microsoft's EDGAR revenue tag changed in FY2021 (RevenueFromContractWithCustomerExcludingAssessedTax); FY2020 is a partial/differently-tagged figure shown for reference only, not for CAGR math.

Cash flow reality (profit is opinion, cash is fact):

Capex/OCF jumping from ~27% to ~47% in three years is the single most important line in this report: Microsoft is reinvesting an ever-larger share of every operating dollar into AI data centers, and FCF growth has already stalled (FY25 FCF is roughly flat vs FY24) even as operating income kept climbing. That's the AI bet showing up in the numbers before it shows up in the stock.

Balance sheet / net debt — why this matters: can the company survive a downturn and still fund its buildout?

(This "cash" line is cash & equivalents only, excluding Microsoft's large short-term investments portfolio — a fuller net-cash picture including investments would look more net-cash-positive than shown. Long-term debt has been roughly flat-to-shrinking even as the balance sheet has grown — Microsoft is self-funding its AI buildout largely from operating cash flow, not new borrowing, so far.)

Source: EDGAR company facts (data.sec.gov, CIK 0000789019), FY2025 10-K.

3. Unit Economics

So what: the number that matters most right now is Azure's growth rate and its capacity constraint — Microsoft is supply-constrained, not demand-constrained.

• Azure revenue growth: guided ~39-40% YoY in constant currency for FY2026 Q4 (gurufocus.com, microsoft.com IR).
• Azure AI backlog: Microsoft has disclosed roughly $80B of Azure orders it cannot yet fulfill due to power and chip constraints (globaldatacenterhub.com, valueaddvc.com) — the clearest "bottleneck, not demand" signal in the whole report.
• Overall operating margin: 45.6% in FY25, among the highest of any company at this revenue scale — pricing power in action.
• Capex guidance: ~$190B planned for calendar 2026 AI infrastructure, up 61% YoY (valueaddvc.com); a separately-cited fiscal-year figure runs near $120B+ for FY2026 (tipranks.com) — the two figures use different year definitions (calendar vs. fiscal), both directionally pointing to capex roughly doubling in ~2 years.

4. Metrics Table (with the "why")

So what: almost every metric here is elite for the company's size — the debate isn't about quality, it's about price.

4b. Sector Valuation Benchmarks

So what: Microsoft trades roughly in line with, or at a modest discount to, its mega-cap tech peers — not obviously cheap, not obviously the most expensive either.

Microsoft is profitable and large, so P/E is the right primary multiple here (EV/EBITDA/EV/Sales are more useful for capital-heavy or pre-profit names).

Discrepancy flagged: financial data sites quote MSFT trailing P/E around 22.8-23.9x using a TTM EPS near $16.6-16.8, while this report's computation from the FY2025 10-K ($101.832B net income ÷ 7.434B shares = $13.70 EPS) implies ~29x. The gap likely reflects those sites using a more recent trailing-twelve-month window (through a later FY2026 quarter, with higher EPS than full FY2025) and/or non-GAAP adjusted EPS. Directionally, MSFT sits roughly in line with or modestly below the broad IT sector's ~40x, whichever EPS base is used — treat the exact multiple as approximate, not precise, and this discrepancy itself as a lesson in why "always state which EPS you used."

5. Ownership & Management

So what: heavy institutional ownership and a CEO who has been a steady net seller — normal for a long-tenured executive, not a red flag on its own, but worth tracking in aggregate.

• Institutional ownership: ~71.7%; insiders ~6.2%; retail ~22.1% (wallstreetzen.com).
• CEO Satya Nadella: holds 786,875 shares as of 2026-07-01; has sold ~1.2M shares since 2021 for an estimated $430M, part of a broader pattern of 122 insider trades since 2021 (12,142 shares bought vs. ~2.1M sold) (ainvest.com, secform4.com).
• Reading it: net insider selling at a mega-cap with heavy stock-based compensation is common and not by itself a bearish signal — largely scheduled diversification (10b5-1 plans, defined below in the glossary), not a sudden vote of no confidence. Still, "insiders dumping despite AI hype" is a headline bears cite; there's no evidence here of unusual timing or cluster selling ahead of bad news.

6. M&A Track Record

So what: Microsoft's largest recent "acquisition" isn't a takeover — it's a deep capital and commercial partnership with OpenAI, structurally different from its past buyouts.

Historically: Activision Blizzard ($69B, closed 2023 — gaming), LinkedIn ($26B, 2016), GitHub ($7.5B, 2018), Nuance ($20B, 2022) — a track record of large, mostly successful integrations rather than value-destroying mega-deals. The OpenAI relationship (cumulative multi-billion-dollar investment, with coverage of the FY26 Q3 print referencing a ~$250B compute/commercial commitment — valueaddvc.com) is now the largest capital commitment in the company's history, run through Azure infrastructure rather than a balance-sheet acquisition.

7. Market & Value-Chain Position

So what: Microsoft is a bottleneck holder in enterprise software (high switching costs) but a price-taker on AI chips (Nvidia/TSMC hold pricing power over Microsoft's own capex).

• Enterprise software/cloud: one of 3 hyperscalers (with AWS and Google Cloud) and the dominant enterprise productivity suite — extremely high switching costs (an enterprise doesn't casually rip out Windows/Office/Azure). This is Microsoft's own bottleneck/moat.
• Upstream, Microsoft is a customer, not the bottleneck holder: it depends on Nvidia (and increasingly its own custom silicon) for AI accelerators, and indirectly on TSMC's advanced fabrication through Nvidia — the $80B unfulfilled Azure order backlog exists precisely because Microsoft can't get enough compute capacity built fast enough, a supply constraint largely outside its own control.
• Named competitors: AWS (cloud share leader), Google Cloud (fastest-growing hyperscaler by some measures), Salesforce/Workday (SaaS), Amazon/Google (search/ads, smaller overlap).

8. PE/Quality Lens

So what: a private-equity buyer would love the recurring revenue and margins, and would be nervous about the capex trajectory relative to near-term cash return.

What a buyer would love: 68.8% gross margins, 45.6% operating margins, near-zero net leverage, massive recurring-revenue base, genuine multi-year moat in enterprise software. What a buyer would fear: capex consuming an ever-larger share of operating cash flow with a multi-year payback profile that depends on AI monetization assumptions nobody can fully verify yet; if Azure AI margins prove structurally lower than Core Azure (the literal Morgan Stanley bear scenario), returns on the $190B 2026 capex program disappoint for years.

Bull / Base / Bear (attributed)

So what: the spread between named institutional price targets ($250 to $795) is unusually wide for a mega-cap — a signal that this is a genuine, unresolved debate, not a settled consensus.

• Bull — Morgan Stanley ($795 scenario target; more recent working "top pick" target $650): if Azure AI economics resemble Core Azure's profitability and Copilot drives material Microsoft 365 acceleration, with operating margins expanding toward ~49% by fiscal 2028 (techtimes.com, 247wallst.com).
• Base — Street consensus: average price target roughly $589-$592, "Strong Buy"/"Buy" from the majority of 60+ covering analysts (247wallst.com, finance.yahoo.com).
• Bear — Morgan Stanley's own downside scenario ($250) and Stifel's Brad Reback: if Azure AI monetizes more like commoditized bare-metal compute and Copilot adoption stalls, margins compress under capex weight before revenue catches up; Stifel specifically flags near-term Azure growth expectations as too optimistic (techtimes.com, 247wallst.com).

Catalysts: FY2026 Q4 earnings (~2026-07-29) Azure growth print; FY2027 capex guidance; any update on the Azure power/chip capacity constraint easing. What would prove the thesis wrong: Azure growth decelerating meaningfully below the 39-40% guided range for two consecutive quarters while capex stays elevated — that combination is the bear case's exact trigger condition.

Practitioner Decision Layer (Educational Framework)

This section is an educational decision framework, not investment advice or a trade instruction.

Reverse Expectations (multiple-and-consensus approximation, not a DCF)

So what: at ~29x earnings, the market is pricing in high-teens-or-better sustained growth for years — plausible given Azure's current trajectory, but with little room for a stumble.

At a ~29x trailing P/E (this report's FY25-based computation) against a company already growing revenue ~15%/yr with operating income growing faster, the price implies the market expects double-digit revenue growth sustained for several more years plus continued operating margin expansion (consistent with Morgan Stanley's bull case assuming margins reach ~49% by FY2028, vs. 45.6% today). Base rates: very few $2.5T+ companies sustain 12%+ revenue CAGR for 5+ years (Amazon and Apple are the rare precedents; most mega-caps decelerate to mid-single digits). Judgment: stretched-but-plausible — Azure's backlog and constrained-supply dynamic (§3) is real evidence for the bull case, but it requires several more years of AI capex translating cleanly into high-margin revenue, unproven at this scale by any company yet. Source dates: price and multiple as of 2026-07-21/22 web search; consensus targets as of 2026-07-17.

Variant Perception

1. Consensus believes: Microsoft's AI capex will pay off, Azure growth sustains near 40%, and the stock deserves its premium — reflected in a ~$589-592 average price target roughly 48% above the current price.
2. This report's differentiated view: Consensus agreement — no supported variant perception found. The evidence gathered here (Azure backlog, margin trend, capex trajectory) is consistent with, not contrary to, the bullish sell-side consensus.
3. Evidence: see §3, §4b, and the bull/base/bear table above.
4. Why the market may be wrong: not independently established here — the honest position is that this is a well-covered, heavily-analyzed stock where the real uncertainty (AI monetization timing) is openly debated by the Street itself, not hidden from it.

Probability-Weighted Expected Value

So what: even a probability-weighted blend of named scenarios lands well above today's price — but that reflects the sell-side's own optimism, not an independently verified forecast.

EV = (0.25 × $650) + (0.50 × $590) + (0.25 × $250) = $162.50 + $295.00 + $62.50 = $520.00

That's ~31% above the current $397.75 price — but this blend uses the sell-side's own scenario targets, and probability weights here are this report's estimate, not independently derived. Scenario discipline, not a forecast.

Pre-Mortem: It's 2028, the Stock Has Halved

So what: all three most-plausible "death" scenarios trace back to the same root cause — AI capex outrunning AI revenue.

1. Cause: Azure AI margins prove structurally lower than Core Azure (the literal Morgan Stanley bear trigger) — capex depreciation drags on operating margin for years. Early signal: Intelligent Cloud segment operating margin declining for 2+ consecutive quarters despite revenue growth. What would invalidate this risk: Microsoft disclosing Azure AI gross margins in filings that match or exceed Core Azure.
2. Cause: a rival (Google Gemini/Vertex, AWS Bedrock, or an open-source model wave) commoditizes the AI layer, making enterprises route AI workloads away from Azure-exclusive stacks, eroding Microsoft's take rate on its OpenAI relationship. Early signal: enterprise customer earnings calls citing multi-cloud AI strategies specifically to avoid Azure lock-in. What would invalidate: Microsoft 365 Copilot attach rates continuing to climb (a stickiness signal independent of the underlying model).
3. Cause: a demand-side AI spending pullback (enterprises pilot Copilot/AI tools, don't renew at scale — "AI ROI winter") stalls revenue while the $190B capex commitment is already locked in. Early signal: Azure growth deceleration below ~30% for two consecutive quarters. What would invalidate: continued backlog growth beyond the current $80B disclosed figure.

KPI Driver Tree & Sensitivity

So what: two numbers matter more than anything else in the whole report — Azure's growth rate and the operating margin on that growth.

Driver tree: Total revenue ≈ (M365/Productivity revenue × seat growth & ARPU) + (Azure revenue × growth rate) + (Windows/gaming/other). Earnings ≈ Revenue × Operating margin (a function of Azure AI gross margin vs. capex depreciation drag).

• If Azure growth rate moves ±10% (e.g., 40% guided → 36% or 44%): given Intelligent Cloud is ~40% of revenue, this is roughly a ±4 percentage-point swing on total company revenue growth — large for a company this size. Rough estimate, not a model.
• If operating margin moves ±10% relative (45.6% → ~41% or ~50%): given FY25 operating income of $128.5B, that's roughly a ±$12-13B swing in operating income — a double-digit percentage swing in earnings, arguably larger than the revenue-driven swing above. This is why margin trajectory (driven by capex depreciation timing) may be the more important variable to watch right now.

Scuttlebutt Proxies

So what: employee sentiment and hiring signals point to a company simultaneously investing heavily in AI talent and trimming elsewhere — consistent with the "reallocating, not simply growing" story management tells.

• Hiring-posting trend: unavailable as an independently sourced data point in this pass.
• Employee-review tone: Glassdoor overall rating 4.0/5 (68,387+ reviews), down ~2% over the trailing 12 months; 77% would recommend to a friend; a recurring theme of "forever layoffs" — small, rolling reductions — alongside reports that AI/skills roles and core product teams are comparatively insulated (glassdoor.com, inc.com). Reading: consistent with a company reallocating headcount toward AI priorities rather than broad-based distress.
• Customer/supplier earnings-call mentions: not captured in this pass — flagged as a gap rather than fabricated.

Decision (Educational — not investment advice)

• Decision: Watch
• Conviction: Medium
• Core assumption: Azure's ~40% growth rate and its $80B unfulfilled order backlog persist through at least two more quarters without meaningful deceleration or a disclosed drop in Azure AI gross margin.
• Review date: 2026-07-29 (estimated FY2026 Q4 earnings date)
• Controlling question: Did Azure constant-currency growth stay in or above the 39-40% guided range, and did management give any signal on Azure AI vs. Core Azure margin trajectory?
• Review trigger: Two consecutive quarters of Azure growth below ~30%, or explicit disclosure that Azure AI gross margins are structurally below Core Azure — either would validate the bear case and warrant re-underwriting the position.

📚 What this company teaches

1. Capex intensity can flip a "cash machine" into a "cash reinvestor" almost overnight — Microsoft's capex/OCF ratio nearly doubled (27% → 47%) in three years without a single acquisition, purely from organic AI infrastructure spend. Watch this ratio on any "quality compounder," not just growth stocks.
2. A backlog you can't fulfill is a different kind of good news than a backlog you can — Microsoft's $80B unfulfilled Azure orders is bullish (demand exceeds supply) but also means near-term revenue is capped by physical build-out speed, not by sales effort — a supply-chain/power-grid problem, not a go-to-market problem.
3. Wide analyst target dispersion ($250-$795) is itself a data point — when a heavily-covered mega-cap has that much spread between named bull and bear cases from the same bank, it signals the market genuinely doesn't know how a multi-year capital cycle will resolve, not that anyone is being sloppy.

Self-test question: If Microsoft's Azure AI segment margins come in permanently 10 percentage points below Core Azure's margins, roughly how would that affect total company operating income, given Intelligent Cloud is about 40% of revenue and Azure is the majority of that segment? (Hint: use the KPI sensitivity approach above.)

📖 Glossary

• P/E (price-to-earnings ratio): how many dollars you pay today for $1 of the company's most recent annual profit.
• EV/EBITDA (enterprise value ÷ earnings before interest, tax, depreciation, amortization): the whole-business price (equity + debt − cash) compared to a cash-flow-like earnings proxy; useful because it ignores how much debt a company carries.
• EV/Sales: enterprise value divided by revenue; used when a company isn't yet profitable, so P/E doesn't work.
• FCF (free cash flow): cash generated by the business after paying for equipment/buildings (capex) — the cash actually left over, as opposed to accounting profit.
• FCF yield: free cash flow divided by market value — like a "cash dividend rate" if the whole company paid out all its free cash.
• Capex (capital expenditure): money spent building or buying long-lived physical assets like data centers and servers.
• Net debt: total debt minus cash on hand; negative net debt ("net cash") means a company could pay off all its debt today and still have cash left over.
• ROE (return on equity): net income divided by shareholders' equity — how efficiently a company turns shareholders' invested capital into profit.
• Bottleneck (value-chain sense): a step in an industry's supply chain that few companies can do, giving whoever controls it outsized pricing power over everyone else in the chain.
• Backlog: contracted-but-not-yet-delivered orders; a form of revenue already "in the bag," just not yet recognized.
• 10b5-1 plan: a pre-scheduled, pre-announced stock sale plan executives use to sell shares on a fixed schedule, specifically to avoid the appearance of trading on inside information.
• TTM (trailing twelve months): the most recent four reported quarters combined, as opposed to a fixed fiscal year — explains why some data sources' EPS figures differ from this report's FY2025 fiscal-year figures.