Microsoft
Recurring enterprise SaaS subscriptions (M365), consumption-based cloud (Azure), per-seat AI add-ons (Copilot), plus licensing, gaming and advertising; ~75% commercial/recurring revenue.
The thesis on this name
State of AI Compute
The demand anchor — Azure + Copilot + the OpenAI stake make Microsoft the single largest AI-capex buyer, setting the pull-through for the whole compute value chain.
State of the AI Cloud
Demand vastly exceeds supply (~$80B unfulfillable Azure backlog) and Copilot/M365 is a captive high-margin AI distribution channel — but the OpenAI anchor loosened.
State of the AI Cloud
~$80B unfulfillable backlog signals demand>supply; Copilot/M365 is captive high-margin distribution — sized below GOOGL/AMZN for the loosened OpenAI anchor.
State of AI for Healthcare
Carried to rule it out — an avoid as a health-AI expression, and specifically NOT a short on the security. Microsoft is the supplier on both sides of the chokepoint: it provides the transcription inside Epic's own AI Charting, while Dragon Copilot separately runs at a reported ~33% ambient category share and 100,000+ clinicians daily at a reported list price of $369–604 per provider per month. That makes it the incumbent with the largest absolute dollar at risk from an EHR-native default and the best hedged against it — it keeps the inference volume while surrendering the customer relationship and most of the ARPU as a list price becomes an OEM component. None of it is expressible. Healthcare is not a reported Microsoft segment and no health or clinical-AI revenue line is disclosed at any level, while the entire contested US ambient pool — $3–6B/yr, itself a framing range derived from current list prices rather than a measurement — is 1.81% of FY2026 revenue of $331.8B (+18%) at its top end, against a $3.45T market cap at $464.72 (31 Jul 2026). A position has to be able to be right or wrong for the reason you took it, and this one cannot. Buying MSFT for health AI is buying Azure and Office with a rounding error attached — the most common of the discrepancies that reject the arms-dealer proxy book.
State of AI for Healthcare
Owns the ambient incumbent — Dragon Copilot at roughly 33% category share and 100,000+ clinicians daily — and supplies the transcription inside Epic's own AI Charting, which makes it the pick-and-shovel on both sides of the chokepoint and still un-ownable here: the whole contested US ambient pool of $3–6B/yr, a framing range built from list prices rather than a measurement, is 1.81% of TTM revenue at its top end.
State of Enterprise AI SaaS
Reference-only diversified arm: the incumbent best-hedged across BOTH sides of the disruption — it still sells M365 seats AND meters Azure/Copilot/Agent365 consumption, so seat erosion at one layer is recaptured as inference/agent consumption at another. The default 'own the platform shift without picking the winner' optionality for a seat-vs-consumption book.
State of Enterprise AI SaaS
The aggregator that wins either pricing regime — Azure consumption +40% AND Copilot seats (20M+ paid), $37B AI run-rate up 123%; seat-compression hits rivals, not the platform that sells the agents.
State of Enterprise AI SaaS
Owns both seat and consumption monetization; dampens book volatility while keeping platform-shift optionality.
State of Frontier AI
The listed OpenAI call option (equity + IP rights) plus Azure inference + 20M paid Copilot seats — de-rated in 2026 on the capex-vs-Copilot-monetization doubt, and that pessimism is the entry.
State of Frontier AI
The listed proxy for OpenAI exposure (equity + IP rights) plus Azure inference demand and 20M paid Copilot seats. De-rated in 2026 on capex-vs-Copilot-monetization doubt (Fortune May'26) — that pessimism is the entry. Owns the OpenAI call option without OpenAI's -122% operating margin.
State of Quantum Computing
Microsoft runs a two-track bet: its own moonshot topological/Majorana qubits PLUS Azure Quantum as a neutral marketplace aggregating other vendors' machines (e.g. Atom Computing GTM). Either way it monetizes the ecosystem. Quantum is a rounding error of optionality inside a ~$3T+ cloud/AI compounder — own MSFT on cloud/AI, treat quantum as a lottery ticket. Reference name; not a pure-play.
State of Quantum Computing
Two-track bet: own topological/Majorana moonshot PLUS Azure Quantum as the neutral multi-vendor aggregator (Atom Computing GTM). Monetizes the ecosystem either way; own MSFT on cloud/AI, quantum is a lottery ticket.
Earnings, margins, COGS & capex
Q3 FY26 (ended 3/31/2026) beat: revenue $82.9B (+18%) vs ~$81.5B consensus, operating income $38.4B (+20%), net income $31.8B (+23%), diluted EPS $4.27 (+23%). Azure +40% (+39% cc) is the growth engine; AI business surpassed a $37B annual run-rate (+123% YoY). Commercial RPO (backlog) nearly doubled to $627B. The market sold off not on the print but on the capex guide — calendar-2026 capex lifted to ~$190B (incl. ~$25B of memory/component-price inflation), pressuring FCF and gross margin.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~33¢ is cost of goods and ~22¢ operating expense, leaving ~45¢ of operating profit.
Revenue trend
Margins
Declining — datacenter depreciation + AI infra drag; lowest since 2022
Up YoY (45.7%→46.3%), resilient despite capex
Down YoY on AI capacity build + rising AI usage, partly offset by Azure/M365 efficiency
Falling sharply as capex outruns operating cash flow ($46.7B OCF − $31.9B capex)
COGS structure
COGS rising faster than revenue, driven by datacenter depreciation (GPUs/CPUs, ~2/3 of capex is short-lived assets), server/network gear, and OpenAI/third-party model compute costs; this is the structural reason gross margin is compressing.
Capex
Q3 FY26 capex $31.9B; guided to >$40B next quarter (incl. ~$5B component-price inflation); calendar-2026 ~$190B total (+~61% YoY), of which ~$25B is higher memory/component pricing from the 2025-26 AI-driven DRAM/HBM shortage.
Latest earnings
Beat on both lines: revenue $90.01B vs ~$87.62B consensus, adjusted EPS $4.74 vs $4.24 est. (LSEG; another aggregator carried $4.33, so the beat magnitude is approximate). GAAP diluted EPS $4.81 (+32%); non-GAAP $4.74 (+23%, excludes the OpenAI investment impact). The stock reaction INVERTED the prior quarter's: +15.51% on Jul 30 ($390.54 -> $451.10), then +3.02% to $464.72 on Jul 31 — the ~$190B capex overhang the entry named as the central bear catalyst was restated down to ~$175B. Quality caveat: Microsoft flagged the $3.2B Anthropic investment mark plus lower VRP expense as a $0.27 EPS benefit vs its own April guidance, so the headline beat is lower quality than it first appears.
Q1 FY27 revenue guided $89.85B-$90.95B (+16-17%); Azure ~45% constant-currency growth (accelerating from +43% in Q4 FY26); Intelligent Cloud +33-34%; Q1 FY27 capex >$50B (inflated by the lease reclassification). FY27 full year: double-digit revenue AND operating-income growth, operating margin down less than one point, free cash flow to remain positive. Calendar-2026 capex expectation revised to ~$175B from ~$190B purely from the useful-life / finance-to-operating-lease accounting change, with underlying investment plans explicitly unchanged. [Guidance figures come from the earnings call via secondary sources — the 8-K states only that guidance was given on the call. No verified full-year FY27 capex dollar figure exists.]
- Azure & cloud services growth
- +43% (Q4 FY26); Azure crossed $100B in annual revenue for the first time in FY26 (+41% YoY); Q1 FY27 guided ~45% constant currency — accelerating, not decelerating
- Microsoft Cloud revenue
- $59.3B (+27%) in Q4 FY26; ~$214B for FY26 (+27%)
- Commercial RPO (backlog)
- $678B (+84% YoY, incl. OpenAI) — up $51B sequentially, though the growth rate decelerated from +99% on a harder comp
- Microsoft 365 Copilot paid seats
- >30M paid seats (newly disclosed), with net seat adds more than doubling quarter over quarter; the new E7 suite showed early traction with usage intensity reported as comparable to Outlook and Teams
- Diluted EPS
- $4.81 GAAP (+32%) / $4.74 non-GAAP (+23%) in Q4 FY26; $17.95 GAAP (+32%) / $17.28 non-GAAP (+22%) for FY26, on 7,443M diluted shares
- Capex
- $35,802M cash capex in Q4 FY26 (calendar-2026 expectation ~$175B, restated from ~$190B on the accounting change, underlying spend unchanged)
- AI investment marks (other income)
- +$3.2B gain on the Anthropic stake in Q4 FY26 vs a ~$480M NET DRAG from OpenAI in the same quarter (-$0.07 EPS); OpenAI contributed +$4.963B / +$0.67 EPS across FY26
- Datacenter footprint
- 31 new datacenters added across five continents in Q4 FY26, 88 across FY26; property & equipment net $313.1B, up from $205.0B YoY
Growth drivers
- Azure consumption (+40%), the single largest growth lever, still capacity-constrained through 2026
- AI run-rate $37B (+123% YoY) — Copilot seats, Azure OpenAI/Foundry inference, Fabric/data
- M365 Commercial seat growth + Copilot per-seat upsell into ~400M+ paid seats
- $627B commercial RPO backlog (incl. OpenAI's $250B Azure commitment) underwriting multi-year cloud revenue
- Security (Defender/Entra/Sentinel) and GitHub/dev-tools attach
Reported financials — SEC EDGAR
Audited GAAP figures pulled from SEC filings · latest filing 2026-07-29. The audited primary-source spine — not financial advice.
Revenue — annual (GAAP)
Margins & balance sheet — FY’26
Bull & bear
At $464.72 / $3.450T and 23.71x forward EPS (Jul 31 2026), the franchise just proved the thing the capex debate hinged on: Azure accelerating to +43% at a $100B+ annual base with ~45% constant-currency guided, FCF margin EXPANDING to 21.8% in a record $35.8B capex quarter, $678B commercial RPO, and >30M paid M365 Copilot seats with net adds more than doubling QoQ. The build is being funded out of a growing operating-cash engine, not against it.
- Azure grew +43% in Q4 FY26 and crossed $100B of annual revenue for the first time (FY26 +41%), with Q1 FY27 guided ~45% constant currency — accelerating at a scale where deceleration is the norm
- FCF margin EXPANDED to 21.8% in Q4 FY26 ($19,639M on $90,007M revenue: $55,441M operating cash flow less $35,802M capex); FY26 FCF $66,987M (20.2%) — operating cash grew faster than a record capex line
- Commercial RPO $678B (+84% YoY, incl. OpenAI), up $51B sequentially — multi-year demand visibility underwriting the spend
- Microsoft 365 Copilot >30M PAID seats newly disclosed, with net seat adds more than doubling quarter over quarter; the E7 suite showed early traction with usage intensity reported comparable to Outlook and Teams — monetized seats, not pilots
- Two-horse frontier-model book: the ~27% OpenAI stake (still accurate) plus an up-to-$5B Anthropic position taken Nov 2025 alongside Anthropic's $30B Azure purchase commitment; Anthropic marked +$3.2B in Q4 FY26
- FY26 delivered $331.8B revenue (+18%, +16% cc) at a 46.8% FY26 operating margin, operating income +21% — growth and margin held through the build year
- FY27 guide is for double-digit revenue AND operating-income growth with operating margin down less than one point and FCF positive (guidance figures from the earnings call via secondary sources)
- MSFT decoupled from the AI-hardware complex: up ~25% over the July 2026 window in which the Philadelphia Semiconductor Index fell 20.6% (its worst month since Oct 2008, -28% from its Jun 22 2026 high) — the market is now discriminating AI capex spenders with visible monetization from AI hardware suppliers
- Still net cash (~$36.5B: $76.8B cash + short-term investments vs $40.3B total debt) — the build does not require balance-sheet stress
Only the accounting got cheaper. Calendar-2026 capex was restated ~$190B → ~$175B purely by extending datacenter/office useful life 15→25 years and shifting some future datacenter leases from finance to operating — management explicitly said underlying investment plans are unchanged, and Q1 FY27 capex is guided >$50B. Gross margin compressed a further ~40bps to 67.20%, the cash cushion fell 18.7% YoY to $76.8B, ~$0.27 of the EPS beat was discrete, and the ~19x valuation cushion that was the entry's margin of safety is gone at 23.71x forward.
- The $15B capex 'cut' is optical, not discipline: Hood said calendar-2026 investment expectations are unchanged, Q1 FY27 capex is guided >$50B, and Q4 FY26 cash capex was a record $35,802M = 39.8% of revenue (FY26 $115,948M = 34.9%). (Calendar-2026 vs fiscal-2027 attribution rests on a two-source majority plus Hood's quoted phrasing, not the primary transcript.)
- Extending useful life 15→25 years flatters FUTURE depreciation and therefore reported margins — a quality-of-earnings item, not an efficiency gain, sitting directly on top of our margin-compression thesis
- The headline beat is lower quality than it reads: Microsoft flagged the $3.2B Anthropic mark plus lower VRP expense as a ~$0.27 EPS benefit vs its own April guidance (GAAP EPS $4.81 / non-GAAP $4.74 vs ~$4.24 consensus, LSEG — one aggregator carried $4.33, so beat magnitude is approximate)
- Other income is now a two-way swing: OpenAI was a ~$480M NET DRAG (-$0.07 EPS) in the same quarter Anthropic added $3.2B, against OpenAI's +$4.963B / +$0.67 EPS across FY26 — diversification bought volatility
- Gross margin 67.20% in Q4 FY26 ($60,482M / $90,007M), down ~40bps further (FY26 67.94%), as property & equipment net rose to $313.1B from $205.0B and 88 new datacenters landed in FY26 (31 in Q4)
- Operating margin 45.11% in Q4, down ~120bps sequentially from Q3 FY26's 46.3%; FY27 guided down less than one point — the direction is still down
- Cash + short-term investments $76.8B at Jun 30 2026, DOWN 18.7% from $94.6B a year earlier as capex outran cash generation; net cash ~$36.5B excludes finance-lease liabilities that grow with the datacenter build
- The valuation cushion is spent: forward P/E 23.71 / trailing 25.89 against the ~19x that followed the June drawdown; the stock traded up through the $420-460 band in two sessions
- UK CMA opened a Strategic Market Status investigation (May 15 2026) into Microsoft business software — Windows, Word, Excel, Teams, Copilot — on bundling, interoperability limits and AI lock-in, with a designation decision due Feb 2027 and power to impose conduct requirements: aimed squarely at the M365 + Copilot attach motion the bull case depends on
- RPO growth RATE decelerated from +99% to +84% on a harder comp — the level is higher, the second derivative is not
- Capex ROIC is still unproven over a multi-year horizon; one quarter of FCF expansion is evidence, not a cycle
What it is worth
Comps + reverse-DCF sanity check, re-anchored to the post-print base (Jul 31 2026) rather than adjusted from the prior ladder
~$350-400
back toward the pre-print $390.54 or the 52-week low of $349.20: underlying spend (unchanged behind the restated ~$175B calendar-2026 headline; Q1 FY27 >$50B) outruns monetization, gross margin keeps sliding from 67.20%, the market discounts the 15->25 year useful-life benefit as accounting rather than earnings, and/or a Feb-2027 CMA conduct remedy bites the M365 + Copilot attach. Roughly 20x or below on forward EPS. Our prior $300-345 bear band — with $345 as the stated bull/bear pivot — was never tested and is superseded by the higher base.
~$465-520
roughly the current price to a modest premium: FY27 guidance delivers (double-digit revenue and operating-income growth, operating margin down <1pt, FCF positive), multiple holds ~23-25x forward as capex ROIC is shown gradually rather than proven, and the CMA process ends in remedies Microsoft can absorb.
~$640-650
the post-print sell-side band (Goldman Sachs $640, raised from $610, Buy; Wells Fargo $650; Bernstein $647, from $646, Outperform). Requires Azure holding near the guided ~45% cc, Copilot paid seats compounding off the 30M base, and FCF margin holding ~20%+ through the FY27 capex peak. Note our prior bull case (~$520-560, stored analyst average ~$561) now sits BELOW the market price and below several post-print targets — the ladder was rebuilt, not nudged.
At $464.72 / $3.450T (Jul 31 2026 close; cross-check 7.43B diluted shares x $464.72 = $3.453T) MSFT trades at 23.71x forward and 25.89x trailing EPS, against FY26 GAAP EPS $17.95 (+32%) / non-GAAP $17.28 (+22%). The ~19x forward multiple that followed the June drawdown was the margin of safety that made the setup skew favorable, and that leg has substantially CLOSED — the stock rose +15.51% on Jul 30 ($390.54 -> $451.10) and +3.02% to $464.72 on Jul 31, +25.3% from its Jun 26 level, trading through the ENTIRE $420-460 base case into the lower band of the bull case. MSFT can no longer be called the cheapest of the megacap-cloud trio: the GOOGL ~24x / AMZN ~28x comparators behind that claim are unverified, so the claim is not made here. What the print did to the underlying debate: demand and cash conversion were demonstrated (Azure +43%, FCF margin 21.8%), so what remains to be priced is the multi-year cost of the build (underlying capex unchanged; Q1 FY27 >$50B) and the Feb-2027 CMA designation risk. Sell-side targets below are cited as sourced datapoints, not endorsements — this is analytical critique, not a price recommendation.
SWOT
Strengths
- Azure at $100B+ annual revenue still growing +43% (Q4 FY26 — FY26 +41%) and guided ~45% constant currency for Q1 FY27
- FY26 revenue $331.8B (+18%) at a 46.8% operating margin with $155.2B operating income (+21%) and $67.0B FCF (20.2%) — the AI build is self-funded from operations
- $678B commercial RPO (incl. OpenAI) giving multi-year revenue visibility
- Widest enterprise AI distribution surface in software — >30M paid M365 Copilot seats with net adds more than doubling QoQ, on top of the M365/Windows install base
- Two-horse frontier-model book (~27% OpenAI stake — up-to-$5B Anthropic position plus Anthropic's $30B Azure commitment) plus in-house MAI models — not single-vendor dependent
- Still net cash (~$36.5B) with $76.8B cash + short-term investments against $40.3B total debt
Weaknesses
- Gross margin 67.20% in Q4 FY26, compressing a further ~40bps as depreciation builds against $313.1B of net property & equipment (from $205.0B YoY)
- Cash cushion thinning fast — $76.8B cash + short-term investments, down 18.7% YoY, now under the ~$80B mark
- Q4 operating margin 45.11%, -120bps sequentially; FY27 guided down less than one point
- Earnings quality — ~$0.27 of the Q4 EPS beat came from the $3.2B Anthropic mark plus lower VRP expense, and the 15→25 year useful-life extension flatters future reported margins
- Capex intensity 39.8% of revenue in Q4 FY26 (34.9% for FY26), with Q1 FY27 guided >$50B and underlying calendar-2026 plans explicitly unchanged
- Consumer/standalone Copilot position is UNMEASURED — there is no verifiable MAU figure to set against ChatGPT or Gemini, and the enterprise paid-seat metric does not substitute for it
- The latest disclosed AI business run-rate (>$37B, +123% YoY) is a Q3 FY26 datapoint — Microsoft did not repeat it in the Q4 8-K, so there is no current-quarter AI revenue figure
Opportunities
- >30M paid Copilot seats is still a small fraction of the M365 commercial base; E7 suite early traction with usage intensity reported comparable to Outlook and Teams
- Capacity coming online against the constraint — 88 new datacenters in FY26 (31 in Q4, five continents), the first Fairwater datacenter (Mount Pleasant, WI) fully operational Jun 23 2026, and a 20-year Chevron power agreement for a new 2.67GW West Texas gas plant ('Project Kilby')
- Build-efficiency partnerships — Azure is the first announced hyperscale cloud to deploy 3M's Expanded Beam Optical interconnect (Jul 15 2026) — a concrete datapoint on the 'capacity online faster' claim that underwrites the ROIC debate
- A second anchor tenant — Anthropic's $30B Azure purchase commitment alongside OpenAI's $250B commitment (the OpenAI figure is unverified)
- Continued re-rating if the market keeps classifying MSFT as an AI monetizer rather than an AI-capex spender — the July 2026 decoupling (MSFT ~+25% vs SOX -20.6%) is the first evidence of that reclassification
Threats
- UK CMA Strategic Market Status designation decision due Feb 2027, with power to impose conduct requirements on bundling, interoperability and AI lock-in across Windows/Word/Excel/Teams/Copilot
- Capex ROIC unproven at scale — underlying calendar-2026 spend unchanged (~$175B stated after the reclassification), Q1 FY27 >$50B, against a depreciation wave already visible in the 67.20% gross margin
- The useful-life extension is a timing benefit — if revenue does not scale into the asset base, the flattered depreciation line only defers the margin problem
- AI-hardware complex stress — SOX -20.6% in July 2026 (worst month since Oct 2008), -28% from its Jun 22 2026 all-time closing high of 14,634.72 to ~10,447.49 — component cost and supply risk to the build persists even though the stock decoupled
- OpenAI co-opetition — a ~$480M net drag in Q4 FY26, free to use rival clouds, and a direct app-layer rival even as Microsoft funds its compute
- Governance/board turnover — Reid Hoffman, a director since 2017, informed Microsoft on Jun 2 2026 he would not stand for re-election
Moats, dependencies & bottlenecks
Moats
Embedded in enterprise workflows + identity (Entra); switching costs measured in years. AI rides this rail for free.
Data, identity, and integrations on Azure/M365 are sticky; backlog $627B reflects multi-year commitments.
Capex scale (~$190B/yr) is a barrier few can match — but AWS, Google, and Oracle are all building at scale too.
Developer + professional graphs feed AI products and partner channel.
Default vendor for regulated enterprises; security suite deepens lock-in.
Dependencies
Strategic AI model + ~27% equity stake + co-opetition ~27% stake worth ~$135B; OpenAI committed $250B to Azure but exclusivity ended — can now use rival clouds and competes in apps. 20% revenue share to MSFT until AGI (panel-verified).
GPUs are ~2/3 of capex and the binding capacity constraint; pricing/allocation power sits with Nvidia. MSFT building Maia silicon to hedge.
2025-26 memory shortage adds ~$25B to 2026 capex; an exogenous cost shock.
Upstream supplier (via Nvidia + in-house Maia) Leading-edge fab capacity underpins all AI silicon Microsoft buys/builds.
Cloud + Copilot spend is discretionary at the margin; a downturn slows Azure consumption.
Advantages
- Largest enterprise distribution surface in software for AI upsell
- #2 cloud growing fastest at scale (+40%)
- Diversified recurring revenue + 46% operating margin + AAA balance sheet to self-fund the build
- Multi-model AI optionality (OpenAI + Anthropic + MAI)
- $627B contracted backlog de-risking near-term cloud revenue
Weaknesses
- Margin + FCF compression from the AI capex super-cycle
- Copilot consumer/agent traction lagging Google + OpenAI
- Component-cost exposure (memory/GPU) it cannot fully control
- Co-opetition + dependence on OpenAI for frontier capability
- Capacity unable to meet Azure demand
Bottlenecks
- Datacenter power + capacity — Azure is supply-constrained through 2026; demand exceeds buildable capacity
- GPU allocation from Nvidia — gates how fast AI capacity comes online
- Memory/HBM shortage inflating cost and limiting builds (~$25B impact)
- Skilled-labor + datacenter construction lead times
- Copilot consumer adoption — a go-to-market/product bottleneck, not a supply one
Top signals & trends
Top signals
Demand is the strongest in the megacap-cloud cohort at this scale.
Headline calendar-2026 capex is now ~$175B, cut from ~$190B purely by an accounting change (datacenter/office useful life 15->25 years; some future leases moving finance->operating) — management explicitly said underlying investment plans are unchanged, and Q1 FY27 capex is guided >$50B, so this is NOT capex discipline. Two things did break the original bear mechanism: FCF margin EXPANDED to 21.8% in Q4 FY26 despite record capex ($35.8B), and the market re-rated the stock +25% rather than punishing the spend. What remains bearish: the underlying build continues (88 new datacenters in FY26, PP&E net $313.1B vs $205.0B), the cash cushion is thinning ($76.8B, -18.7% YoY), and the useful-life extension flatters future depreciation and therefore reported margins — a quality-of-earnings item.
Backlog nearly doubled — multi-year revenue visibility underwriting the spend.
Structural AI-mix margin dilution + depreciation.
Forward P/E 23.7x / trailing 25.9x after the +25% re-rating (Jul 31 2026) — the valuation-discount leg of the bull case has substantially closed. Whether MSFT is still cheapest of the megacap-cloud trio is unverified: the entry's GOOGL ~24x / AMZN ~28x comparators were not re-checked and may themselves be stale.
Microsoft newly disclosed Microsoft 365 Copilot >30M PAID SEATS with net seat adds more than doubling QoQ — enterprise monetization is the counter-evidence to the app-layer mindshare gap, and it is not comparable to a consumer-chatbot MAU number. The ~150M Copilot MAU figure and the ChatGPT 900M / Gemini 750M comparators were not independently confirmed and should be treated as unverified until refreshed.
Trends
Inference + agent workloads drive Azure consumption + Copilot seats.
Drives Azure growth but compresses near-term FCF/margin; ROIC unproven at $190B/yr scale.
~$25B cost inflation + capacity constraint.
More model choice on Azure, but OpenAI free to use rivals.
Defender/Entra/Sentinel take point-vendor share.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Primary AI GPU supplier; ~2/3 of capex, the binding capacity constraint.
Alternative AI accelerators (MI-series) + EPYC CPUs; Microsoft's second-source hedge.
DRAM/HBM memory — at the center of the shortage adding ~$25B to 2026 capex.
Networking silicon + custom-ASIC partner for hyperscale datacenters.
Leading-edge foundry behind Nvidia GPUs + Microsoft's in-house Maia silicon.
$250B Azure commitment — Microsoft's largest AI compute customer + partner.
Global SI deploying M365 Copilot + Azure across enterprise clients; demand-amplifier.
~400M+ paid M365 seats + Azure consumers; Copilot upsell base.
Cloud leader ~28-30% share; AWS grew ~19% (slower than Azure) but largest scale. Mkt cap ~$2.44T, fwd P/E ~28x.
Google Cloud +63% (fastest), vertically integrated TPU+Gemini; Gemini 750M MAU pressures Copilot. Mkt cap ~$4.2T, fwd P/E ~24x.
OCI +93% YoY, aggressive AI-infra pricing + RPO; smaller (~$440B cap) but fast-growing challenger.
Competes in enterprise SaaS + AI agents (Agentforce) vs M365 Copilot/Dynamics.
Enterprise workflow + AI agents overlapping Microsoft's automation push.