Microsoft · NASDAQ: MSFT · Post-earnings decision memo
Azure accelerated. The capital bill still decides the stock.
FY2026 Q4 delivered a broad operating beat, 43% Azure growth and an 84% surge in commercial backlog. The unresolved question is whether AI monetization can outrun cloud-margin compression, rising depreciation and a $329 billion future datacenter lease pipeline.
Reported July 29, 2026Quarter ended June 30, 2026Post-market snapshot: $420.30 at 6:35 PM ETEvidence through 6:35 PM ET, July 29, 2026FY2026 10-K filed
Raise the business view; do not chase the first reaction.
Revenue beat the common pre-print bar by 2.7%, Azure growth beat consensus by 2.5 points, and every segment exceeded the midpoint of Microsoft’s April outlook. Commercial remaining performance obligations rose 84% to $678 billion, while Microsoft Cloud revenue grew 27%. S1S3S5
Headline EPS was less clean than the operating result. Microsoft reported $4.81 GAAP and $4.74 excluding OpenAI, but said Anthropic gains, retirement-cost timing, severance and an Xbox impairment produced a net $0.27 benefit versus its April guide. The resulting ~$4.47 guide-comparable proxy is analyst-derived, yet still about 5% above the $4.24 Street bar. S1S5
The 45% constant-currency Azure outlook for FY2027 Q1 is the strongest new evidence: it implies another sequential acceleration despite supply remaining below demand. The stock deserves a better operating assessment. At the same time, Microsoft Cloud gross margin is 65%, reported calendar-2026 capex is lower only because of lease accounting, and the post-call price already moved 7.6%.
Beat a low bar; Windows and Xbox still declined. S1
What changed: the debate is no longer whether demand is sufficient. It is whether Microsoft can convert an unusually large, partly concentrated backlog into revenue and cash returns before depreciation, energy and lease obligations overwhelm the margin narrative.
02 · Earnings quality
The revenue beat is cleaner than the EPS headline
$4.81GAAP diluted EPS
−$0.07OpenAI investment net gain
$4.74Company non-GAAP, ex-OpenAI
~$4.47Less $0.27 net discrete benefit; analyst proxy
Why the adjustment matters
Microsoft’s non-GAAP definition removes only OpenAI investment effects. It does not remove the $3.2 billion Anthropic gain or the net effect of retirement, severance and Xbox impairment items. Management disclosed the aggregate $0.27 EPS benefit but not a component-by-component after-tax bridge. S1
The ~$4.47 proxy simply subtracts that disclosed aggregate from $4.74. It avoids double-counting OpenAI and preserves the company’s own framing, but remains an inference rather than an audited adjusted figure.
Why the print still clears the bar
The proxy is about 5.4% above the common $4.24 Street expectation. More importantly, revenue, operating income and all three segments exceeded guidance. The operating beat survives the quality screen even though the headline EPS surprise shrinks materially.
Consensus definitions vary across public providers. This report uses $4.24 for the current-quarter comparison and labels the five-quarter chart as provider-reported rather than perfectly accounting-consistent. S5S6S8
Fallback: FY25 Q4 $3.65 vs $3.38; FY26 Q1 $4.13 vs $3.66; Q2 $4.14 vs $3.92; Q3 $4.27 vs $4.07; Q4 $4.74 vs $4.24. Series is provider-reported and Microsoft’s adjusted basis changed during FY2026. S1S8
03 · Financial trend
Growth accelerated while infrastructure absorbed more cash
Fallback: revenue rose from $76.4B in FY25 Q4 to $90.0B in FY26 Q4. Operating margin stayed near 45–49%; GAAP net income is distorted by OpenAI marks in FY26 Q1–Q2 and an Anthropic gain in Q4, so it should not be read as a smooth operating series. S1S4
Top-line cadence
Q4 revenue grew 18% reported, matching the full-year growth rate. Intelligent Cloud grew 32%, far ahead of PBP at 14% and MPC at −4%. S1
Operating leverage
Q4 operating income rose 18% to $40.6B and operating margin was 45.1%, essentially flat year over year despite the AI infrastructure build. S1
Cash conversion
Q4 CFO was $55.4B, but $35.8B of cash PP&E additions left $19.6B of simple FCF. FY2026 simple FCF was $67.0B versus $115.9B of cash capex. S1
04 · Operating drivers
The AI engine accelerated; the application layer became more mixed
Segment revenue and operating margin
Fallback: PBP $37.8B revenue / 57.9% operating margin; Intelligent Cloud $39.3B / 40.6%; MPC $12.9B / 21.4%. Margins are analyst-derived from official segment revenue and operating income. S1
Five-quarter cloud KPI trajectory
Fallback: Azure growth rose from 39% to 43% over the last five reported quarters, while Microsoft Cloud gross margin fell from 68% to 65%. S3
Azure / infrastructure
43% growth and FY2026 Azure revenue above $100B are the central upside signals. The result beat both guidance and consensus. S1
Copilot / applications
M365 Copilot exceeded 30 million paid seats, but M365 Commercial cloud growth slowed to 14% reported and Dynamics 365 slowed to 13%. Paid seats alone do not establish usage or incremental margin. S1S3
Consumer / gaming
Windows OEM and Devices fell 7%; Xbox content and services fell 10%. The Xbox impairment also reduced comparability, and MPC remains the least important piece of the AI thesis. S1
05 · Forward guide
The outlook raises the top-line bar; capex accounting lowers the headline, not the spend
Metric
Management outlook
Relevant prior bar
Read
Source
FY2027 Q1 revenue
$89.85B–$90.95B
$89.71B pre-print consensus
Midpoint is about 0.8% above the prior Street bar; 16–17% growth.
Q1 revenue at the midpoint is about 0.8% above the pre-print consensus, Azure is expected to accelerate to roughly 45% in constant currency, and Intelligent Cloud should grow 33–34%. Management also expects another year of double-digit revenue and operating-income growth.
Guide burden
The guide still absorbs heavy infrastructure costs: Q1 COGS grows 23–24%, fiscal-year capex grows, and FY2027 operating margin is expected to decline by less than one point. Windows, Devices and gaming remain material offsets.
Capital discipline test: The calendar-2026 capex expectation moved from about $190B to $175B because future datacenter leases will shift from finance to operating classification. Management said the underlying investment plan is unchanged, expects fiscal-year capex to grow, and guided Q1 capex to more than $50B. Lower reported capex is therefore not evidence of lower economic capital intensity.
Accounting lens: beginning in FY2027, Microsoft will extend the useful life of datacenter and office buildings from 15 to 25 years and expects future datacenter leases to be classified as operating rather than finance leases. Management said the operating-income benefit from the useful-life change should be minimal, while the lease change reduces reported capex without reducing the underlying capacity plan. S9
06 · Transcript and Q&A
The questions investors should carry into the next quarter
The summaries below paraphrase the official FY2026 Q4 call rather than reproducing the transcript. Bias labels and falsifiers are analyst judgments.
Karl Keirstead · UBS · Multi-model architecture
How does Microsoft benefit as enterprises use frontier, open-weight and custom models rather than standardizing on one provider?
Satya Nadella and Amy Hood: Enterprise context, memory and governance should sit outside any one model so models can be swapped. Customers will mix frontier, open and custom models; Azure can monetize the workload regardless of the selected model family.
BullishInvestor read: Azure is being positioned as the neutral control plane, reducing dependence on a single frontier lab. Falsifier: Model choice fails to produce Azure consumption or customers bypass Microsoft's data, orchestration and governance layers. S9
Brent Thill · Jefferies · Azure acceleration and capacity
What drives Azure from 43% growth to roughly 45% in Q1: easing constraints, better execution or demand?
Amy Hood: Demand continues to exceed supply. Small gains in CPU and GPU fleet efficiency and faster capacity deployment are immediately monetized, and Microsoft's scale makes those improvements material.
BullishInvestor read: The acceleration is demand-backed and aided by execution, rather than a temporary price or mix artifact. Falsifier: Azure decelerates sharply once the first fleet-efficiency gains are exhausted or newly deployed capacity fails to fill. S9
Mark Moerdler · Bernstein Research · Overbuild and component inflation
How does Microsoft protect returns if AI capacity is overbuilt while accelerator and component prices rise?
Amy Hood and Satya Nadella: Short-lived CPU and GPU assets are the largest capex component and purchases can be slowed. Land and datacenter builds can be staged, while a broad customer, geography and workload mix lets Microsoft late-bind capacity. Contract pricing, product mix and efficiency must preserve customer ROI.
MixedInvestor read: The portfolio is more flexible than the headline commitment stack suggests, but component inflation can still pressure price and margin. Falsifier: Utilization weakens, impairments rise or long-dated lease obligations prove less flexible than management describes. S9
Adam Wood · Morgan Stanley · M365 Copilot monetization
How are Copilot pilots converting into deployments, usage and higher monetization through seats, premium SKUs and consumption?
Satya Nadella and Amy Hood: Chat, Cowork and agentic features are converging into an enterprise super-app; time from license purchase to usage has fallen from months to days. Seat plus usage pricing, E7 and Agent 365 expand ARPU and the addressable pool, while token-spend governance becomes part of customer value.
BullishInvestor read: Microsoft is broadening Copilot from an add-on seat into an operating layer with multiple monetization paths. Falsifier: Paid seats rise without durable usage, renewal uplift, ARPU growth or measurable customer outcomes. S9
Brad Zelnick · Deutsche Bank · Agentic cybersecurity
What do new frontier models and Project Perception mean for Microsoft's security business and customer trust?
Satya Nadella: Microsoft is applying red-team, blue-team and remediation agents across security signals, while preserving a multi-model approach for cost and resilience. The combined identity, endpoint, network and application context should improve detection and response.
BullishInvestor read: Security can become both a direct AI monetization surface and a control point for broader enterprise-agent adoption. Falsifier: Agentic security creates new trust failures, fails to improve outcomes or cannot differentiate from stand-alone security platforms. S9
Gabriela Borges · Goldman Sachs · AI return on investment
How has management's view of AI ROI and monetization changed over the last year, and what still drives incremental economics?
Amy Hood: Confidence now comes from a larger addressable market and more margin levers: application and infrastructure efficiency, first-party silicon, model diversification, token efficiency, agent monetization and hyperscale operations. Management framed improvement as continuous execution rather than one discrete unlock.
MixedInvestor read: The ROI framework is broader and more credible, but remains qualitative while cash capital intensity is rising. Falsifier: The additional usage and TAM fail to lift lease-adjusted cash returns or cloud margin after depreciation and operating costs. S9
07 · Cash, commitments and concentration
The P&L is absorbing AI better than the cash-flow statement
Cash bridge
Fallback: Q4 CFO $55.4B minus cash PP&E additions $35.8B equals $19.6B simple FCF. This excludes finance leases and future lease commencements. S1
Commitment stack
$329.1B of datacenter-heavy leases had not yet commenced at June 30; they begin across FY2027–FY2033 and can run 1–20 years. S2
$34.6B was committed for new buildings and improvements, primarily datacenters. S2
$26.7B of PP&E purchases remained in accounts payable, up from $6.9B a year earlier. S2
Microsoft reset calendar-2026 reported capex from about $190B to $175B because future datacenter leases will be presented as operating leases. Management explicitly said the underlying investment expectation did not change and guided FY2027 Q1 capex to more than $50B. S9
Cash and short-term investments were $76.8B; current and long-term debt was about $40.3B. S1S2
Concentration is now visible: Microsoft disclosed $24.1B of FY2026 revenue from OpenAI, a $6.0B OpenAI receivable, an approximately 25% ownership interest, and $13.0B of funding commitments. Commercial RPO has 2.3 years of weighted duration and only about 30% is expected to convert in the next 12 months. Backlog quality therefore depends on counterparty durability, capacity delivery and long-dated economics—not just the headline growth rate. S2
08 · Tape and valuation
The reaction was positive; the valuation still prices successful execution
One-year price path
Fallback: MSFT closed at $390.54 on July 29, 2026, down 23.9% over the trailing year. The one-year range was roughly $349–$555. The post-market snapshot was $420.30 at 6:35 PM ET, +7.6% versus the regular close. S7
Post-market tape
Fallback: the post-market move was volatile around the release and call; the last observation used here is timestamped rather than presented as a final closing price. S7
FY27 P/E
~21.7×
Post-market price / pre-print FY27 EPS of $19.38 S8
Tape caveat: July 29 was also an FOMC decision and press-conference day. The regular-session move and some post-market volatility should not be attributed solely to Microsoft’s print. S16
FY27 EPS / Multiple
18×
21×
24×
Interpretation
$18.00 bear
$324
$378
$432
AI growth decelerates; margin and capex pressure dominate.
$19.38 consensus
$349
$407
$465
Pre-print consensus snapshot; likely to move after guidance. S8
$20.50 bull
$369
$431
$492
Azure stays above 40% and operating leverage offsets depreciation.
Scenario prices are mechanical EPS × P/E outputs, not price targets. They exclude net cash adjustments because equity EPS multiples already capitalize the whole enterprise. Forward inputs were captured before the print and may be stale until consensus refreshes. S8
09 · Read-throughs and event map
The ecosystem is broadening while the economic exposure concentrates
What changed in the last quarter
April 27, 2026
OpenAI agreement reset.
Microsoft remains the primary cloud partner, but OpenAI gained more multi-cloud flexibility; Microsoft’s license remains non-exclusive through 2032 and the revenue-sharing mechanics changed. S10
June 16, 2026
Copilot Cowork became generally available.
A concrete usage and ARPU checkpoint for enterprise agent monetization. S11
June 22–29, 2026
Capacity and model breadth expanded.
Microsoft announced a 2GW Pecos datacenter project and made Claude generally available in Foundry, broadening Azure beyond a single model vendor. S12S13
July 1–6, 2026
Pricing and cost base reset.
M365 packaging/pricing changes took effect and Microsoft announced 4,800 role reductions alongside a voluntary-retirement program. S14
Twelve-month anchor — November 18, 2025: Anthropic committed to purchase $30B of Azure capacity and Microsoft agreed to invest up to $5B. That transaction now matters twice: it supports backlog and multi-model strategy, but also produced part of Q4’s below-the-line gain. S15
Cross-company read-throughs
NVDA / AMD: rising short-lived accelerator purchases support near-term compute demand, although Microsoft’s own Maia/Cobalt road map creates long-run substitution risk.
OpenAI / Anthropic: Microsoft is diversifying model supply and monetizing investment marks, but the 10-K makes commercial and credit concentration much more explicit.
Datacenter power, cooling and construction: the $329B uncommenced lease pipeline and new 2GW project support a multi-year infrastructure cycle, with timing and return-on-capital risk.
Enterprise software: 30M Copilot paid seats validate distribution; slowing M365 and Dynamics growth means investors still need proof of usage intensity and incremental economics.
Next dated event: the FY2027 Q1 earnings date was not confirmed at publication. The monitoring window is late October 2026 based on Microsoft’s historical cadence; this is an analyst inference, not company guidance.
10 · Catalysts, falsifiers and action discipline
A better business print does not remove the need for price discipline
Upside gates
Azure stays above 40% while capacity constraints ease.
Copilot usage and ARPU rise faster than paid-seat count.
RPO converts without customer concentration or renegotiation surprises.
FY27 EPS revisions move above $20 without higher capital intensity.
Falsifiers
Azure growth falls below the high-30s before supply catches demand.
Microsoft Cloud gross margin drops below management’s path without offsetting revenue acceleration.
OpenAI receivables, commitments or revenue concentration deteriorate.
Simple FCF remains below $70B while leases and capex keep expanding.
M365/Dynamics deceleration shows Copilot is cannibalizing rather than expanding wallet share.
Positioning framework
Existing holders: Hold the core position. The Azure and backlog evidence improved, but use cloud gross margin, lease-adjusted capital intensity and RPO conversion as the conditions for adding.
New money: Do not chase a 7.6% post-call move. A small starter is defensible near 21× FY2027 EPS; scale only on a pullback or when free-cash-flow and margin evidence confirms the revenue acceleration.
Do not do: annualize the $0.27 discrete EPS benefit or value the $678B backlog as if it were short-duration, unconcentrated recurring revenue.
Bottom line: Microsoft delivered the operating print the bull case needed and then guided Azure to another acceleration. That strengthens the thesis materially. The remaining constraint is not demand; it is whether a lease-heavy infrastructure program can turn that demand into attractive incremental cash returns.
11 · Evidence ledger and method
Primary filings for reported facts; public consensus only where needed
Regular-session one-year history and timestamped post-market observations. Post-market data can be volatile and is not an official close.
Market data
S8
Point-in-time consensus snapshot
Pre-print FY2027/FY2028 revenue and EPS aggregation; captured July 29, 2026. Used only for clearly labeled forward valuation inputs and subject to revision.
FOMC decision and press conference on July 29, relevant to same-day market attribution.
Primary · Official calendar
Method and limitations: Reported financials and operating KPIs come from SEC/IR primary sources. Simple FCF equals operating cash flow minus cash PP&E additions and is not lease-adjusted. Segment margins, the ~$4.47 EPS proxy, valuation ratios, scenarios and investment conclusions are analyst calculations. Public consensus data are point-in-time and not fully standardized across providers. No management quote is reproduced verbatim beyond short labels.