Azure Holds 39% on a Bigger Base, Capex Steps DOWN $6.6B Q/Q, RPO Sprints to $627B — The Q2 Bear Case Just Got Materially Weaker
Initial Read: Clean operational beat with two thesis-critical surprises — Azure +39% cc holds the line one point above the 37–38% guide, and capex came in at $30.9B vs. Street $35.3B (a $4.4B undershoot and a $6.6B sequential step-DOWN from Q2's record $37.5B). Commercial RPO accelerated to $627B (+99% Y/Y). Upgrading from Hold back to Outperform; expect the AH tape to run +5–8%.
Key Takeaways
- Triple beat across the headlines: Revenue $82.9B vs. $81.5B Street (+1.8%, +18% Y/Y, +15% cc). EPS $4.27 vs. $4.04 Street (+5.7%, +23% Y/Y). Operating income $38.4B (+20% Y/Y, +16% cc) — the first quarter in the FY26 sequence where revenue, OI, and EPS all printed comfortably above consensus with no offsetting concern.
- Azure & other cloud services +39% constant currency — at the high end of the company's prior 39–40% framing, one point above the 37–38% guide given at Q2, and most importantly flat sequentially after three consecutive prints at 39% cc (FY25 Q4, FY26 Q1, FY26 Q2). The "first sequential deceleration" call that anchored the Q2 downgrade did not materialize. Azure compounding at 39% on a substantially bigger base is structurally bullish.
- Capex $30.9B — the single most important number in the release. Down $6.6B sequentially from Q2's $37.5B (the largest single-quarter capex print in corporate history) and $4.4B below the $35.3B Street consensus. This directly refutes the Q2 bear thesis that capex-without-revenue would compound into FCF compression. Operating cash flow of $46.7B vastly exceeded the spend, restoring meaningful FCF after Q2's collapse to $5.9B.
- Commercial RPO $627B (+99% Y/Y) vs. $625B at Q2 — a $2B sequential add on a $625B base after the $250B OpenAI commit landed in Q1. The growth metric remains the cleanest hyperscaler bookings signal in the market and the leading indicator that justifies the capex curve.
- AI annual revenue run rate $37B, +123% Y/Y — this is the first explicitly disclosed AI revenue dollar figure in the FY26 sequence. At $37B against ~$95B of FY26 capex (run-rate), the AI business is now a non-trivial monetization engine, not just a capex story. Microsoft Cloud at $54.5B (+29% Y/Y, +25% cc) crossed the line on a single-quarter basis.
- Rating: Upgrading from Hold to Outperform. Our Q2 Hold was anchored to three concerns: (1) Azure step-down, (2) capex inflation outrunning FCF, (3) memory-pricing tail risk. Concerns (1) and (2) were directly refuted by this print. Concern (3) remains unaddressed pre-call. Upgrade is supported by the operational beat plus the disciplined capex pivot, with the recap re-evaluating after the call's capex/Copilot/memory commentary.
Results vs. Consensus
| Metric | Actual (FY26 Q3) | Consensus | Beat/Miss | Magnitude |
|---|---|---|---|---|
| Revenue | $82.9B | $81.5B | Beat | +1.8% (+18% Y/Y, +15% cc) |
| Diluted EPS (GAAP) | $4.27 | $4.04 | Beat | +$0.23 (+5.7%, +23% Y/Y) |
| Operating Income | $38.4B | ~$36.5B (implied) | Beat | +5.2% (+20% Y/Y, +16% cc) |
| Net Income | $31.8B | n/a | Beat | +23% Y/Y |
| Microsoft Cloud Revenue | $54.5B | ~$53.6B | Beat | +29% Y/Y (+25% cc) |
| Azure & other cloud (cc) | +39% | +37–38% (mgmt guide) | Beat | +1–2 pts above guide |
| Capital Expenditures | $30.9B | $35.3B | Beat (lower) | -$4.4B undershoot |
| Operating Cash Flow | $46.7B | n/a | Strong | FCF rebuild after Q2 collapse |
Quality of Beat
- Revenue: Beat is broad-based and operationally clean. The +18% headline (+15% cc) is the strongest growth rate in the FY26 sequence (FY26 Q1 was +18%, FY26 Q2 was +15%). FX is a 3-pt tailwind, not a gimmick — underlying constant-currency growth is steady. Growth is concentrated in Intelligent Cloud (+30% Y/Y) which is also the highest-margin and highest-quality segment. P&BP at +17% is the strongest print of the FY26 sequence as well.
- Margins: Operating income +20% Y/Y vs. revenue +18% — operating margin expanded ~70 bps despite the AI infrastructure ramp. This is the single most important operating-leverage data point in the release: the bear case has been that AI capex would compress unit economics, but Q3 shows operating leverage even as capex remains elevated. Constant-currency operating income growth at +16% slightly trails revenue at +15% cc — tiny, but worth flagging.
- EPS: $4.27 vs. $4.04 is an operational beat, not a tax or share-count beat. Net income +23% Y/Y vs. EPS +23% Y/Y means share-count tailwind was minimal — buybacks contributed but were not the swing factor. The 18% cc EPS growth confirms underlying earnings power.
- Capex undershoot is the surprise: $30.9B is $4.4B below Street and $6.6B below the prior quarter. After three consecutive quarters of capex prints exceeding expectations (Q1 $34.9B vs. ~$30B expected, Q2 $37.5B vs. ~$32B expected), Q3 broke the pattern in the opposite direction. Two interpretations: (a) supply discipline — Microsoft has chosen to slow incremental builds as pre-existing capacity comes online, or (b) timing — some Q3 deliveries slipped to Q4. The recap will need management's framing on the call to distinguish, but on the press release alone this is unambiguously bullish for FY26 FCF.
Segment Performance
| Segment | FY26 Q3 Revenue | Y/Y (rep / cc) | Op Income | Our Assessment |
|---|---|---|---|---|
| Productivity & Business Processes | $35.0B | +17% / +13% cc | $21.0B | Strongest P&BP print of FY26. Copilot ARPU monetization is showing up in M365 Commercial cloud +19% / +15% cc — a sequential acceleration vs. FY26 Q2. |
| Intelligent Cloud | $34.7B | +30% / +28% cc | $13.8B | The headline driver. Azure +39% cc holds the line; total IC growth at +28% cc indicates server/on-prem is still alive and contributing. Op income $13.8B is 39.8% segment margin — healthy given AI ramp. |
| More Personal Computing | $13.2B | -1% / -3% cc | $3.7B | The drag, but contained. Windows OEM -2% / -3% cc and Xbox -5% / -7% cc reflect the well-known consumer-PC and gaming softness. Search ads (ex-TAC) +12% / +9% cc is a quietly bright spot. |
Productivity & Business Processes — Copilot Monetization Showing Up
P&BP at $35.0B (+17% Y/Y, +13% cc) is the strongest segment print since the FY25 Q4 cycle peak. Inside the segment, the standout is Microsoft 365 Commercial cloud at +19% (+15% cc), which is a sequential reacceleration from FY26 Q2's deceleration concern. M365 Consumer cloud at +33% (+29% cc) is benefiting from Copilot Pro / Copilot for Consumer pricing flow-through. Dynamics 365 at +22% (+17% cc) continues to be the under-appreciated SaaS franchise — at this growth rate, Dynamics is now a $20B+ revenue ramp on a multi-year horizon.
LinkedIn at +12% (+9% cc) is a respectable print given the macro backdrop on hiring; this is the strongest LinkedIn quarter of the FY26 sequence and indicates the enterprise hiring corner has turned.
Segment operating income of $21.0B implies a 60.0% segment margin — this is the most profitable segment by far and continues to fund the IC capex ramp.
Intelligent Cloud — Azure Holds 39% on the Biggest Base in the Stack
Intelligent Cloud at $34.7B (+30% Y/Y, +28% cc) is the headline. Azure & other cloud services at +39% constant currency is the data point that defines this print. To recap the FY26 trajectory: FY25 Q3 was +35% cc, FY25 Q4 was +39% cc, FY26 Q1 was +39% cc, FY26 Q2 was +39% cc — and FY26 Q3 prints +39% cc again. Azure has now compounded at 39% cc for four consecutive quarters on a base that has roughly doubled over the period.
Critically, management's own Q3 guide (issued at the Q2 print) was for 37–38% cc, framed as "the first sequential deceleration." That deceleration did not happen. The 1–2 point beat against the guide is the difference between a Hold and an Outperform on the press release alone.
The +28% cc total IC growth rate (vs. +39% cc Azure-only) implies the on-prem server/legacy bucket is decelerating but not collapsing — likely growing in the mid-single digits constant currency. This is consistent with the secular framework: server licenses are migrating to Azure, but the migration is steady, not cliff-edge.
Segment operating income of $13.8B against $34.7B revenue is a 39.8% segment margin. Given the AI infrastructure ramp running through this segment, this margin level is genuinely impressive — it means the AI workloads are running at unit economics that are not dilutive to legacy Azure margins, which has been one of the open questions across the FY26 sequence.
More Personal Computing — Contained Drag
MPC at $13.2B (-1% Y/Y, -3% cc) is a soft print but not a thesis-relevant one. Windows OEM and Devices -2% (-3% cc) reflects the broader consumer PC softness; Xbox content & services -5% (-7% cc) is the well-telegraphed gaming consumption decline. The bright spot is Search and news advertising (ex-TAC) at +12% (+9% cc) — Bing AI and Copilot-driven search continue to take incremental ad share. Segment OI of $3.7B implies 28.0% segment margin, reasonable given the mix.
For thesis purposes, MPC is the smallest of the three segments and not the marginal driver of the multiple. The drag here is fully contained and was already in the model.
Operating Margin Walk
| Segment | Revenue | Op Income | Segment Margin | Y/Y Rev |
|---|---|---|---|---|
| Productivity & Business Processes | $35.0B | $21.0B | 60.0% | +17% |
| Intelligent Cloud | $34.7B | $13.8B | 39.8% | +30% |
| More Personal Computing | $13.2B | $3.7B | 28.0% | -1% |
| Total | $82.9B | $38.4B | 46.4% | +18% |
Consolidated operating margin of 46.4% (vs. 45.7% in FY26 Q2 and ~45.5% Y/Y prior) is the highest single-quarter operating margin in the FY26 sequence. The expansion is most pronounced in P&BP (Copilot ARPU at high incremental margin) and stable in IC despite the AI ramp.
Capex / AI Infrastructure Spend — The Pivot Point
| Quarter | Capex (PP&E + finance leases) | Operating Cash Flow | Implied FCF | Capex / OCF |
|---|---|---|---|---|
| FY25 Q3 | ~$21B | ~$37B | ~$16B | ~57% |
| FY25 Q4 | ~$24B | ~$42B | ~$18B | ~57% |
| FY26 Q1 | ~$34.9B | ~$45B | ~$10B | ~78% |
| FY26 Q2 | $37.5B | $43.4B | $5.9B | 86% |
| FY26 Q3 | $30.9B | $46.7B | ~$15.8B | 66% |
FCF rebuilds materially. Implied free cash flow of ~$15.8B in Q3 is roughly 2.7x the Q2 trough of $5.9B and is back into the historical range. Capex/OCF ratio of 66% is meaningfully better than the 78–86% prints of FY26 Q1–Q2 and back toward the FY25 baseline. This single quarter does not undo the multi-year capex cycle — but it is the first concrete signal that capex is being pulsed against demand visibility rather than ramped open-loop.
FY26 capex run-rate math: Cumulative FY26 capex = $34.9B + $37.5B + $30.9B = $103.3B through nine months. Annualized to ~$138B at a Q3-rate run-rate, or ~$140–145B if Q4 returns toward the Q2 level. The $190B FY26 figure circulating in pre-print previews now looks high relative to actual nine-month cadence — unless Q4 prints a step-up. The bear case requires Q4 to be $50B+ to validate the $190B number; the recent trajectory points lower.
The bear-case test on capex is now on the call. Pre-print, the bear thesis required (a) capex continuing to ramp, (b) FCF compression continuing, and (c) Azure decelerating — all three to compress the multiple further. None of the three printed in Q3. The recap will need to test whether (i) capex moderation is structural or timing-driven, and (ii) whether management commits to disciplined capex framing for FY27.
Capital Returns
- Total returned to shareholders: $10.2B in dividends and buybacks combined.
- Pace is steady relative to recent quarters — Microsoft's capital return cadence has been deliberately understated through the AI capex cycle, with management prioritizing reinvestment over accelerated buyback. With FCF rebuilding, capital return capacity should expand into FY27 even with continued capex commitments.
- The relative-value framing: at $10.2B/quarter of capital return against ~$130B of operating cash flow on an annualized FY26 basis, MSFT is returning roughly 30% of OCF to shareholders versus 75–80% to capex. This is the inverse of the historical ratio (which was ~50/50) and is the financial signature of the AI build-out cycle. The Q3 capex moderation begins to rebalance this.
Bookings & Commercial RPO — The Demand Validator
- Commercial Remaining Performance Obligation: $627B, +99% Y/Y. Up from $625B at Q2 (+$2B sequential) and $392B at Q1 (where the $250B incremental OpenAI commit landed). The base is so large now that incremental sequential growth is harder to read — but the +99% Y/Y growth rate confirms the demand backdrop has not weakened.
- AI annual revenue run rate: $37B, +123% Y/Y. First explicit AI revenue dollar disclosure in the FY26 sequence. At $37B run rate, AI is now ~11% of consolidated annualized revenue (~$330B). Run-rate growth at +123% Y/Y is decelerating from prior-year hyper-growth but remains the highest growth rate of any disclosed business line at scale globally.
- Microsoft Cloud revenue: $54.5B, +29% Y/Y, +25% cc. Single-quarter; first time crossing $54B. Annualized at $218B+ Microsoft Cloud — a business unit larger than 90% of S&P 500 companies on standalone revenue.
Reading the RPO Cadence
The RPO step-up pattern across FY26 looks like: ~$368B (FY25 Q4) → $392B (FY26 Q1, with $250B OpenAI booking) → $625B (FY26 Q2, the major OpenAI booking flowed through) → $627B (FY26 Q3, +$2B sequential, +99% Y/Y). The $2B sequential add in Q3 looks small, but on a $625B base it implies Microsoft is still booking incremental commitments at a ~$8B/quarter run-rate after digesting the OpenAI mega-deal. Annualized that's another ~$32B/year of incremental contracted revenue building on an already-massive backlog. RPO duration almost certainly extended further given the OpenAI deal structure (through 2030–2032).
Notable Items in the Release
Capex Sequential Step-Down
Capex of $30.9B is $6.6B below the Q2 record print and $4.4B below Street consensus of $35.3B. This is the first sequential decline in capex in the FY26 sequence and the first material undershoot of expectations. The press release does not characterize the moderation explicitly; management framing on the call will be the key disambiguator between "supply-side discipline" (bullish) and "deliveries slipped to Q4" (neutral).
Assessment: On the press release alone, this is the single most bullish data point relative to the Q2 thesis. If management confirms structural pacing on the call — even directionally — the capex-without-revenue bear case is materially weakened.
AI Run-Rate Disclosure ($37B, +123% Y/Y)
Microsoft has been reluctant to disclose AI revenue figures historically. The explicit $37B run-rate disclosure in the press release (in Nadella's quote) suggests management's confidence has reached a level where they're willing to let the Street size the AI business directly. At $37B run-rate, AI is roughly 11% of MSFT's consolidated revenue, and at +123% Y/Y growth, it is the single fastest-growing $30B+ revenue stream in the public markets.
Assessment: Strategic disclosure shift. The $37B is the dollar number that will replace the abstract "Azure compounding" framing in Street models post-print. Recap to dig into composition — Copilot, Azure AI services, OpenAI revenue passthrough.
OpenAI Partnership Restructuring (April 27, 2026)
Two days before this print, Microsoft announced a restructuring of its OpenAI partnership that eliminated outbound revenue-share payments while preserving Azure priority through 2032. Not in the press release directly, but materially relevant to interpreting the Q3 numbers and forward margins.
Assessment: Margin tailwind that begins to flow through Q4 onward. The operating margin expansion in Q3 already reflects partial benefit. Recap to quantify with management commentary.
Nadella's "Agentic Computing" Framing
The CEO quote uses "agentic computing era" — a framing shift from the prior "AI infrastructure" or "Copilot transformation" framing. This is consistent with broader industry rhetoric (autonomous AI agents as the next platform shift) and signals the next product narrative MSFT will sell into.
Assessment: Marketing/positioning, not a number. But worth tracking — agentic compute monetization will be the FY27 story.
Guidance & Outlook
Microsoft does not provide forward-looking quantitative guidance in the press release (consistent with their convention — guidance is given on the earnings call). The press release language reads:
"Microsoft will provide forward-looking guidance in connection with this quarterly earnings announcement on its earnings conference call and webcast."
The implicit setup, however, is constructive. The Q3 operational beat plus capex undershoot give management room to guide Q4 confidently on Azure (likely 36–38% cc range as the comp gets harder), reaffirm or modestly raise FY26 capex (likely $140–150B vs. earlier-cycle $190B speculation), and signal continued operating-margin expansion. The recap will dissect the actual Q4 and FY27 framing.
What to watch on the call:
- Q4 FY26 Azure cc growth guide — bullish if 37%+, bearish if 35% or below
- FY26 total capex framing — bullish if <$150B, bearish if $170B+
- FY27 capex directional framing — Hood typically gives early FY framing on the Q3 call
- Memory pricing pass-through to capex / Azure GMs
- Copilot seat count update (last disclosed: 15M paid seats at FY26 Q2, +160% Y/Y on seat-adds)
Questions for the Call
- Capex moderation — structural or timing? The $6.6B sequential step-down is the headline of the print. Bullish: management confirms supply-side discipline, frames Q3 as the new run-rate, and signals FY26 total capex below $150B. Bearish: management attributes Q3 to timing slippage, signals Q4 catch-up to $40B+, and reaffirms higher-than-FY25 framing.
- FY27 capex framing. Hood typically gives early FY framing on the Q3 call. Bullish: FY27 capex moderates in absolute dollars or grows at single-digit percentages — implying the AI infra build-out is reaching a sustainable plateau. Bearish: FY27 capex grows another 20%+ — implying multi-year FCF compression.
- Azure trajectory through Q4 and into FY27. Q3 held 39% cc against a 37–38% guide; the comp gets harder from here. Bullish: Q4 guide at 37%+ cc with a confident mid-30s framing for FY27. Bearish: Q4 guide at 35% cc or below, signaling the four-quarter 39% sequence was a peak.
- AI revenue composition. The $37B run-rate is the most-watched new number. Bullish: majority is Copilot ARPU and Azure AI services with healthy unit economics; minimal OpenAI passthrough revenue. Bearish: dominated by lower-margin OpenAI compute resale at near-cost pricing.
- Memory pricing impact. Hood flagged this as a multi-vector risk on the Q2 call. Bullish: contracted memory pricing has stabilized, and the impact on capex / Azure margins is contained. Bearish: memory contracts re-priced higher, with multi-quarter margin headwind.
- Copilot paid seat count and ARPU. Last disclosed 15M paid seats at FY26 Q2. Bullish: 20M+ paid seats with rising ARPU mix shift; M365 Commercial Cloud +15% cc growth is durable. Bearish: seat-add growth slowing as the early-adopter wave matures, with rising churn risk.
- OpenAI partnership restructuring P&L impact. April 27 announcement eliminates outbound rev-share. Bullish: ~$X billion annual margin tailwind, beginning Q4. Bearish: structural changes also reduce IP-share economics that may be offsetting.
- Search advertising +12% (+9% cc). Quietly the strongest growth in MPC. Bullish: Bing/Copilot AI search is taking incremental ad share at a structural pace. Bearish: the comp benefited from a soft prior-year quarter.
Market Reaction
Earnings released April 29, 2026, 4:00–5:00 PM ET. After-hours trading data not yet available at time of writing. Based on the print, we expect MSFT to trade up 5–8% in the after-hours session, driven by:
- Capex undershoot of $4.4B vs. Street (the most thesis-relevant surprise)
- Azure +39% cc beating the 37–38% guide (resolves the Q2 deceleration concern)
- EPS beat of +5.7%
- Operating margin expansion to 46.4% despite AI ramp
The risk to a larger gap-up is the call commentary on capex framing — if Hood signals Q4 catch-up to $40B+ or reaffirms FY26 capex at the high end, the AH gain could fade. If Hood confirms structural moderation, +8% is achievable on the close tomorrow. Either way, the print is unambiguously above the bar set by the Q2 reaction.
Model Implications
| Item | Prior View (Q2 Recap) | Post-Q3 FY26 | Reason |
|---|---|---|---|
| Azure cc growth FY26 exit | 35–37% (decelerating) | 37–39% (holding) | Q3 printed +39% cc against 37–38% guide; sequential deceleration thesis didn't print. |
| FY26 Capex | ~$170–190B | $140–150B | Nine-month cumulative $103.3B; Q3 $30.9B sets a lower run-rate. |
| FY26 FCF | ~$80–90B | ~$95–105B | Q3 FCF rebuild to ~$15.8B; lower capex flows directly to FCF. |
| FY26 EPS | ~$15.50 | ~$16.20 | Q3 $4.27 ahead of plan; Q4 setup constructive. |
| Operating margin FY26 | ~45% | ~46% | Q3 prints 46.4%; Copilot ARPU and OpenAI restructuring tailwind both positive. |
| AI revenue run-rate trajectory | Implied ~$30B | $37B disclosed | Explicit Q3 disclosure exceeds prior implied number. |
| Commercial RPO | $625B (Q2) | $627B (+99% Y/Y) | Steady incremental bookings on top of OpenAI base; demand backdrop intact. |
Valuation: An FCF re-acceleration toward $100B+ in FY26, combined with Azure compounding at 39% on a doubling base and operating margin expanding back toward 46%+, restores the high-quality compounder framing that anchored the original Outperform initiation. At a forward FCF yield of ~2.7% on a $100B FCF base and $3.7T market cap, the multiple has room to re-rate higher if the call confirms FY27 capex discipline. Our prior Hold framework was anchored to FY26 FCF in the $80–90B range; the Q3 print pulls the FY26 trajectory back into the >$95B range and removes the principal financial concern.
Thesis Scorecard
| Thesis Point | Status | Notes |
|---|---|---|
| Bull #1: Azure compounding at 25%+ on a broadening workload mix | Confirmed | Q3 prints +39% cc on a doubled base — fourth consecutive quarter at 39% cc. Materially exceeds the bull thesis bar. |
| Bull #2: Copilot ARPU drives M365 Commercial Cloud reacceleration | Confirmed | M365 Commercial cloud +19% (+15% cc); P&BP segment margin 60% — Copilot pricing flowing through. |
| Bull #3: AI-side unit economics hold through capex ramp | Confirmed | Intelligent Cloud segment margin 39.8% despite AI infrastructure ramp; consolidated OM 46.4% expanded Y/Y. |
| Bull #4: OpenAI relationship is structurally favorable | Confirmed | April 27 restructuring eliminated outbound rev-share while preserving Azure priority through 2032. Net margin tailwind. |
| Bear #1 (Q2): FCF compression to ~$80-90B FY26, structural through asset depreciation | Refuted (this quarter) | Q3 FCF rebuild to ~$15.8B vs. Q2 $5.9B trough. Capex undershoot is the cleanest single-quarter refutation. Recap will retest on full-year math. |
| Bear #2 (Q2): Azure as a managed metric / portfolio-LTV-optimization output | Neutral | The metric printed cleanly above guide this quarter, which is supportive — but the underlying allocation question Hood raised at Q2 is not directly addressable from the press release. |
| Bear #3 (Q2): Memory pricing as multi-vector unquantified tail risk | Neutral / Open | Not directly addressed in the press release. The 46.4% consolidated operating margin suggests no material flow-through yet, but the call commentary is the test. |
| Bear #4 (new): Capex run-rate compounds into multi-year FCF compression | Refuted (preliminary) | Q3 capex $30.9B is the first sequential step-down. If the call confirms structural pacing, this bear point is materially weakened. |
Overall: Thesis materially strengthened. The two financial concerns that drove the Q2 downgrade to Hold — Azure decelerating sequentially and capex outpacing FCF — both reversed in Q3. The third concern (memory pricing) is unaddressed pre-call but the operating margin expansion suggests it is not yet flowing through. The fourth bull pillar (OpenAI relationship) was independently strengthened by the April 27 restructuring.
Preliminary Action: Upgrade from Hold to Outperform. The numbers in this press release directly refute two of the three financial concerns from the Q2 downgrade. Unless the call introduces a new material negative (memory pricing pass-through quantified, or Q4 capex framing reverses the moderation trend), the upgrade should hold through the recap. Position sizing should reflect the rapid round-trip: Outperform → Hold (Q2) → Outperform (Q3) within 90 days indicates how data-dependent the financial profile has become quarter-to-quarter, and risk controls should respect that volatility.