Three hyperscalers reported heavy AI capex weeks within days of each other in July 2026, and the market drew three different conclusions. Alphabet raised capex and fell 7%. Meta's free cash flow cratered and it fell 9%. Microsoft reported the same evening as Meta — and rose almost 9%, adding roughly $260 billion of market value in a single session. Same theme, opposite outcome. The difference is worth understanding in detail, because it's not simply "Microsoft had a better quarter" — part of the reaction rode on a one-time gain and an accounting change, not pure fundamentals.
The Headline Numbers
| Metric | Q4 FY2026 | Estimate | Result |
|---|---|---|---|
| Revenue | $90.0B | ~$87.6B | +18% YoY, beat |
| Adjusted EPS | $4.74 | $4.24 | Beat by $0.50 |
| GAAP diluted EPS | $4.81 | — | +32% YoY |
| Intelligent Cloud revenue | $39.3B | ~$38.2B | +32% YoY |
| Azure growth (constant currency) | 43% | 39–40% guided | Blew past guidance |
| Commercial RPO (backlog) | $678B | — | +84% YoY, +$51B QoQ |
| M365 Copilot paid seats | 30M+ | — | Up from 20M+ in April |
Every headline number cleared consensus, several by a wide margin. Azure didn't just beat its own guided range — it blew past it, pushing full-year Azure revenue above $100 billion for the first time in company history, a milestone CEO Satya Nadella specifically flagged on the call. Operating margin actually ticked up slightly to 45%, even as gross margin dipped to 67% on heavier AI infrastructure costs — evidence Microsoft is managing the cost side of the AI buildout better than the "AI spending crushes margins" narrative that dogged this earnings season elsewhere.
Why the Market Rewarded This Quarter Specifically
Four things lined up at once. The beat was broad, not concentrated in one flattering line. Guidance for the new fiscal year came in well ahead of models — Microsoft guided September-quarter revenue to $89.85–90.95 billion (above the ~$89.7B analysts expected) and Azure growth to 45% constant currency, sharply above the roughly 41% analysts had modeled. Capex guidance came in below what the Street had priced in, and investors read that as spending discipline. And CFO Amy Hood used the call to directly address the market's biggest lingering worry about Microsoft's AI story — customer concentration — disclosing that nearly 90% of Microsoft Cloud revenue now comes from customers outside the frontier AI labs, a specific, quotable number that undercuts the idea Microsoft's growth rides entirely on OpenAI.
Our take: that Azure backlog number — $678 billion, up $51 billion in a single quarter, +84% YoY — is the single most important figure in this release, more important than the EPS beat or the capex guide. It's the closest thing to hard proof, across this entire earnings season, that AI infrastructure spending is converting into contracted, forward-visible revenue rather than speculative build-out. Alphabet's bull case rests on an argument that its Cloud margins are already expanding; Microsoft's rests on a backlog number nobody has to interpret or estimate. That's a meaningfully stronger form of evidence, and it's very likely the single biggest reason the market treated Microsoft's capex story so differently than Alphabet's and Meta's the same week.
Two Reasons to Read the Headline Beat Carefully
This wouldn't be a clean earnings story without at least one number that overstates itself, and Microsoft has two worth flagging.
A $3.2 billion gain on Microsoft's stake in Anthropic added $0.33 to this quarter's EPS — out of a total beat of roughly $0.50 over consensus. That means close to two-thirds of the headline earnings beat came from marking up a private investment, not from selling more Azure capacity or Office seats. Microsoft doesn't revalue this stake every quarter; it updates the number only when a fresh financing round gives it a new price to point to, so this isn't a repeatable quarterly tailwind.
The "lower capex" story is partly an accounting change, not a spending cut. Calendar-year 2026 capex guidance came down to roughly $175 billion, from about $190 billion guided back in April — and the market cheered it as discipline. But Reuters reported Microsoft is changing how it accounts for long-term data-center leases, spreading them over 25 years instead of 15. Stretching a depreciation schedule lowers how much spending shows up in any single year's capex line without necessarily reducing what Microsoft has actually committed to spend in total — the same building costs the same amount; it just gets expensed more slowly. An investor reading "capex came down" as proof AI infrastructure spending has peaked would be reading this report more generously than the numbers underneath it fully support.
Neither of these caveats makes the quarter weak — Azure's 43% growth and the $678 billion backlog are genuinely demand-driven, cash-generating numbers that no accounting choice can manufacture. But the size of the stock's reaction, an 8.88% single-day pop, was helped along by both a non-recurring gain and a friendlier way of counting the same underlying spend, not purely by stronger core fundamentals. The next couple of quarters — specifically whether Azure keeps growing near 40%+ without one-time items propping up EPS — will tell you which story was closer to the truth.
The One Soft Spot
More Personal Computing (Windows, Xbox, Surface) revenue fell 4.4% YoY to $12.9 billion, though it still beat the roughly $12.2 billion estimate. Pricier PCs — driven by an industry-wide memory-chip shortage — and a tough Xbox comparison weighed on the category; Omdia estimated global PC shipments fell 3.6% in Q2 2026, including a 4.2% decline in notebooks. Windows OEM revenue is guided down in the high teens for FY2027. Worth remembering: not every part of Microsoft is riding the AI wave at the same speed, and the memory-chip cost pressure hitting PCs here is the same industry dynamic squeezing margins across the hardware supply chain broadly this year.
Sector Read-Through
Lay the three hyperscaler reactions from late July side by side and the pattern is instructive. Alphabet raised 2026 capex to $195–205 billion and beat on cloud growth — stock fell about 7%. Meta's free cash flow collapsed to $784 million from $8.5 billion — Meta's own reported basis, which nets off finance-lease principal; our series, on operating cash flow minus capex, puts the quarter at $1,746M — with no forward capex number for 2027, and the stock fell about 9%. Microsoft lowered its headline capex number (even if partly via accounting) and backed it with the clearest demand evidence of the three, an $678 billion, 84%-growing backlog — stock rose 8.88%. The market isn't punishing AI capex uniformly; it's punishing capex that arrives without a matching, verifiable demand signal, and rewarding capex that arrives with one. Microsoft is currently the cleanest example of the latter.
Analyst Reaction
Heading into the print, roughly 50 analysts covered the stock with a Strong Buy consensus and an average target near $550–557. Post-earnings, Bernstein reiterated Buy at $646, Wedbush Outperform at $625, Morgan Stanley Overweight at $600, and Truist Buy at $575 — all well above the pre-earnings average, reflecting upward revisions on the beat and raise. One outlier: Scotiabank actually cut its target to $470 from $550 the same morning, before the print's full reception was clear, a reminder that not every analyst call lines up neatly with the market's eventual verdict.
Even after the rally, valuation isn't stretched relative to its own recent history: trailing P/E moved from roughly 21.75x pre-earnings to about 23.69x after, in the same 20–24x range the stock has traded in through this earnings cycle rather than a new, expensive multiple.
What to Watch
- Whether Azure sustains 40%+ growth without one-time items inflating EPS in the next one or two quarters — the real test of whether this print marked a genuine inflection or an unusually strong outlier.
- How the $678 billion RPO actually converts to revenue over time, and at what margin, as the AI infrastructure it's backed by comes online.
- Whether the 25-year lease depreciation change gets scrutinized more closely by analysts modeling true cash capex versus reported capex going forward.
- More Personal Computing and Windows OEM, both guided down for FY2027 — a real, if smaller, drag worth tracking separately from the AI/cloud story.
The Bottom Line
Microsoft delivered a genuinely strong quarter — Azure's acceleration, the ballooning backlog, and Copilot's seat growth are demand-driven signals that don't require any accounting interpretation to trust. But the size of Wednesday night's reaction rode partly on a non-operating investment gain and a longer depreciation schedule for the same underlying spend, not purely on stronger fundamentals. If Azure keeps growing near 40% or higher on its own, this report will look like the moment the AI capex story genuinely turned a corner — and turned it specifically because Microsoft, unlike its hyperscaler peers this earnings season, could point to a concrete number proving the demand is already contracted, not just hoped for.
Microsoft Corporation (NASDAQ: MSFT) reported Q4 FY2026 revenue of $90.0B (+18% YoY) and adjusted EPS of $4.74 (beat by $0.50, roughly two-thirds from a one-time Anthropic stake gain). Shares rose approximately 8.9% in after-hours trading, adding roughly $260B in market value.