KEY POINTS
- Microsoft beat Q3 2026 estimates across the board — revenue of $82.9B, adjusted EPS of $4.27, Azure growth of 40% in constant currency.
- The AI business crossed $37B annualized run rate, up 123% year over year, with Copilot paid seats surpassing 20 million — up 250% YoY.
- The company plans to spend $190 billion in capital expenditures across calendar 2026 — and is still supply-constrained. That's the number every investor needs to understand.
Microsoft reported Q3 fiscal 2026 results on April 29 after the close. The stock moved higher in after-hours trading. Most of the coverage you'll read summarizes the beat and moves on. Which brings me to the part that actually matters — what this report reveals about where the company is placing its bets, what's working, and what's getting quietly glossed over in the prepared remarks.
What they reported
Microsoft — the company that sells cloud infrastructure, office software, and increasingly AI-powered agents to basically every large enterprise on earth — posted Q3 fiscal 2026 revenue of $82.9 billion, up 18% year over year in constant currency, with Azure growing 40% and its AI business crossing $37 billion in annual run rate while Copilot paid seats surpassed 20 million for the first time.
Beat or miss
Beat. Revenue came in at $82.9 billion against a consensus expectation of roughly $81.4 billion. Adjusted EPS was $4.27 — ahead of the ~$3.45 the Street had modeled, though that figure is adjusted for the accounting treatment around the OpenAI investment, so the raw GAAP number looks different depending on which line you're reading. Operating margin held at 46%, roughly flat year over year despite an enormous ramp in AI spending.
What the CEO actually said about the future
Satya Nadella said a lot of words. What he actually meant is roughly this: Microsoft is betting its entire next decade on the idea that every knowledge worker, every developer, and every security team will eventually run AI agents alongside them — and that Microsoft will charge for both the seat and every token those agents consume. That's the model transition he kept circling without ever saying plainly on the call.
It sounds great. The catch is buried in CFO Amy Hood's remarks about "supply constraints persisting at least through 2026" — which means they're essentially telling investors that demand is genuinely outrunning their ability to build data centers fast enough to capture it. $31.9 billion in capital expenditure this quarter alone. Q4 is going over $40 billion. Calendar year 2026 total CapEx is expected to hit $190 billion, with $25 billion of that just from component price inflation. So yeah. They're spending faster than most countries' entire GDP on compute infrastructure, betting that Azure growth "modestly accelerates" in the second half of the year. That phrase — modest acceleration — is doing a lot of work in this guidance.
Nadella also mentioned their Wisconsin data center came online six weeks ahead of schedule, which he seemed genuinely excited about. Which tells you what kind of operational grind they're in right now — celebrating a six-week pull-forward. Gaming is quietly bleeding — Xbox content revenue down 5-7% — and they're quietly trimming headcount through a voluntary retirement program that'll cost roughly $900 million in one-time charges next quarter. Neither of those facts led any press release.
What it means for the stock in plain terms
The stock went up after hours. Of course it did.
But here's the thing — Microsoft trades at roughly 30x forward earnings. That's a valuation that prices in continued strong execution, Azure growth re-accelerating, Copilot seats compounding into genuine consumption revenue, and $190 billion in annual CapEx somehow generating returns that justify the spend. If any of those assumptions wobbles — if Azure stalls instead of accelerating, if the seat-to-consumption model transition takes longer than Nadella's optimistic framing suggests, if the renegotiated OpenAI deal creates friction down the road — the multiple compresses fast.
Institutional money will stay long here because the AI infrastructure narrative is real and the numbers support it right now. The commercial remaining performance obligation hit $627 billion — a staggering backlog. So the revenue is contracted. Getting it out the door fast enough, with constrained supply, is the actual operational problem. The risk isn't that Microsoft is lying. It's that they're in an extraordinarily capital-intensive bet-everything moment, and the ROI on $190 billion in annual CapEx is still largely theoretical at the applications layer.
The one thing beginners should take away
"Beating estimates" doesn't mean a company is cheap, or that buying right now makes sense. Microsoft beat estimates and the stock is priced at a significant premium to its own history because the market is buying the AI story — not Q3's actual numbers. The lesson: always separate whether a quarter was good from whether the price is justified. They're different questions entirely, and confusing them is how beginners overpay for genuinely great companies at bad entry points.