Microsoft closed the year ended June 30, 2026 with $331.8 billion in revenue, up 18%, or 16% once currency swings are stripped out. That is an enormous business. If you want to see what AI is doing to software, it is the wrong number.
In the fourth quarter of that year, Intelligent Cloud grew 32%. Azure and other cloud services grew 43%. More Personal Computing fell 4%, and Xbox content and services fell 10%. Same company, same quarter, three different economies. Azure is sprinting. The PC and gaming businesses are taking a knee. Average them into one growth rate and you will miss the only split that matters.
That internal split is the useful version of a point Citi’s Tyler Radke has been making in public: AI is sorting software into winners and losers. The sorting has already started inside Microsoft.
One company, three speeds
The fourth-quarter segment print, from Microsoft’s July 29, 2026 release, is where the story gets specific. Two cloud figures sit side by side and look like they contradict each other. They do not. Intelligent Cloud is the Azure-heavy segment. Microsoft Cloud is a broader bucket that also pulls in commercial Microsoft 365 and other cloud revenue.
- Intelligent Cloud delivered $39.3 billion, up 32%, or 31% in constant currency. Azure and other cloud services grew 43%. Microsoft does not publish a quarterly dollar figure for Azure here, and it did not give a constant-currency rate for that line. The 43% is a growth rate. The dollar milestone is annual: chief executive Satya Nadella said Azure revenue surpassed $100 billion for the first time this year.
- Microsoft Cloud was $59.3 billion, up 27%. That is the figure chief financial officer Amy Hood highlighted. It is larger than Intelligent Cloud because it counts cloud revenue living in other parts of the company.
- Productivity and Business Processes was $37.8 billion, up 14%. Microsoft 365 commercial cloud rose 14% as reported and 16% on an adjusted basis. Last year’s comparison was helped by two percentage points of revenue recognized in that period, so the reported 14% understates the adjusted 16%. Still mid-teens. Nowhere near a cloud-style surge.
- Inside that segment, the speeds differ again. Microsoft 365 consumer cloud grew 24%, or 22% in constant currency. LinkedIn grew 12%, or 10% in constant currency. Dynamics 365 grew 13%, or 12% in constant currency.
- More Personal Computing was $12.9 billion, down 4%, or 5% in constant currency. Windows OEM and devices fell 7%. Xbox content and services fell 10%. Search advertising, excluding what Microsoft pays partners for traffic, rose 10%, or 9% in constant currency. One bright spot. The rest of the segment is shrinking.
A sentence that says Microsoft is growing above 20% because of AI is wrong on the company total. It is right only if you are talking about cloud, and especially Azure. Nadella also said Microsoft 365 Copilot had reached more than 30 million paid seats. That is the cleanest adoption number in the release. It counts seats. These results do not say what Copilot contributed to the top line.
Commercial remaining performance obligation (signed contracts not yet counted as revenue) stood at $678 billion, up 84%. That says contracted demand at Microsoft has not collapsed. It does not say every customer’s AI budget grew. Multi-year cloud deals can swell that backlog even while the applications sitting on top of the cloud grow at a mid-teens pace.
Budgets do not grow. They move.
“AI splits software” is a claim about a finite IT budget. Software, in this telling, is intact. In a Bloomberg Television interview, as summarized by BeInCrypto, Radke, Citi’s co-head of US software equity research, described AI projects absorbing corporate technology spend. Platforms tied to infrastructure, data, and bundled copilots collect the money. Traditional application vendors and smaller point tools get squeezed.
He named Microsoft, Palantir, and Snowflake on the winning side. He put Adobe, and smaller application vendors, under pressure.
For anyone buying software, the live issue is procurement. The stock chart is the smaller part of the story. AI for business is not showing up as a blank check. Picture the budget meeting. The pilot needs money, and the envelope is flat, so something else gives. A renewal that used to pass on habit gets a harder look. Seats nobody opens come off the order. A project that cannot show an outcome gets killed. The copilot gets bundled into a bill the company already approved. That is how AI automation and AI agents get funded without the IT budget growing.
Bundling is the advantage. Microsoft can put a copilot next to tools people already pay for. A smaller vendor has to win a fresh budget fight every quarter. Distribution wins the default fight. It does not win every renewal. A specialist that owns a workflow people refuse to leave, and can show a result the bundle cannot, is not automatically a loser.
AI for sales is a decent stress test. Dynamics 365 grew 13% and LinkedIn grew 12%. Healthy. Not heroic. The fast money sits in the cloud under the agent. The sales application on top of it does not. A pilot that cannot point to a shorter cycle or a better win rate is living in the part of the stack that gets squeezed.
A 20% expectation, and a year that grew 18%
In that same Bloomberg appearance, as BeInCrypto reported it, Radke said he expects Microsoft’s revenue growth to top 20%. That is an account of a television interview, not a transcript and not Microsoft’s guidance. Which revenue line, which period, and whether the figure is a formal Citi model or a verbal expectation are not established.
What Microsoft reported is clearer. Full-year revenue of $331.8 billion rose 18%. That does not confirm a company-wide rate above 20%. It also is not a clean miss of a forecast that was never pinned to this year or this total. The headline and the result talk past each other until someone specifies the line and the period. The June-quarter company-wide growth rate was not broken out in the release, so do not treat 18% as a quarterly figure either.
Profit is the easy way to talk yourself into the wrong conclusion. Full-year GAAP net income was $133.7 billion, up 31%. Exclude OpenAI investment accounting and net income was $128.8 billion, up 22%, or 20% in constant currency. OpenAI gains added $4.963 billion to the GAAP profit figure. A year earlier, OpenAI losses had pulled the other way. Operating income did clear 20% on a reported basis ($155.2 billion, up 21%, or 19% in constant currency). That is still not revenue. A 30% profit print is not evidence that AI already pushed Microsoft’s top line past 20%.
Liking the franchise, paying a lower multiple
A separate Citi moment, from before the July 29 report, is easy to mash into the television remarks. It should not be. Barchart reported that Radke and his team cut their price target to $570, a level then described as about 43% upside. The cut was pinned on multiple compression in enterprise software, framed as 25 times estimated 2028 earnings. The prior target was not disclosed.
He stayed positive. The note cited checks particularly on Copilot and described Microsoft as strategically positioned for an era of optimizing token spend (what it costs to run AI models) and AI efficiency. The caveat going into the print was that investors would have to absorb heavier capital spending and a conservative fiscal 2027 operating-margin outlook while Microsoft invests in AI infrastructure. The earnings release does not settle that margin outlook. The target, the upside, and the share-price context around them are pre-earnings snapshots.
The combination is the point. Citi can like the franchise and still pay a lower multiple for it. Microsoft sits on the winning side of Radke’s map because of distribution, cloud scale, and a copilot attached to an installed base, not because the valuation is expanding. Tokens without a result are a cost center with a brand name. Seats without a result are a temporary courtesy.
Questions worth sitting with
Is Microsoft already growing revenue faster than 20%?
Not on the company total for the year just reported. Revenue rose 18%, or 16% in constant currency, according to Microsoft. Radke has said he expects growth to top 20%, but the interview account does not specify the line or the period. The year does not confirm the headline. It also is not a clean miss of a fully specified forecast.
Where is growth actually above 20%?
In cloud. Fourth-quarter Intelligent Cloud rose 32%. Azure and other cloud services rose 43%. Microsoft Cloud revenue was $59.3 billion, up 27%, and Nadella said Azure revenue passed $100 billion for the year. Productivity software is mostly mid-teens, though Microsoft 365 consumer cloud grew 24%. Personal computing, Windows, and Xbox declined.
Does the profit growth confirm a revenue call above 20%?
No. Full-year GAAP net income rose 31%, but OpenAI investment gains added $4.963 billion and inflated that rate. Excluding OpenAI, net income rose 22%. Profit growth and revenue growth are different claims. Revenue is the comparison that matches what Radke was reported to have said, and company revenue grew 18%.
What does “AI splits software” mean if you buy or sell software?
Budgets are being reallocated. They are not magically enlarged. In the framing reported from Radke’s interview, Microsoft, Palantir, and Snowflake sit on the side that captures spend tied to AI infrastructure, data, and bundled copilots. Adobe and smaller application vendors have to fight for what is left. Your vendor’s problem is whether it sits in a bill already approved or has to re-win its budget every quarter.
Why cut the price target and stay bullish?
Ahead of the July 29, 2026 report, Citi cut its target to $570 (still about 43% upside at the time) because enterprise-software multiples had compressed, framed as 25 times estimated 2028 earnings. The positive case was strategic position and Copilot checks, not a richer multiple. That target and the upside math are pre-earnings snapshots.
What is worth watching instead of the headline growth rate?
Commercial remaining performance obligation of $678 billion, up 84%. That is signed contracts not yet counted as revenue. It says demand at Microsoft has not collapsed. It does not say every AI budget grew, and it does not say what Copilot contributed to the top line. Copilot’s more than 30 million paid seats are adoption. They are not revenue. The filter that decides who keeps the money is whether the spend in your own budget is tied to a measurable result.
The mistake is to hear “AI” and expect every software line to levitate. Microsoft’s own year refuses that story. One engine is compounding well above 20%. Another is mid-teens, trying to attach copilots to seats people already pay for. A third is shrinking. If you buy software, sell it, or sit on the board that funds both, stop asking whether AI grows the category. Ask which budget it just took, and whether you are the one who received it.