Microsoft is not an expensive stock that got more expensive. Over the past 52 weeks MSFT has returned 1.34% while fiscal 2026 revenue grew 17.7% and diluted EPS grew roughly 25% — the multiple did the falling, not the business. That is the first thing to understand about any Microsoft MSFT stock prediction made at today’s $513.53: you are paying 26.0x forward earnings for a company that was on 33x a year ago, and the market has spent twelve months quietly re-rating the safest name in software.
Here is the reason, and it is the single number that decides both the bull and the bear case. In fiscal 2026 Microsoft’s operating cash flow rose $46.8bn, from $136.16bn to $182.94bn — a 34% increase, the largest in the company’s history. Free cash flow fell, from $71.61bn to $66.99bn. Capital expenditure absorbed every dollar of that record cash-flow growth and $4.6bn more besides. Having tracked hyperscaler capex cycles since the 2023 Azure inflection, I have not previously seen a quarter where Microsoft’s own CFO volunteered, unprompted, that the company expects “to remain free cash flow positive” — a sentence no chief financial officer says about a business earning $35.8bn a quarter unless the arithmetic has become genuinely close. The bull case at $675 and the bear case at $400 are two readings of that same cash-flow statement.
The pattern is not a software pattern — it is a utility pattern
What Microsoft is doing to its balance sheet has a precise precedent, and it is not in technology. It is the capital cycle that ran through US telecoms between 1997 and 2002 and through regulated power utilities in the 2000s: a business with excellent unit economics decides that the constraint on future revenue is physical plant, borrows against its cash flows to build that plant, and in the process converts itself from an asset-light compounder into a capital-intensive one. Investors do not pay software multiples for capital-intensive businesses. They pay infrastructure multiples.
The evidence that this repricing is already underway is in the gross margin. Microsoft’s quarterly gross margin has fallen from 69.35% in the September 2024 quarter to 67.20% in the June 2026 quarter — 215 basis points of erosion during the most profitable stretch the company has ever had. Depreciation on data-centre assets flows through cost of revenue, and the depreciation base is compounding at roughly 80% a year. That is the mechanism by which capex becomes a margin story, and it works with a two-to-three-year lag, which means most of it has not arrived yet.
The counter-argument is equally concrete, and it sits in the backlog. Commercial remaining performance obligation — contracted revenue Microsoft has not yet recognised — grew 84% to $678bn. That is more than twice fiscal 2026’s entire revenue, booked and signed. A utility builds capacity against a regulator’s demand forecast. Microsoft is building against $678bn of executed contracts. Whether that distinction is worth 35x earnings or 20x earnings is the whole argument.
Key facts
- Spot price $513.53 (close, 28 August 2026); market capitalisation $3.81trn on 7.43bn shares — StockAnalysis.com, 31 August 2026
- Q4 FY2026 revenue $90.0bn, +18%; operating income $40.6bn, +18%; net income $35.8bn, +31%; diluted EPS $4.81, +32% — Microsoft Q4 FY2026 press release, 29 July 2026
- Azure and other cloud services grew 43% in constant currency; Microsoft Cloud revenue $59.3bn, +27% (Microsoft, 29 July 2026)
- Commercial remaining performance obligation $678bn, +84% year over year — Q4 FY2026 earnings call transcript, 29 July 2026
- FY2026 operating cash flow $182.94bn (+34.4%); cash capex $115.95bn (+79.6%); free cash flow $66.99bn, down 6.5% from $71.61bn — StockAnalysis.com quarterly cash-flow data
- Q1 FY2027 capex guided at over $50bn including finance leases, against $41bn in Q4 FY2026 (earnings call, 29 July 2026)
- Analyst consensus: 55 analysts, Strong Buy, average target $569.45, high $870, low $400; FY2027 consensus revenue $391.08bn and EPS $19.75 — StockAnalysis.com forecast page, 31 August 2026
- 52-week closing range $352.83 (25 June 2026) to $542.07 (28 October 2025)
What is actually happening: the depreciation clock got moved
The most consequential disclosure of Microsoft’s fiscal 2026 results was not the revenue beat. It was an accounting change that most coverage treated as housekeeping.
On the earnings call, CFO Amy Hood said: “We are extending the estimated useful life of our data centers and office buildings from 15 to 25 years, reflecting our operating history and expected use of these assets.” Mechanically, this spreads the cost of a building over 25 annual instalments instead of 15, which reduces the annual depreciation charge on that portion of the asset base by roughly 40% and increases reported operating income by the same amount.
Two things make this worth pausing on. First, the change applies to shells and land improvements, not to silicon: management noted that roughly two-thirds of recent capital spending has gone to short-lived assets, primarily CPUs and GPUs, which continue to depreciate on a much faster schedule. So this is not a wholesale flattering of earnings — but it is not nothing either, because the building share of a $116bn annual programme is still tens of billions of dollars.
Second, and more interesting: Microsoft is now depreciating its data-centre buildings over a period longer than the entire commercial life of Azure, which launched in February 2010 — sixteen and a half years ago. The company is asserting, in its accounting policy, a confidence about 2051 that no one can currently test. That is a defensible judgement given how long well-sited data-centre shells genuinely last. It is also, unavoidably, a judgement that raises reported profit today and defers the cost to a decade in which today’s management will not be running the company.
The same reclassification is why headline capex figures moved around confusingly this summer. Microsoft’s expected capital spending for calendar 2026 was restated from roughly $190bn to about $175bn, a change that reads as a $15bn cut but is not one. Hood addressed it directly: “outside of this useful-life impact, our calendar year 2026 capex investment expectations remain unchanged.” The money is being spent. It is simply being classified differently.
Meanwhile the physical build is running into physical limits. The Guardian reported on 17 August 2026 that Microsoft’s AI roadmap is being constrained by a shortage of chips rather than a shortage of demand — a framing Hood corroborated on the call when she said Microsoft remains “focused on delivering efficiencies that help us bridge the gaps we see as customer demand continues to exceed supply.”
What Microsoft and its neighbours are actually doing about it
The named response across the sector is remarkably consistent, and it points the same way: everyone is trying to own more of the stack in order to protect the margin the stack is eating.
Microsoft presented its Maia 200 AI accelerator at Hot Chips 2026 on 26 August, its second-generation in-house training and inference silicon. The strategic logic is straightforward — every workload moved from merchant GPUs to internal silicon converts a supplier’s gross margin into Microsoft’s own. The execution risk is equally straightforward: in-house accelerators have repeatedly under-delivered against roadmaps across the industry, and Microsoft’s own capacity commentary implies it is still buying merchant parts as fast as it can get them. Our coverage of Nvidia’s bull and bear case lays out the other side of that trade in detail, and AMD’s positioning shows how contested the merchant tier has become.
The power constraint is drawing a different kind of response — from communities rather than companies. Tom’s Hardware reported on 29 August 2026 that a Microsoft-backed data centre is facing multiple complaints over 62 unpermitted gas turbines, alongside allegations of unpermitted construction and noise pollution. This matters commercially rather than reputationally: permitting friction is now a real determinant of how fast guided capex can actually be deployed, and a project that cannot be energised is capital sitting idle on the balance sheet earning nothing.
Elsewhere in the peer group the same capital cycle produces sharply different balance-sheet outcomes. Oracle is funding its build with materially more leverage; CoreWeave is a pure-play version of the same bet with none of Microsoft’s cash cushion; and Meta has pushed part of its build off balance sheet entirely. Microsoft is the only one of the four financing the whole programme out of operating cash flow while still returning capital. That is the bull case’s real foundation, and it is not a small one.
Market impact: the coverage cushion is thinner than it looks
Combine two datasets that are usually read separately — the cash-flow statement and the capital-return schedule — and a number emerges that neither shows on its own.
In fiscal 2025 Microsoft returned $42.50bn to shareholders ($18.42bn of buybacks, $24.08bn of dividends) against $71.61bn of free cash flow: a payout of 59%. In fiscal 2026 it returned $48.72bn ($22.27bn of buybacks, $26.45bn of dividends) against $66.99bn of free cash flow: a payout of 73%. The cushion between what Microsoft generates and what it has committed to distribute narrowed by fourteen percentage points in a single year, and it narrowed while the business was performing exceptionally.
Now layer on the guidance. Q1 FY2027 capex is guided at over $50bn including finance leases, against $41bn in Q4 FY2026. Annualise even a flat $50bn per quarter and the fiscal 2027 programme is $200bn-plus on that basis. Operating cash flow would need to grow another 25–30% simply to hold free cash flow at today’s level — on top of the 34% it grew last year.
That is the context in which Hood’s line lands. “We expect to remain free cash flow positive in FY ’27” is not a boast. It is a floor being set, and the fact that a floor was thought necessary tells you where the distribution of outcomes now sits.
The two cases, side by side
| Bull case — $675 | Bear case — $400 | |
|---|---|---|
| Implied move | +31.4% | −22.1% |
| Market cap | $5.02trn | $2.97trn |
| FY2027 P/E | 34.2x | 20.3x |
| Anchor | The first $5 trillion close | Lowest of 55 published targets |
| Requires | Azure holds ~45% growth; capex intensity peaks in FY2027; RPO converts on schedule | Azure decelerates below 35%; capex runs past FY2027; depreciation compresses margins |
| Breaks if | Free cash flow turns negative in any quarter | RPO conversion accelerates and capex growth flattens |
The $675 bull case is deliberately anchored to a round structural milestone rather than a target price: at $675 Microsoft’s market capitalisation is $5.02trn on 7.43bn shares — the first $5 trillion close in market history. On consensus FY2027 EPS of $19.75 that is 34.2x, above Microsoft’s own five-year average and well inside the $870 high target that 55 analysts currently span. It requires Azure to hold something close to the ~45% constant-currency growth guided for Q1, and it requires fiscal 2027 to be the year capital intensity peaks.
The $400 bear case is anchored to something harder: it is the single lowest of those 55 published targets, and it sits 13.4% above the $352.83 close Microsoft actually printed on 25 June 2026. This is not a crash scenario. It is a return to a price the stock traded at ten weeks ago, on a multiple — 20.3x forward earnings — that Microsoft has held for extended periods within the last decade. The bear case does not require anything to go wrong. It requires the capex cycle to last one year longer than the bulls expect.
Regulatory and physical tension
The binding constraints on Microsoft’s plan are increasingly neither financial nor competitive. They are permits, transformers and turbines.
The unpermitted-turbine complaints reported in August are one instance of a broader pattern: US grid interconnection queues now run to multiple years in several of the regions where hyperscale capacity is most wanted, and local objections to noise, water use and air permits have become a routine feature of siting. For a company that has guided to more than $50bn of quarterly capital spending, a six-month permitting delay is not a public-relations problem — it is billions of dollars of deployed capital producing no revenue while its depreciation clock runs.
There is a second regulatory axis with a direct earnings consequence. Depreciation-life extensions are a recognised area of audit and SEC comment-letter attention precisely because they move reported profit without moving cash. Microsoft’s 15-to-25-year change is well within the range peers have adopted and is supported by genuine operating history, so it is unlikely to be challenged. But it does mean that any investor comparing Microsoft’s fiscal 2027 operating margin to its fiscal 2025 operating margin is comparing two different accounting policies, and should adjust before drawing conclusions about operating leverage.
The third is competitive-regulatory: Microsoft’s cloud licensing terms remain under scrutiny in the European Union, and any remedy that makes it cheaper for customers to run Microsoft software on rival clouds would attack Azure growth precisely where the bull case is least tolerant of disappointment.
What happens next
Three predictions, with the reasoning attached.
First: fiscal 2027 free cash flow lands between $55bn and $75bn, and the range matters more than the midpoint. The causal chain is arithmetic. If operating cash flow grows 25% to roughly $229bn and capex on a cash basis grows in line with the guided total to roughly $160bn, free cash flow is about $69bn — flat. Shift either variable by ten percentage points and the answer moves $20bn. Hood’s “remain free cash flow positive” language is consistent with a management team that models the bottom of that range as reachable.
Second: the gross margin bottoms in fiscal 2028, not fiscal 2027. Depreciation on assets placed in service during the fiscal 2026 and 2027 build lands with a lag, and the short-lived GPU share of that spend depreciates fastest. Expect the quarterly gross margin to print in the 65–67% band through fiscal 2027 before stabilising. A print below 65% would be the clearest single signal that the bear case is winning.
Third: the $678bn RPO becomes the most-watched line on the release. Backlog is what justifies the build. If RPO growth decelerates from 84% toward the 30s while capex is still rising, the market will reprice the spending as speculative rather than contracted — and that, not a revenue miss, is the specific sequence that takes MSFT toward $400.
Satya Nadella, Chairman and CEO, framed the whole strategy on the results as follows: “We are advancing the frontier on the cost-to-outcome curve, ensuring every customer can turn tokens into business results.” That is a precise description of the bet. Microsoft is spending an unprecedented amount of capital to make each unit of AI output cheaper, on the conviction that cheaper output expands the market faster than it compresses the price. For most of the last three years that conviction has been rewarded. Fiscal 2027 is the year the cash-flow statement finally puts a number on it.
Frequently asked questions
What is the Microsoft MSFT stock prediction for 2027?
Our framework sets a bull case of $675 and a bear case of $400 against a spot price of $513.53. The $675 bull case corresponds to a $5.02trn market capitalisation and 34.2x consensus FY2027 EPS of $19.75. The $400 bear case is the lowest of 55 published analyst targets and 20.3x the same earnings figure. The consensus average sits at $569.45.
Why did Microsoft’s free cash flow fall in fiscal 2026?
Operating cash flow rose 34.4% to $182.94bn, but cash capital expenditure rose 79.6% to $115.95bn. Because capex grew faster than the cash it was funded from, free cash flow fell from $71.61bn to $66.99bn — a 6.5% decline in the same year the company posted record revenue and earnings.
What does Microsoft’s change to a 25-year useful life mean?
Microsoft extended the depreciation period for data centres and office buildings from 15 to 25 years. This lowers the annual depreciation charge on those assets by roughly 40% and raises reported operating income accordingly. It does not change cash flow, and it does not apply to servers, CPUs or GPUs, which remain on much shorter schedules.
Is Azure still accelerating?
Yes, as of the most recent report. Azure and other cloud services grew 43% in constant currency in the June 2026 quarter, up from 40% in the prior quarter, and management guided to approximately 45% constant-currency growth for the September quarter. Management also stated that customer demand continues to exceed available supply.
What would invalidate the $675 bull case?
A quarter of negative free cash flow, a gross margin print below 65%, or commercial RPO growth decelerating from 84% toward the 30s while capital expenditure is still rising. Any of the three would signal that the spending has outrun the contracted demand supporting it.
How does Microsoft’s capex compare with its peers?
Microsoft is funding a roughly $116bn annual cash capital programme entirely from operating cash flow while still returning $48.7bn to shareholders. Oracle is using materially more leverage, Meta has moved part of its build off balance sheet, and CoreWeave is a pure-play version of the same bet without a legacy cash engine behind it.
This article is analysis and information, not investment advice. Prices and market data are as of 31 August 2026 and will change.