Microsoft

IThe business

Microsoft makes most of its money from companies. They pay for its software by the seat and for its computing by use. In FY2026 revenue was $331.8B, up 18%, and operating income was $155.2B, a margin of 46.8%.² Since FY2017 revenue has grown 14.7% a year and operating income 20.5% a year.¹

The three reporting segments are three different businesses. Productivity and Business Processes, which holds Microsoft 365, LinkedIn and Dynamics, earned a 60% operating margin in FY2026. Intelligent Cloud, which holds Azure, server software and GitHub, earned 41%. More Personal Computing, which holds Windows, XBOX, search and devices, earned 27% and shrank 1%.²

How it makes money

• Microsoft 365 for companies. $102.0B in FY2026, products and cloud together.R (reported) Microsoft 365 commercial cloud revenue grew 17% on 6% more paid seats, so revenue per seat rose about 10%.C (calculated) Volume-licensing customers are generally invoiced once a year, at the start of each contract period. Microsoft does not disclose Copilot revenue.²
• Azure. $101.9B in FY2026, up 40% on the September 2026 restatement.C (calculated) Microsoft’s own figure on its old definition is 41%. Customers pay for what they use, and demand is larger than the capacity Microsoft can switch on.⁴⁸
• LinkedIn, XBOX, Windows and search. $73.9B together.C (calculated) LinkedIn grew 11% to $19.8B. XBOX fell 7% to $21.8B. Windows and Devices brought in $17.1B and search advertising $15.2B.²
• OpenAI. OpenAI paid Microsoft $24.1B in FY2026, including revenue-sharing payments, or 7.3% of revenue.C (calculated) Microsoft owned about 25% of OpenAI at 30 June 2026.²

Azure is shown on the restated basis Microsoft adopted in September 2026; the other lines are on the FY2026 10-K basis. On the new basis, everything outside Azure brought in $229.9B in FY2026, up 9.9%.C (calculated)²⁸

What changed after FY2023

Since FY2023 Microsoft has spent like Amazon and Alphabet. Counting datacenters taken on finance leases, capital spending went from 14.7% of revenue in FY2023 to 42.4% in FY2026 on the filing basis, or $140.6B; Microsoft’s quarterly figures sum to $145.3B.C (calculated)¹⁴ Net property and equipment went from $95.6B to $313.1B in three years, and depreciation from $11.0B to $34.3B.¹² Nadella says Microsoft will roughly double its total capacity in two years.⁴

Software pays for the build, and the build’s returns set the value.

IIThe moat

I think Microsoft has two moats. The one in its software is widening. The one in its cloud is real but shared, and its returns have come down.

The first is the switching cost in its software. Over a million of Microsoft’s 1.6 million security customers use four or more of its security products.⁶ A company that runs its identity, security, mail and meetings on Microsoft would have to move every employee at once to leave. The evidence is in pricing. Productivity and Business Processes raised its operating margin from 53.2% in FY2023 to 59.9% in FY2026 while its revenue grew 49%.C (calculated) Revenue per Microsoft 365 commercial seat rose about 10% in FY2026, up from about 2% in FY2023, while seat growth slowed from 11% to 6%.C (calculated)²¹⁰¹²

The AI products carry that moat into the next platform. Microsoft has over 30 million paid Microsoft 365 Copilot seats, and net seat additions more than doubled from the March quarter to the June quarter. EY put the new E7 suite on 400,000 employees. GitHub has 225 million users, and 50 million of them use GitHub Copilot. All of the June quarter’s growth in contracted revenue came from customers other than the frontier model labs.⁴

The second moat is scale in cloud infrastructure. The 10-K names three economies of scale: large datacenters cost less per unit, pooled demand raises utilization, and shared tenancy cuts maintenance work.² Amazon and Alphabet have them too. I think scale lets the top three price below a small entrant; it does little to separate them from each other. Synergy Research puts Microsoft at 20% of cloud infrastructure in both the June 2025 and June 2026 quarters, while Google rose from 13% to 15%, and it names CoreWeave, OpenAI, Oracle and other second-tier clouds as the fastest growing.²⁶³⁰

The returns show what the build costs. Each dollar of capital added from FY2023 to FY2026, excluding the goodwill and intangibles bought with Activision, earns about 23 cents a year after tax, down from 36 cents for FY2020 to FY2023 and level with the 22 cents for FY2022 to FY2025.C (calculated) Including Activision the figure is 18%. Both are well above my 9% cost of capital, so the build creates value. Returns have held because the rest of the business got more profitable: its operating margin before depreciation rose from 47.0% to 57.1% in three years, while each dollar of net new capital brought $0.43 of new revenue in FY2026.C (calculated) Before FY2020 Microsoft’s tangible capital was close to zero, so earlier returns on it are not meaningful.¹²

Microsoft Cloud gross margin fell from 70% in FY2022 to 66% in FY2026, and to 65% in the June quarter, as AI capacity took a larger share of the mix; gross profit still rose from about $64B to $141B.C (calculated)²⁴¹³ OpenAI, the largest single customer, gained the right on 27 April 2026 to serve all its products from any cloud.¹⁹

Customer demand continues to exceed available capacity.

Microsoft 365 revenue per seat rose about 10% in FY2026, on 6% more seats.

IIIThe competitive advantage

Microsoft’s advantage over Amazon and Alphabet is the customer relationship. It already holds the enterprise contract, and its Agent 365 control plane had nearly 40 million AI agents registered two months after launch. Nearly 90% of FY2026 Microsoft Cloud revenue came from customers other than the frontier model labs.⁴ The backlog is less spread out. OpenAI was about 45% of commercial contracted revenue at December 2025, and the rest grew 28%, 26% and 25% in the December, March and June quarters.⁴⁵⁶

The comparison is close but inexact. Microsoft’s trailing year is FY2026. Its quarterly capital figure including finance leases was $41.0B, and after $24.6B of datacenters taken on finance leases its FY2026 free cash flow was $42.4B.C (calculated) Amazon’s capital spending is net of proceeds and incentives. Microsoft does not report Azure’s operating income; its Intelligent Cloud segment, which adds server software and GitHub, earned $16.0B on $39.3B in the quarter, a 41% margin.C (calculated)^1342425

Where Microsoft leads

Distribution comes first: Copilot and agents reach buyers through contracts they already hold. Funding is tighter than the reported figures show. On reported figures FY2026 operating cash flow of $182.9B covered $115.9B of cash capital spending and $48.7B of dividends and buybacks.² Adjusted for $30.8B of payment and tax timing, operating cash flow of about $152B did not cover $164.6B of cash capital spending and payouts, and cash and short-term investments fell $17.7B.C (calculated)¹² Alphabet’s cash flow covered its build with $53.3B to spare, and it still sold $49.6B of common and mandatory convertible preferred stock in June; Amazon’s trailing free cash flow was negative.²⁴²⁵ Landlords fund part of Microsoft’s build. At 30 September it had signed $364.6B of datacenter leases that have not yet started, up $35.5B in a quarter and equal to 9.3% of its market value.C (calculated)³⁴²

The second lead is neutrality. Azure’s Foundry offers more than 11,000 models, including OpenAI’s and Anthropic’s, and Anthropic committed in November 2025 to $30B of Azure compute.⁴²⁰ Hood’s case is that the same capacity serves whichever model wins:

Think about that infrastructure as being pretty fungible.

Where it trails

Google Cloud is growing twice as fast as Azure from a smaller base, and its 82% includes Workspace and sales of TPU systems, the chips Alphabet designs for itself and now sells to others. AWS has 8 more points of market share.²⁵²⁶ The deeper gap is the cost of computing. About two-thirds of Microsoft’s capital spending went on short-lived assets, mainly CPUs and GPUs, in the June quarter, and about three-quarters in September. It fills its fleet with chips from NVIDIA and AMD alongside its own.³⁵⁴ NVIDIA earned a 71% gross margin on $216B of sales in its year to January 2026; that margin is part of the price of every GPU Microsoft buys from it.C (calculated)³¹ Microsoft’s own chip, Maia 200, is earlier in its life; the company says it delivers 30% better performance per dollar than the latest hardware in its fleet, a claim I could not check.⁴ And since April 2026 OpenAI can take its work to any cloud.¹⁹

IVManagement

Satya Nadella has run Microsoft since February 2014 and chaired its board since June 2021. Amy Hood has been chief financial officer since 2013. Under Nadella, revenue rose from $86.8B to $331.8B and diluted earnings per share from $2.63 to $17.95, which the company puts at 11.8% and 17.4% a year.²⁹ Including dividends, the stock returned 24.9% a year from 31 January 2014, the last month-end before he took over, to Q4 2026.C (calculated) Most of that is earnings. From June 2014 the price rose 12.7 times: earnings per share rose 6.8 times, and the price-to-earnings multiple went from 15.9 to 29.5.C (calculated)²⁹³²

Nadella holds both the chair and the chief executive role. Sandra Peterson is the lead independent director, and 11 of the 12 board nominees are independent.⁹

Capital allocation

The deal record is mixed, and the best deal was an investment.

On Microsoft’s own figure, $11.9B of funding bought a stake worth about $135B, and OpenAI paid Microsoft $24.1B in FY2026 alone.²¹⁸ Activision was the largest deal and is the weakest so far: XBOX revenue shrank 7% in FY2026 and XBOX took impairment charges.²⁴

From FY2017 to FY2026 Microsoft paid $180.2B in dividends and $205.3B for its own shares. $38.8B of that went to withholding employee taxes on vested stock, so program buybacks were $166.5B, and dividends plus program buybacks took 64% of free cash flow on Microsoft’s definition.C (calculated) In FY2026 the program bought 36 million shares for $16.7B, an average of about $464 a share, against my base value of $367.C (calculated) The share count fell 4.9% in ten years and has been flat since FY2023, because buybacks now offset the shares issued to employees.¹² Stock pay was $12.4B in FY2026. In September the quarterly dividend rose 8% to $0.98.¹²¹

Incentives

Nadella’s FY2026 pay was $90.6M on the summary table, and his stock award is all performance shares. Its target value was $50M; the proxy says the required accounting model shows it higher in the tables.⁹ Those shares vest on four revenue growth measures, each against the business plan: Azure, weighted 35%; the rest of Microsoft Cloud, 35%; consumer services, 15%; and XBOX content and services, 15%. A three-year comparison of shareholder return with the S&P 500 then scales the award between 75% and 150%. Microsoft’s latest three-year return ranked at the 46th percentile, so the scale stayed at 100%.⁹

His cash bonus pays 70% on Incentive Plan revenue and operating income against the budget, and for FY2027 it moves to half revenue, half operating income.⁹ Operating income carries the build’s depreciation, so the bonus feels its cost. The stock award, 86% of his reported pay, does not.C (calculated) Neither measures return on capital or free cash flow, in years when Microsoft reinvests three-quarters of its after-tax operating profit. Pay does follow the share price: the proxy’s compensation actually paid, which marks unvested stock to market, was $30.1M for FY2026 against $131.1M for FY2025, and $100 in Microsoft stock at June 2021 was worth $143 at June 2026 against $260 for the Nasdaq Computer Index.⁹ Nadella owns 918,226 shares, about $487M at $530.14, or about five years of his FY2026 reported pay.C (calculated)⁹ I give management high marks for operating and for the OpenAI bet, and a low mark for a stock award that pays on revenue growth when the open question is the return on the capital that buys it.

The stock award pays Nadella to grow the segments that absorb the capital, and his $487M stake is five years of pay. Over the cycle, Activision cost $75.4B and XBOX revenue is falling, while $11.9B put into OpenAI became a stake worth about $135B.²¹⁸

VFinancials

Operating income rose 5.3 times from FY2017 to FY2026. Owner free cash flow rose 14%.C (calculated) Both are true because Microsoft now reinvests most of what it earns.¹

I define owner free cash flow as operating cash flow less cash capital spending, less datacenter assets taken on finance leases, less stock pay. Microsoft’s own free cash flow leaves out the last two. The leased assets are datacenters Microsoft pays for over a weighted average of 13 years at 4.5%, and the stock pay is a cost the buybacks now cover. From FY2027 most new datacenter leases will be operating leases, which never appear as capital spending. The tables here leave them out, because the filings do not split datacenters from other operating leases, so owner free cash flow from FY2027 looks better against earlier years than it should. My model counts all capital, owned or leased.¹²⁴

The share reinvested rose every year from 22% in FY2021 to 76% in FY2026. In FY2017 it was below zero: depreciation and customer prepayments covered all the spending.¹

Growth equals the share reinvested times the return on it. In FY2026 Microsoft reinvested 76% of after-tax operating profit at about 23% on new capital, which gives about 17.5%; after-tax operating profit grew 18.1%.C (calculated) Microsoft guides FY2027 free cash flow, on its own definition, to be positive.¹⁴

Trailing twelve months to September 2026

Q1 FY2027 adds a quarter to the picture. Revenue was $92.3B, up 19%, and operating income $45.9B, a 49.8% margin. Operating cash flow was $58.0B, but cash capital spending was $39.4B against $19.4B a year earlier. Owner free cash flow was $10.3B in the quarter and $13.5B in the quarter a year earlier. Over the trailing twelve months it is $26.7B, against $30.0B in FY2026, and Microsoft reinvested about 80% of its after-tax operating profit, against 76% in FY2026.C (calculated)¹³⁴

The trailing column adds Q1 FY2027 to FY2026 and takes out Q1 FY2026. Finance-lease assets are the right-of-use assets obtained in exchange for finance lease obligations in the 10-Q. Microsoft’s call figure for the quarter is $7.4B because it counts principal at lease start. The table leaves out operating leases. Assets obtained on them were $6.5B in the quarter against $1.4B a year earlier, and $9.7B over the trailing twelve months. That covers all operating leases, not only datacenters. Taking all of it out would leave the trailing figure at $17.1B, against $25.4B for FY2026 on the same basis.C (calculated)¹³⁴³⁵

Operating margin rose from 30.1% in FY2017 to 46.8% in FY2026, through the whole build.¹ Spending discipline paid for it: research and development fell from 13.5% of revenue to 10.7%, and headcount went from 221,000 in FY2022 to 223,000 in FY2026 while revenue per employee rose from $0.90M to $1.49M.C (calculated)¹²¹³ Gross margin peaked at 69.8% in FY2024 and fell to 67.9% as lower-margin Azure revenue grew.¹

That discipline now runs into depreciation. On the capital in my base case, depreciation rises from 10.3% of revenue in FY2026 to 16.8% in FY2028 and about 22% by FY2030, because two-thirds of the spending is on servers written off over six years.E (estimate) To hold the FY2026 margin, costs before depreciation would have to fall by about 11 points of revenue by FY2030; I assume about 5.E (estimate)¹²⁴

Accounting to watch

1. Asset lives. Microsoft has lengthened useful lives three times in six years. Servers went from three to four years in FY2021 and to six in FY2023, which added an estimated $3.7B to FY2023 operating income. Datacenter buildings go from 15 to 25 years from FY2027.⁴¹³ Hood describes the spending as short-lived assets, mainly CPUs and GPUs.⁴ On the $215.9B of servers and network equipment at cost in June 2026, each year cut from the six-year life would add depreciation of about $7.2B a year at five years and $18.0B at four, or 2.2 and 5.4 points of FY2026 margin.E (estimate) In my model a five-year life lowers the base value to $348 and a four-year life to $326; a seventh year, if Microsoft extends again, lifts it to $384.E (estimate)²
2. Cash timing. Property purchases still owed to suppliers rose from $6.9B at June 2025 to $26.7B at June 2026, so FY2026 cash spending ran $19.8B below the spending booked.C (calculated) Cash taxes of $21.2B were $11.0B below the tax charge, mostly through $14.2B of deferred taxes, a deferral that lasts as long as spending keeps growing.C (calculated) Pay the suppliers and FY2026 owner free cash flow was about $10.2B; reverse the tax deferral too and it was about zero.C (calculated)¹²
3. Investment gains. GAAP net income of $133.7B includes a $5.0B after-tax net gain from OpenAI over the year and a $3.2B pretax gain on Anthropic in the June quarter. Adjusted earnings per share of $17.28 remove the OpenAI effect only. Without the June quarter’s one-off items, which include the Anthropic gain, they are about $17.01.C (calculated)²³⁴

VIValuation

How this goes badly wrong

Before the numbers, I ask what would most likely wreck this and work back. I see three ways.

1. The capacity outruns the demand. Microsoft has signed $364.6B of datacenter leases that have not started, and they arrive whether or not the customers do.³⁴² If capacity catches up by FY2028, new capital earns about 9% to 11%, near its cost. This is the most likely of the three, and my bear case is built on it.
2. The backlog is narrower than it looks. OpenAI was about 45% of contracted revenue at December 2025, and the rest has slowed from 28% to 25% growth.³³⁴
3. The software earns less per seat. Paid Microsoft 365 seats grew 6% in FY2026, down from 11% in FY2023.²¹² If agents do work seats used to do, seats and prices soften, as in my bear case.

What an owner can take out is less than the profit. Trailing owner free cash flow is $26.7B, after stock pay and after all capital spending.¹³⁴ Earnings without investment gains are about $132B, the $3,937B market value divided by 29.8.C (calculated)³³³⁶ So an owner gets about a fifth of reported earnings in cash, a 0.68% yield against an earnings yield of 3.4%., and the rest goes into datacenters. I cannot tell from the filings how much of that only keeps the business standing still.

I cannot know whether each new dollar of capital keeps bringing $0.42 of revenue or climbs toward $0.90. On that input alone my base runs from $210 to about $670 a share.C (calculated) So $367 is the middle of a wide range. It is not a measurement.

What the price implies

At $530.14, my base case needs a discount rate of about 7.5% to reach the price, about 2.5 points over the 4.98% ten-year Treasury.C (calculated) Microsoft’s market value is $3,937B, 29.8 times trailing adjusted earnings without the investment gains and 26.9 times the FY2027 consensus. Its earnings yield of 3.4% sits below the Treasury’s 4.98%, and trailing owner free cash flow yields 0.68%.C (calculated)^333363738

Run my model backwards, holding its margins and capital needs. The price then needs revenue to grow 15.2% a year for ten years, to about $1.37 trillion in FY2036.C (calculated) With everything outside Azure growing as in my base, about 5% a year, that means Azure compounding about 25% a year, to about $925B, against 18% and $526B in my base.C (calculated) If the operating margin held at FY2026’s 46.8% instead of falling with depreciation, the price would need 12.8% revenue growth.C (calculated) Microsoft grew 14.7% a year over the last nine years, from a base less than a third the size, with a quarter of today’s capital intensity.¹⁸

The price needs Azure to grow about 25% a year for ten years, to about $925B, on my margins and capital needs.

The base-case inputs are my estimates; each needed value comes from solving my model for the price with every other input unchanged.

Growth and capital separate my value from the price. If growth needed no new capital at all, my base revenue would be worth about $670 a share, above the price; if each new dollar of capital keeps bringing only the $0.42 my base starts from, it is worth $210.E (estimate) Swapping one bull input at a time into my base adds $63 for growth, $52 for capital needs, $39 for margins and $19 for growth after FY2041.E (estimate) Taken together, in the order the chart shows, they reach the consensus case at $588.¹

What the price pays for the cloud gives the same answer from the other side. I value the two software segments at multiples of their FY2026 after-tax operating profit, at a 20% tax rate, and read off what is left for Intelligent Cloud.E (estimate) The segments changed to Microsoft’s two-segment basis with its Q1 FY2027 report. The table keeps the FY2026 segments, because the new basis has no full year yet. In Q1 Agents and Infra earned $40.6B on $76.8B of revenue, a 53% margin, and Devices and Consumer $5.4B on $15.5B.²³³³⁶

At the middle row the price pays 46 times Intelligent Cloud’s FY2026 after-tax profit, 1.8 times the multiple I give the software that funds it.C (calculated)²³⁶

Bear, base and bull

The model runs fifteen years, FY2027 to FY2041. Revenue is Azure plus everything else, with ten years of explicit growth and five in which growth fades in a straight line to the terminal rate. The operating margin is a margin before depreciation, which I estimate, less depreciation, which the model builds from the capital the growth needs: two-thirds of gross capital on servers over six years and one-third on buildings over 25, on top of the assets Microsoft already holds.²⁴ Each year’s owner free cash flow is after-tax operating profit less the net new capital needed for the next year’s growth. Tax is 20%, Microsoft’s guide for FY2027. Stock pay counts as a cost. I add cash and investments and subtract debt and finance leases, a net $6.3B at June 2026, after adding my carrying value for equity investments. At 30 September the same sum was minus $4.1B, because Q1 spent more than it earned after leases; the model starts in June and already contains that spending.¹²⁴³⁴

I discount at 9.0%. The capital asset pricing model gives 9.1%: the 4.98% Treasury plus Damodaran’s 4.20% implied equity risk premium for 1 October 2026, times Microsoft’s ten-year beta against the S&P 500 of 0.97.C (calculated) Its five-year beta of 1.10 gives 9.6%. Damodaran states his 4.20% against a 5.29% Treasury yield.R (reported) On that pair the model gives 9.4% at the ten-year beta, and I keep 9.0%.C (calculated) The 7.5% the price needs implies an equity premium of about 2.6 points at a beta of 0.97.C (calculated)^27282937

My base FY2027 now has revenue up 18% and the operating margin flat at 46.8%, on gross capital of about $210B.E (estimate) Q1 is why. Before the report my base had revenue up 16%, the margin down to 45.8% and gross capital of $196B. Microsoft’s own guide points to a similar amount of capital: it now guides calendar 2026 capital spending to about $170B, down from about $175B, because some third-party deliveries slipped into the second half. That leaves about $97B for the six months to December after $72.9B in the first half, and Q1 was $45.2B with Q2 guided above $50B.C (calculated)⁴⁵³⁵ My base already assumes the return on new capital rises from 16% in FY2027 to 29% in FY2036, as capacity built now starts earning, and new revenue per $1 of new capital from $0.42 to $0.90. Owner free cash flow reaches 27% of revenue by FY2036, about where it was in FY2021.E (estimate)¹

The bear case is stranded capacity. Capacity catches up with demand by FY2028, OpenAI spreads its work across clouds, and the $364.6B of signed leases arrive anyway, so new capital brings $0.30 of revenue for three years and earns about 9% to 11%, near the cost of capital. The margin falls to about 31% as depreciation lands on slower revenue.E (estimate)²¹⁹

My bull case is the consensus case. Its after-tax operating profit per share lands within 6% of consensus earnings per share in every year from FY2027 to FY2030: $20.23 against $19.71 in FY2027, and $32.63 against $33.49 in FY2030, a year that rests on four estimates.C (calculated) At my 9% rate the consensus case is worth $588; the price is 10% below it, so the market either doubts consensus or uses a higher rate than I do. At my bull revenue, consensus FY2030 earnings need a 48% operating margin while depreciation reaches about a fifth of revenue.C (calculated) My base now sits level with consensus for FY2027, which Q1 supports, and 23% below by FY2030.³⁸

The value after FY2041 is 58% of my base, 46% of the bear and 61% of the bull.C (calculated) None of the cases removes OpenAI’s revenue-sharing payments when they end in 2030, or adds back the payments Microsoft no longer makes to OpenAI; neither amount is disclosed.²¹⁹

Only the 7.5% row comes near the price, and only with growth after FY2041 of 3.5% does it reach it; at 3.0% it is $526, $4 short.C (calculated) One point on the discount rate moves my base by 18% to 26%: $301 at 10% and $462 at 8%.C (calculated)¹³⁶

Two items sit outside the cases. I carry the OpenAI stake at book value inside $36.3B of investments. At Microsoft’s October 2025 figure of about $135B, less the $11.9B funded as a stand-in for the carrying value, which is not disclosed, and after 21% tax, it adds about $13 a share.E (estimate) The IRS seeks $28.9B plus penalties and interest for 2004 to 2013, about $3.90 a share, and Microsoft contests it. With both, my base is about $376.C (calculated)²¹⁸

The margin of safety

• Price$530.14Close, Q4 2026 [36]
• Base value$367My estimate E (estimate)
• Margin of safety below$27525% under my base value C (calculated)

I set the margin of safety at 25% below my base value, which gives $275. My rule for a discount to base value is 25% for a business with net cash and no customer above a tenth of revenue, 30% when one cycle or two customers drive results, and 35% when fixed charges such as leases sit ahead of the shares. Microsoft meets the 25% test only on a narrow reading. Cash and short-term investments less debt was $29.0B at 30 September; subtract finance leases too, as my model does, and it is minus $4.1B. OpenAI was 7.3% of FY2026 revenue and 10.8% of the September quarter’s, $10.0B of $92.3B; over the trailing year it is under a tenth. I keep 25%, and that is a judgement call. At 30% the line is about $257. At 35%, for the $71.0B of finance leases and $364.6B of signed leases that sit ahead of the shares, it is about $239. If a full year of OpenAI passes a tenth, I move to 30%. The discount also covers one point on the discount rate, which moves my base by 18% to 26%.²³⁴ The price is 93% above that line, so the stock would have to fall about 48% to reach it.C (calculated) The lowest close of the past year was $352.83 on 25 June 2026, still 28% above it; the stock had fallen 35% from its $542.07 close on 28 October 2025, and has risen 50% since.C (calculated)²⁷

The next best use of the same dollar is a check. The closest comparable on this site is Alphabet; my base case returns 5.4% a year from its price, against 4.5% for Microsoft. The ten-year Treasury yields 4.98%.³⁷

How good the business is and how good the price is are two questions. The business earns a 46.8% operating margin and 23.0% on capital added since FY2023.¹² The price is 44% above my base value. On those numbers the first answer is easier to defend than the second.

The case also needs several things at once: Azure near 25% a year, a margin near 40% and about $0.90 of revenue per new dollar of capital. If I give each a 60% chance, which is my judgement and not a measurement, all three together come to about 22%.E (estimate)

Where this could be in 5 and 10 years

Value in a year is my model’s value at that date: the owner cash after it, discounted at 9.0%, per today’s share count and in that year’s dollars, with the value after FY2041 included. It leaves out today’s net cash. Year 5 is FY2031 and year 10 is FY2036. Owner cash is the sum for FY2027 to FY2036, not reinvested.

Return a year is price growth plus owner cash, from $530.14 to the year-10 value.

For my base to be right at five years, Azure has to reach about $366B in FY2031, up from $102B, with each new dollar of capital bringing about $0.60 of revenue and the operating margin near 39%. Owner cash that year is $111B, against $75B in FY2030. That puts the base about level with today’s price. At ten years it needs revenue of $913B, a margin near 40% and about $0.90 of revenue per new dollar of capital. Owner cash is then $248B a year and still growing.

The cases differ on three numbers. In the bear, Azure slows to about 4% in FY2028, new capital brings only $0.30 to $0.60 per dollar and the margin falls to about 32%. In the bull, Azure compounds 22% a year to $753B, the margin stays near 46% and capital brings $0.50 rising to $1.10. All three use 9.0%. At 8%, my base today is $462 instead of $367.

The most likely way my base is wrong is capital. If each new dollar keeps bringing only the $0.42 it starts from, my base today is $210, and the year-5 and year-10 values drop to about $320 and $410.

A reader’s guide: these are estimates of value, not forecasts of the share price, and the price can sit above or below value for years.

VIIThe case for the price

What most people miss

Azure beat its guide on less new capacity than planned. It grew 48% against a guide of 44% to 45%, while capital spending came in at $45.2B against more than $50B, because third-party deliveries ran late.⁴⁸³³³⁵ Hood said engineers found underused CPU capacity they could charge for, and got more out of the GPU fleet.³⁵ Capital is the input I doubt most, and better capital needs alone add $52 to my base.E (estimate) Hood also said these gains are harder to predict.³⁵

The second fact answers my bear case, stranded capacity. In January Hood said most of the capital, and a lot of the GPUs, are “already contracted for most of their useful life”. She added that margins on sold-out assets improve with age.⁶ I can’t check that. The 10-Q gives one contract length, 2.3 years on average, and it mixes Azure with Microsoft 365.³⁴

The bull case

Bill Ackman makes the clearest public case. Pershing Square started building its stake in February and disclosed it on 15 May.⁴⁰ He wrote on X that it paid “21 times forward earnings”.⁴¹ He called the OpenAI change “a deliberate pivot toward a more open, multi-model architecture”, and backed the spending as the source of future revenue.⁴⁰ Press reports of his post say he also argued the multiple left out the OpenAI stake, about $200B on his count.⁴¹⁴⁸ The 13F for 30 June lists 6,206,730 shares, $2.3B, 11.9% of the firm’s reported total, up from 5,654,078 at 31 March.C (calculated)⁴²⁴³

Brad Gerstner said on CNBC that Microsoft is “misunderstood” on what it gets from AI, as Insider Monkey reported in October 2025. That is a year old.⁴⁴ Altimeter’s 13F for 30 June lists $473M, 4.8% of what it reported.C (calculated)⁴⁵ After the September quarter Morgan Stanley raised its target to $650. CNBC quoted its note: the AI investments can drive “durable high-teens revenue growth and > 20% earnings growth”.⁴⁶ The 10-Q puts the stake at about 24%, or about $204B at March’s $852B round, with liquidation rights that differ from that share. On my method that is about $20 a share, against the $13 I use.E (estimate)⁴⁷³⁴

Against reality

Ackman’s shares were valued at about $370 each at 31 March.C (calculated)⁴³ The price is $530.14, or 26.9 times the FY2027 consensus rather than 21.³⁶³⁸ My base is $367, 31% under the price.

In my model, say Azure compounds 22% a year to $753B by FY2036, and the rest grows about 6%. Say the operating margin settles near 45.5% by FY2030, against 40% in my base. Say each new dollar of capital brings $0.50 of new revenue, rising to $1.10. That is worth $555 a share at my 9% rate, with growth after FY2041 left at 3.0%.E (estimate) If Microsoft does those three things and hits those targets, the price is possible.

Year one is close. Azure is running at 48% to 50%, against the 49% that path needs for FY2027.³³³⁵ The margin before depreciation was 61.8% in the quarter, against 61.0% needed, though last year’s September quarter also ran above its full year.C (calculated)³⁴

The distance is in the years after. Azure must then average about 19.5% for nine years, to reach 7.4 times its FY2026 revenue.C (calculated) Depreciation climbs from 12.0% of revenue to about 20.5% by FY2030 while the margin barely falls, so other costs must keep shrinking.E (estimate) Roughly three-quarters of Q1 capital was short-lived, against two-thirds in my model, which brings depreciation forward.³⁵ New revenue per $1 of capital was $0.43 in FY2026, $50.1B on $116.2B, below the $0.50 where that path starts.C (calculated)¹

What they have to hit

Thresholds come from my signposts and my model. Where they differ, I show both.

Where I come down

The efficiency evidence is the best thing here, and it helped move my FY2027. It did not move FY2036. One good quarter is not ten years, and I can’t check Hood’s contract claim. If new revenue per $1 passes $0.60 by FY2028, with Azure above 25% and depreciation under 15%, I would raise my values.

VIIIRisks, and what would change my view

The risk Microsoft names in its own filing is the one I weight most:

Overestimation of demand or misalignment of capacity investments may result in underutilization

1. OpenAI. OpenAI was 7.3% of FY2026 revenue and about 45% of commercial contracted revenue at December 2025. In the September quarter it paid $10.0B, 10.8% of revenue, and Microsoft’s stake was about 24%, down from 25%.^{34 26} If all $24.1B were Azure, Azure excluding OpenAI was about $78B in FY2026.C (calculated) Since April 2026 OpenAI can serve its products from any cloud, and its revenue-sharing payments to Microsoft are capped and end in 2030. The same amendment ended Microsoft’s payments to OpenAI and kept Azure as OpenAI’s primary cloud, where its products ship first.²⁸¹⁹ Its $250B incremental Azure commitment from October 2025 is only as good as OpenAI’s own funding.¹⁸ The 10-K says contracted revenue involves judgment about “potential renegotiation of commitments”.²
2. The capital commitment. Microsoft had $743.8B of contractual obligations at June, including $329.1B of datacenter leases signed but not started, with terms of up to 20 years. By September the leases signed but not started were $364.6B.²³⁴ Hood’s answer in July was that two-thirds of capital spending goes on short-lived servers and chips, and “if the demand environment changes, you just slow down”. The share was about three-quarters in September.³⁵ She says the land and building commitments are flexible too; the $364.6B of signed leases says they are less so.²⁴ My bear case prices that in.
3. Margins. Microsoft Cloud gross margin was 66% in the September quarter, guided to about 65% for December, as the mix moves to Azure. Depreciation rose from $15.2B in FY2024 to $34.3B in FY2026 and was $11.1B in Q1 alone, up from $7.1B, and in my base it passes a fifth of revenue by FY2030.E (estimate)^1243435
4. Seats. AI agents could let companies do the same work with fewer Microsoft 365 seats. Seat growth has already slowed from 14% in FY2022 to 6% in FY2026, and was 5% in September on a measure that now includes GitHub seats. The 10-K now lists AI-first application companies among Office’s competitors, and Microsoft is moving to seat-plus-usage pricing to get ahead of it.²⁴¹³
5. Regulation and the rest. Microsoft runs independent compliance functions for the EU’s Digital Markets Act and Digital Services Act, and says US and foreign antitrust authorities “continue to scrutinize our business”.² XBOX is shrinking, and Windows OEM and devices revenue was guided down in the high teens for FY2027; in Q1 it fell 22%.⁴³⁴

What would move me

The first test: Q1 FY2027, graded

Microsoft reported on 28 October. Revenue was $92.3B, up 19%, against a guide of $89.85B to $90.95B. Operating income was $45.9B, up 21%, a 49.8% margin. Adjusted earnings per share were $4.89, up 18%. Azure grew 48%, in constant currency as well, against a guide of 44% to 45%. Capital spending including finance leases was $45.2B, below the guide of more than $50B, because some third-party capacity arrived late. Cash paid for property and equipment was $39.4B.^48223335

The quarter confirms the near term and does not change the decade. I raised FY2027 Azure growth from 40% to 45% and the FY2027 operating margin from 45.8% to 46.8%, then lowered the fade so Azure still reaches $526B in FY2036. A faster start needs more capital sooner, which cancels the benefit: my base stays at $367, my bear moves from $170 to $174 and my bull from $590 to $588. The bear moves because 32% Azure growth in FY2027 is no longer possible with the first half running near 49%.E (estimate)³³³⁵

Four things went the other way. Leases signed but not started rose $35.5B in a quarter, to $364.6B. OpenAI was 10.8% of revenue. Owner free cash flow was $10.3B in the quarter against $13.5B a year earlier, and $26.7B over the trailing year against $30.0B in FY2026. And on 29 October, the day after the report, Microsoft priced $25.0B of notes due 2028 to 2066, $23.0B at fixed coupons from 4.850% to 6.350% and $2.0B at floating rates, after Hood said debt markets were an option it had not yet used.C (calculated)³⁴³⁵³⁹ The notes add cash and debt together, so my net cash figure does not change.

The stock barely moved. It closed at $531.01 before the report and $530.14 after it, against $524.51 on the Friday before, and the 10-year Treasury yield moved from 5.00% to 4.98%.^36371623 Consensus earnings per share for FY2027 moved from $19.65 to $19.71.³⁸¹⁷