Microsoft (MSFT) — The Life Change the 10-Q Leaves Out
Microsoft's September quarter grew Azure 48% and guided the next to 49–50%, while depreciation rose 56% against revenue up 19%. The 10-Q the July note said would confirm the move of building lives from 15 to 25 years does not mention it, though the lease note shifted the way that change predicts. We value Microsoft at $365.64, from $403: a higher cost of equity explains almost all of the fall, the quarter's operating news nets to about zero, and the method changes and the OpenAI stake roughly offset a halved earnings weight. The 28 October close of $531.01 is 45% above fair value.
In July we valued Microsoft as the best-run machine in the AI buildout, with earnings that run ahead of the bill for building it: a server clock that spreads equipment cost over two to six years, and a new building clock, extended on the July call from 15 years to 25. The July note said it would check the second claim against the first quarterly filing of fiscal 2027. That filing arrived on 28 October with a 48% Azure quarter, $45.2 billion of capital spending, and no sentence about building lives. This note asks what the silence means, what the quarter did to the value of the business, and what $531.01 now requires.
What this note answers
- What does the 10-Q show about the building-life change July said it would verify?
- What did the quarter earn, and how fast is depreciation arriving against revenue?
- How large is the capital bill, and how much of it now sits off the capex line?
- What is Microsoft worth after the print, what moved the number from $403, and what changed in how we value it?
- What does $531.01 require, and what would change our view, on dated tests?
Terms used throughout, defined once. Capex: capital expenditures including finance leases, the basis on which Microsoft guides; where a figure is cash only, it says so. Finance and operating leases: a finance lease is treated like a purchase on borrowed money (the asset sits in property and equipment and counts in Microsoft's capex); an operating lease is treated like rent (it stays out of capex and its cost runs through operating expenses). Free cash flow (FCF): operating cash flow less cash additions to property and equipment. FCFE conversion: free cash flow to equity, after finance-lease principal, as a share of net income, so it runs below free cash flow over net income; share-based compensation stays an expense. Re-mark: the house restatement of server depreciation onto a three-year life (the house clock) against Microsoft's two-to-six-year range; a model estimate, not an accounting restatement, and Microsoft's reported figures comply with GAAP. The 84% terminal conversion is the share of net income left as free cash flow once spending settles at the level that clock implies. Price-to-earnings leg: 24 times re-marked trailing earnings, weighted 20% in the base case. Cost of equity: the return a shareholder requires, here the ten-year Treasury yield plus beta (the stock's measured sensitivity to the market, over five years of monthly returns and adjusted a third of the way toward 1.0) times a 4.5% equity risk premium, the same rule at every update. Remaining performance obligation (RPO): contracted revenue not yet recognized. Growth stages: periods of the forecast, each lasting as long as the thing that drives it. Fiscal calendar: Microsoft's year ends 30 June; FQ2-27 is the December 2026 quarter. In quotations, "3P" is third-party and "H2" the second fiscal half.
The case for the price, in the company's terms
Microsoft has a strong case. Azure revenue was $33.2 billion, up 48%, in the first quarterly report to give Azure in dollars, and it beat the 44–45% the September outlook implied even though, in the CFO's words, "some third-party capacity deliveries we expected in Q1 were delayed." The guide for the December quarter is 49–50% at constant currency. Commercial RPO is $677 billion, up 73%, and the part beyond the next twelve months grew 98%. On the call the CFO said "Customer demand continues to exceed available supply," and, on the efficiency work, "each dollar of capex has the potential to support more revenue and higher returns over time." Operating margin rose to 49.8% from 48.9% while depreciation grew 56%: the company absorbed the bill rather than passed it through. Return on invested capital is 35.1% against a 9.8% cost of equity. And Microsoft owns about 24% of OpenAI, worth about $27.46 a share before tax at the last closed round, which July's cash-flow model did not value separately. Free cash flow stayed positive at $18.6 billion while cash capital spending doubled.
On the clock, the evidence that favors the company is real. GPU rental markets still price the 2020-vintage NVIDIA A100 at $1.54 an hour (Silicon Data's A100 rental index for neo-cloud providers and GPU marketplaces, 3 October 2026), six years after launch, which is what a long economic life looks like. If servers earn for five years or more, the house re-mark overstates the expense. The cost of equity is a measured number, not a law: raw beta is 1.086 (1.058 after the house adjustment) with a standard error of 0.17; one standard error lower, the fair value would be $393, and at July's 8.9% cost of equity, $416. At $531.01 the stock trades at 27.7 times trailing earnings, below the roughly 37 times it briefly paid in 2025 (July note), for a company growing operating income faster than revenue. The note runs one multiple, July's 24 times (§4 argues it); each turn moves fair value by about $3.32.
This note tests four of these claims: the building clock against the filed record; the depreciation wedge against revenue; the capital bill, on and off the capex line, against the cash the business earns; and the growth and discount rate the price implies. It cannot test, inside an explicit model ending in fiscal 2033, whether AI consumption turns Azure into a business that grows faster than the economy for a decade, and it cannot verify the server share of capex beyond what the filed property table and the CFO's statement show. The Azure stage length is carried as a priced sensitivity and a tripwire, not a verdict.
The building clock: what the 10-Q shows, and what it leaves out
The claim. On the 29 July call, as reported in three articles by two news outlets and cited in the July note, the CFO said datacenter and office buildings would be depreciated over 25 years instead of 15, effective fiscal 2027. She said more future datacenter leases would be classified as operating as a result, which was why calendar-2026 capex including finance leases was now expected near $175 billion rather than about $190 billion. The July note counted it as the fourteenth life extension across the large cloud companies in its comparison group since 2020 (thirteen on servers, this one on buildings), against one reversal, and said the first fiscal-2027 10-Q would verify it and quantify the benefit.
What the 10-Q says. Nothing. It carries no change-in-estimate note, no useful-life table and no mention of 25 years; "useful lives" appears once, in the standard list of estimates, and property and equipment is not among the critical accounting estimates. Depreciation was $11.1 billion for the quarter, against $7.1 billion a year earlier. The fiscal-2026 10-K's policy still reads "buildings and improvements, five to 15 years," and the October 8-K that recasts that 10-K for the new segments does not update it (it states that it does not reflect events after the 10-K was filed). The 28 October call did not mention useful lives.
What the record does show. Two things July did not have. First, the company's own written record of its July outlook: the 2 September 8-K reproduces the first-quarter guidance "As provided on July 29, 2026", with capex "Expected to be over $50 billion including the impact of the useful life update." That is a furnished document, not a filed financial statement. It confirms an update sat inside the guidance, but it names no asset class and no number of years; the buildings and the 15-to-25-year detail still rest on the call reports. Second, the lease note moved the way such a change predicts. A lease that runs for most of an asset's economic life is a finance lease, so a longer building life moves datacenter leases to operating treatment, outside capex. Operating-lease assets added in the quarter were $6.5 billion, from $1.4 billion a year earlier and more than all of fiscal 2026; finance-lease additions fell from $9.1 billion to $5.2 billion; and the average remaining operating-lease term rose from six years to eight in one quarter, the strongest sign (lease extensions would lengthen it too). The 10-Q gives no cause; delayed third-party capacity could account for part of the fall in finance leases, but not for the longer operating-lease book.
Why the 10-Q is silent. Not yet effective is unlikely: the July outlook already included "the impact of the useful life update" for the first quarter. Not adopted: the lease note argues against it, though no filed statement confirms adoption. Immaterial fits best. US GAAP requires a change in an asset's service life to be disclosed with its income effect when material, and described when immaterial now but reasonably certain to be material later, in interim reports too (ASC 250-10-50-4 and 50-5; ASC 270-10-50-1; the house's reading of the standard). Silence is consistent with the company judging the effect small, as the CFO's July guide of "a minimal benefit" to fiscal-2027 operating income said. That guide does not cover the larger consequence July reported, about $15 billion moved out of calendar-2026 capex through lease classification, which no filing describes. Exhibit 1's small depreciation step is consistent with a small benefit but does not prove one: late third-party capacity also means less equipment in service.
Exhibit 1 — Depreciation in the quarter against growth in the server book
How to read it. Billions of dollars a quarter, September 2025 to September 2026. Gold bars are depreciation expense as filed; outlined bars are the growth in the gross book of servers, network equipment and software over the quarter (net of retirements). The label under each gold bar is the increase in depreciation from the quarter before. Compare the step with the outlined bar beside it. The one thing it shows: the September step (+$0.8 billion) matched the smallest of the year while the server book grew by a record $29.0 billion.
How large could it be? The filing bounds it. If the 25-year life applied to the whole $182.7 billion book of buildings and improvements at June, and those assets were on average two years old, annual depreciation would fall from about $12.2B to about $6.9B, a benefit near $5.3B a year, about 2.9% of the quarter's annualized operating income. That is an upper bound; "minimal" implies a much narrower set of assets, such as new buildings or shells. Depreciation is not cash, so in the cash-flow model the change is worth nothing; in the price-to-earnings leg the upper bound is worth $2.73 a share of fair value at the 20% weight. We carry it at zero, the company's own description.
What it means. The building-life change stands as the July note described it: sourced to the call, now backed by the company's written outlook and a lease note that moved as predicted, but with no filed statement of the asset class, the years or the effect. July's test, that the 10-Q would quantify it, failed as written, most likely because the company treats the change as immaterial. The fiscal-2027 10-K, due around late July 2027, is the first filed statement whose policy table can show the new life (tripwire 5).
The wedge: Azure accelerates, and depreciation runs faster
The wedge. Depreciation grew 56% against revenue up 19%; it reached 12.0% of revenue from 9.1%. Operating margin still rose, because operating expenses grew 6%, headcount fell 2% and, in the company's account, engineering work let it "monetize underutilized capacity." The December quarter will be harder: cost of revenue is guided up 26–27% against revenue up 18–19%, and operating margin "down slightly." July's model path for fiscal-2027 server depreciation was about $25.6B on Microsoft's own schedule. On the fiscal-2026 capex actual, the guided run-rate and the server share measured from the filing (§3), the same engine now gives $33.7B, and $50.7B for fiscal 2028. Total depreciation is already running at $44.4B a year before any further step.
Exhibit 2 — Azure revenue by quarter, on the restated definition
How to read it. Azure revenue a quarter, billions of dollars, September 2024 to September 2026, on the definition Microsoft adopted in September 2026 (GitHub, Security Copilot and healthcare cloud moved out); the dashed bar is the December 2026 quarter at the midpoint of the 49–50% guide. Year-on-year growth is printed under each bar from September 2025. Compare the growth labels across the row. The one thing it shows: growth held near 40% for three quarters, then stepped up twice, and the guide steps it up again.
Azure. The September restatement gave the history the dollar figure needs: on the new definition Azure grew 40%, 39%, 40% and 42% through fiscal 2026, to $101.9 billion. The September quarter's 48% and the 49–50% guide are a real acceleration, achieved while capacity was short. The CFO tied the guide to "continued efficiency gains across our CPU and GPU fleet helping to offset some 3P GPU as a service contracts that have been delayed to H2", and she warned that later quarters face "a strong prior-year comparable." Outside Azure, revenue grew 6.9%.
The rest of the quarter. Revenue was $92.3 billion, up 19%. Operating income was $45.9 billion, up 21%, a 49.8% margin. Net income was $36.6 billion, or $4.92 a diluted share; excluding a $207 million net gain from the OpenAI investment, $4.89, up 18%, against $4.13 a year earlier. Agents and Infra, the new segment that holds Azure, Microsoft 365 cloud and licensing, grew 25% to $76.8 billion. Devices and Consumer fell 3% to $15.5 billion as Windows OEM revenue dropped 27%. Revenue from commercial arrangements with OpenAI was $10.0 billion, 10.8% of the total, the 10-Q discloses.
The bill: what the capex line counts, and what it leaves out
Capex including finance leases was $45.2 billion, with $7.4 billion of finance leases; cash paid for property and equipment was $39.4 billion, double the year before. The company's capex measure is not a cash figure, so the three do not add, and the 10-Q's finance-lease assets obtained ($5.2 billion) are measured differently again. The company now expects calendar-2026 capex of approximately $170 billion, "due to the 3P capacity delays," against about $175 billion in July. For the December quarter it guides more than $50 billion, and for the second half of the fiscal year it expects spending "to accelerate." The model reads that as fiscal-2027 capex of about $202B including leases ($96B in the first half, the second half 10% higher), of which about $176B is cash. The bear case's second half runs 20% above the first ($212B for the year); the bull case's is level with it ($192B).
What sits outside the line. Because of the building-life change, more datacenters will arrive as operating leases, which never enter capex. The 10-Q shows how large that pipeline is. Leases signed and not yet commenced, primarily datacenters, were $364.6 billion at 30 September (the 10-Q does not split them between finance and operating), up $35.5B in the quarter and more than three times the $106.2 billion of a year earlier, with commencement now running to fiscal 2034. The capex figure the house's calendar-2026 trigger tracks is therefore partly an accounting artifact, and the next update reports an all-in figure beside the company's (tripwire 6). For the valuation the shift changes line items more than economics. Operating-lease rent is paid out of operating cash flow, while finance-lease principal is repaid below the free-cash-flow line, so the reclassification, if anything, makes reported free cash flow a fuller measure of the bill.
Exhibit 3 — Leases signed and not yet commenced, by quarter
How to read it. Billions of dollars at each quarter end, September 2025 to September 2026, as disclosed in each 10-Q and the 10-K; below each bar, the last fiscal year in which those leases begin. Compare bar heights, then the window. The one thing it shows: the signed pipeline more than tripled in a year and lengthened by three fiscal years.
How much of the spend is servers. The re-mark needs one number the filing does not give: the share of capex that is short-lived equipment. July assumed 45% and called it unverifiable. Three readings now say it is higher. The gross book of servers, network equipment and software grew $83.0 billion in fiscal 2026, 62% of the growth in all property and equipment, and $29.0 billion in the September quarter, 69%. On the call the CFO said "Roughly three quarters of our capex was for short-lived assets, primarily CPUs and GPUs." The update uses the filed 62% for fiscal-2026 and later spending, the lowest of the three readings, and keeps 45% for earlier years, where the July model was checked against Microsoft's disclosed 2023 life change. On that basis the house three-year clock defers $21.5B of fiscal-2027 depreciation ($13.3B in July) and $33.2B in fiscal 2028. Trailing GAAP earnings of $19.15 a share, including investment gains and losses, re-mark to $17.70, an overstatement of 8.2%, from 5.8% in July. The price-to-earnings leg uses $16.59, the same figure without the net OpenAI gains, because the stake is valued on its own; other investment items stay in. Because the price-to-earnings leg now carries a 20% weight, the share assumption moves fair value only from $367 at 45% to $365 at 75%.
Who pays. Free cash flow fell 27% to $18.6 billion, 51% of net income excluding the OpenAI gain. Shareholders received $10.2 billion in dividends and buybacks, and repurchases under the buyback program fell to $2.9 billion from $4.0 billion. The CFO's frame for the year is that Microsoft will "remain free cash flow positive in FY27" and "access additional sources of funding when we believe it is attractive." On the base path, fiscal-2027 free cash flow to equity is about $33B, or 23% of net income.
Exhibit 4 — Return on invested capital against the cost of equity, twelve quarters
How to read it. Trailing-twelve-month return on invested capital (after-tax operating income over average equity plus debt less cash), percent, December 2023 to September 2026; the dashed line and shaded band below it mark the 9.8% cost of equity used in this note, held flat. Compare the gold line's slope with its distance above the dashed line. The one thing it shows: the return is more than three times its cost and has fallen about ten points in three years.
Return on capital: strong, and fading. Return on invested capital is 35.1% on a trailing basis (56% with acquired goodwill and intangibles removed), 40.0% a year ago and 45.1% at the start of the window. On the house's rule the trend reads as fading: well above the hurdle, with the spread down 4.8 points in four quarters. The cause is the capital base. Average invested capital roughly doubled, from $182B to $373B, while after-tax operating profit rose 60%, because servers and datacenters go on the books before the revenue they will carry. A return this far above its cost is the strongest fact in the company's favor. The direction is the question the next updates answer.
What Microsoft is worth after the print, and why the number moved
The engine is the July note's: revenue, a net margin excluding investment gains, and the share of net income that becomes free cash flow to equity, discounted at the cost of equity, with a terminal value on the last year's earnings and a price-to-earnings leg at 24 times re-marked trailing earnings. Share-based compensation stays expensed and diluted shares (7,446 million) are held flat, as in July: repurchases have roughly offset stock issued to employees, and the diluted count fell 0.3% in a year (7,466 million to 7,446 million).
The cost of equity, the largest change. It is rebuilt on the house rule: the ten-year Treasury yield of 5.05% on 28 October (4.67% in July) plus a beta of 1.058 (five years of monthly returns against the S&P 500, raw 1.086, adjusted a third of the way toward 1.0; July used 0.90 from a year of daily prices) times a 4.5% equity risk premium (4.75% in July): 9.8%, from 8.9%, worth −$37.37. Measured the July way today (one-year daily beta 1.015, 4.75% premium), the cost of equity would be about 9.9% and fair value $362.10, so the rule itself adds +$3.55 a share; the move comes from the higher Treasury yield and a higher measured beta, partly offset by the lower equity risk premium. The house now measures every cost of equity the same way at every update; the alternatives are in the sensitivities below.
How long Azure grows faster than normal. Each length is argued from the thing that drives it; the model workbook's Stages sheet carries the year-by-year paths.
| Part of the business | What drives the faster growth | How long, and why | Then | Growth while it lasts (bear / base / bull) |
|---|---|---|---|---|
| Azure | Capacity coming online against contracted demand: RPO $677 billion, the part beyond twelve months up 98%, demand above supply | About 3 years, fiscal 2027–29: 30% of the book converts in twelve months and its weighted-average duration is 2.3 years from September 2026, so the average contracted dollar is recognized in fiscal 2029; part of the book runs longer, which is why a longer stage is priced and armed | Four years of fade to nominal GDP growth by fiscal 2033, steepest first (base 14, 10, 7, 4%) | Fiscal 2027 42% / 45% / 47%; then 24–15% / 30–20% / 36–26% |
| Rest of Microsoft (Microsoft 365, licensing, industry, devices and consumer) | Seats and premium per-seat pricing (the E5 and E7 suites, Copilot) against falling Windows OEM and licensing | No supernormal stage: nothing in the filings or the call dates faster growth than today's mid-single digits | Glide to nominal GDP growth by fiscal 2033 | 5.0% / 6.0% / 6.5% in fiscal 2027 |
The stage ends when the contracted book is delivered, not when demand ends; after that, Azure grows with enterprise IT and AI budgets against competitors' capacity. Above-average growth rarely persists once its driver stops: across US companies from 1951 to 1998, sales growth stayed above the median five years running for 6.3% of firms (Chan, Karceski and Lakonishok, 2003). Hence a four-year fade, steepest in the first year after the stage, when, on average, about half of today's contracted book has been delivered and the comparison base is largest, landing on terminal growth by fiscal 2033. The Azure stage is the input most in the price's favor; at one year longer the fair value is $388, at one year shorter $341.
Together the two parts give company revenue growth of 18.0% in fiscal 2027 (July: 15.5%), on Azure of $147.9B (+45%), and $646B of revenue by fiscal 2033. Net margin excluding investment gains is 37.3% in fiscal 2027, the fiscal-2026 level of 37.6% less the September quarter's year-on-year decline (0.2 point) and a further 0.1 point for the December quarter's operating margin, guided "down slightly." In fiscal 2028 it carries half of the extra depreciation step the new capex path brings ($9.8B more than July's), to 35.9%: the September quarter absorbed all of its depreciation increase, so half is the middle reading; the bull case absorbs all of it, the bear case none. FCFE conversion starts at 23% in fiscal 2027 (July: 48%), from the capex path above and finance-lease principal against operating cash flow at fiscal 2026's ratio to net income, and rises to the house-clock terminal of 84% by fiscal 2033.
| $ a share | Bear (25%) | Base (50%) | Bull (25%) |
|---|---|---|---|
| Fair value (base: 80% FCFE, 20% price-to-earnings, plus the OpenAI stake; bull and bear: FCFE plus the stake) | $228.14 | $365.64 | $454.90 |
| OpenAI stake a share (bear: recapitalization mark after tax; base: last round after tax; bull: last round untaxed) | $13.07 | $22.03 | $27.46 |
| Fiscal 2027 revenue growth / net margin / FCFE conversion | 16.5% / — / 14% | 18.0% / 37.3% / 23% | 19.0% / — / 31% |
| Base FCFE alone / price-to-earnings leg alone | — | $329.99 / $398.09 | — |
The expected value at 25 / 50 / 25 is $353.58 (July: $398). Terminal value is 79% of the base case's cash-flow value, which is why terminal growth and terminal conversion are priced below. Excluding OpenAI gains, the price-to-earnings leg is $398.09 (July's, which included them, was $407.09); like for like, trailing earnings grew about 2%. Read as a single-stage multiple at the 9.8% cost of equity and 84% payout, 24 times implies perpetual growth of about 6.1%, above the long-run GDP growth the cash-flow model uses. The multiple is July's, unchanged, and that tension is the second reason its weight is not restored.
What moved the number, and what did not. Each row re-runs the same engine with one more factor changed from July's value to today's, in order, so the rows add to the move. The first row is the cost of equity, re-measured on the house rule; the next seven apply the quarter on July's method; the next four change how the cash-flow model is calculated; the next two value the OpenAI stake on its own; the last is the separate decision on weighting.
| Factor | Kind | $ a share | Fair value after |
|---|---|---|---|
| July note, reproduced (published $403) | start | 403.12 | |
| Cost of equity 8.9% -> 9.8% on the house CAPM rule: 10-year Treasury 4.67% -> 5.05% (2026-10-28); beta 0.90 (one year of daily returns) -> 1.058 (60 monthly returns, raw 1.086, adjusted a third of the way toward 1.0); equity risk premium 4.75% -> 4.5% | cost of equity | -37.37 | 365.75 |
| Valuation date 30 Jul -> 28 Oct (one quarter nearer; Q1's $18.6bn FCF already earned) | data | +3.40 | 369.15 |
| Earnings base: FY2026 EPS $17.95 -> trailing $19.15; diluted shares 7.453bn -> 7.446bn; tax 19.4% -> 20% | data | +10.70 | 379.85 |
| Revenue path: Azure +48% and a 49-50% guide -> FY27 growth 15.5% -> 18.0% | data | +2.85 | 382.70 |
| Net margin: FY27 from the print and the Q2 guide; FY28 carries half the extra depreciation step | data | -2.78 | 379.92 |
| Capex: FY27 incl. leases ~$202bn on the guide -> FCFE conversion 48% -> 26% (before the lease deduction) | data | -9.92 | 370.00 |
| Re-mark: the house depreciation model's capex at the FY26 actual and FY27-28 at the new run-rate (was $115bn flat) | data | -0.63 | 369.37 |
| Re-mark: server share of FY26+ capex 45% -> 62% (measured from the filed gross book) | data | -2.76 | 366.61 |
| Terminal growth = long-run nominal GDP (Fed SEP, 16 Sep 2026): base 4.0% unchanged; bull 4.25% -> 4.0%; bear 3.5% -> 3.0% | dcf-method | +0.00 | 366.61 |
| Fades land on terminal: explicit years FY27-31 -> FY27-33; growth fades on the stage path, conversion rises in a straight line, both land on terminal | dcf-method | -0.52 | 366.09 |
| Mid-year discounting (house convention) from 28 Oct, replacing July's year-end steps | dcf-method | +3.44 | 369.53 |
| Finance leases: FCFE deducts finance-lease principal, a financing cash outflow (FY27 $4.6bn: Q1 actual + scheduled payments less interest); the path still lands on the all-in terminal conversion | dcf-method | -1.59 | 367.94 |
| OpenAI: price-to-earnings leg on earnings excluding net OpenAI gains (trailing $1.12 a share, mostly the $7.6bn net gain in the December 2025 quarter) | openai stake | -10.71 | 357.23 |
| OpenAI: the 24% stake valued once as a non-operating asset, $22.03 a share (24% of the $852bn round after 21% tax on the gain; bull pre-tax $27.46; bear at the recapitalization mark $13.07) | openai stake | +22.03 | 379.26 |
| Weighting: P/E leg 40% -> 20% (earnings converting to cash at about 50%) | weighting | -13.62 | 365.64 |
| Fair value 403.12 → 365.64 | -37.48 | 365.64 |
The model workbook's Attribution sheet carries the same rows. The July configuration reproduces its published $403 ($403.12 unrounded) before any change. Kinds: cost of equity (the house rule), data (the quarter, on July's method), dcf-method (how the cash-flow model is calculated), openai stake (the stake valued once) and weighting (a separate judgment on which methods count).
Exhibit 5 — From $403 to today's fair value: the market, the quarter, then the method
How to read it. Dollars a share; the scale starts at $320, not zero. The first and last black bars are the July and current fair values; each bar between adds (gold) or subtracts (gray) a group of rows from the table, in its order, with the cash-flow-model changes, the OpenAI stake (net of its two rows) and the weighting decision after the dashed divider. Compare bar heights. The one thing it shows: the cost-of-equity bar is larger than every operating bar combined.
Read it plainly. The quarter's operating facts were good and worth about nothing net: one quarter of time (+$3.40), higher trailing earnings (+$10.70, about $4 of it OpenAI's loss quarter leaving the trailing year, which the OpenAI rows take back out) and faster revenue growth (+$2.85) are offset by the margin path (−$2.78), the heavier capex path (−$9.92) and the larger re-mark (−$3.39). Faster Azure growth adds little because July's path already assumed strong growth for five years; the new stage is faster early and fades on schedule.
What the price requires. Hold every other base-case input and solve for one. The cost of equity would have to be 7.7%, against 9.8%. Or revenue growth would have to run far ahead of the base path, enough to reach $944B of revenue in fiscal 2031 against $588B in the base case and $652B in the bull. A fair value equal to the price needs the bull case's operating path and a discount rate below July's.
Exhibit 6 — The value field: scenarios, the July value and the price
How to read it. Dollars a share on one scale. The gold band runs from the bear to the bull case; marks above the line are the bear case, the fair value, the bull case and the 28 October close; marks below are the probability-weighted expected value, the July fair value and the 8:40 p.m. ET after-hours print. Compare the close with the right end of the band. The one thing it shows: the price sits above the bull case.
The inputs a reader will want to move. At July's 8.9% cost of equity the fair value would be $416; at July's Treasury yield, $387; with beta one standard error lower or higher, $393 or $342; on a three-year raw beta (1.23, adjusted as the house beta, a 10.2% cost of equity), $347; on the peer median raw beta (1.43 across Alphabet, Amazon and Oracle, the other three largest US cloud providers, adjusted likewise, 10.9%), $319; a full point of cost of equity either way gives $427 or $323. Terminal growth half a point either way: $348 or $386. Terminal conversion at 80% or 88%: $356 or $376; at the 90% that assumes Microsoft's own clock describes the economics, $380. The Azure stage one year shorter or longer: $341 or $388. The price-to-earnings weight at 40% or none: $379.26 or $352.02. Each turn of the 24 times multiple moves fair value by about $3.32 at the 20% weight.
The receipt: seven tests the house had armed, scored
The July note armed seven triggers on Microsoft. All are graded against the release, the 10-Q and the call; two are not yet due and carry their readings so far.
The 10-Q carries no change-in-estimate note, no useful-life table and no mention of 25 years; July's test as written, that it would quantify the benefit, did not happen. The company's written July outlook names a 'useful life update' without class or years, and the lease note moved the way a longer building life predicts; the 10-Q states no cause (see section 1).
Azure 48% ($33.16bn, the first quarterly report in dollars, new definition), from 42% in FQ4 on the restated basis; Q2 guided 49-50% cc.
$45.2bn on the company's definition, including $7.4bn of finance leases (cash paid for PP&E $39.4bn): 'some third party capacity deliveries we expected in Q1 were delayed'. Q2 guided over $50bn; H2 to 'accelerate'.
Depreciation +63% vs revenue +18% in FQ4-26, then +56% vs +19% in FQ1-27. The growth gap narrowed (46 -> 38 points) but the condition (depreciation growth below revenue growth) is far off, and depreciation reached 12.0% of revenue from 9.1%.
Azure held (48%), but cash capex/revenue rose again: 25.0% -> 36.8% -> 37.3% -> 39.8% -> 42.7% (FQ1-26 to FQ1-27). A light quarter on delivery timing, with H2 guided to accelerate, is not a structural turn.
FQ1 depreciation $11.1bn, a $44.4bn annual run-rate before any further step. The July path was ~$25.6bn server-only for FY2027; on the FY2026 capex actual, the guided run-rate and the server share measured from the gross book, the same engine now gives $33.7bn. The quarterly step was +$0.8bn after +$0.8bn, +$1.1bn and +$1.3bn, while the server book added a record $29.0bn.
The company now guides calendar 2026 capex to approximately $170bn (from ~$175bn in July) 'due to the 3P capacity delays'. Under the trigger as written, that is under on timing. The headline also omits operating leases: $6.5bn of operating-lease assets came on in Q1 alone, more than all of FY2026 ($4.6bn), and leases not yet commenced rose $35.5bn to $364.6bn.
Settled close $531.01 on 28 Oct, 33% above the level. Superseded by this note's tripwire on the new fair value.
No server-fleet life or utilization disclosure. Adjacent evidence, logged on both sides: GPU rental markets (Silicon Data A100 rental index for neo-cloud providers and GPU marketplaces, 3 Oct 2026) still price the 2020-vintage A100 at $1.54/hr, which favors longer economic lives. On the share leg, the filing points the other way from the falsifier: the server line was 62% of FY2026 gross PP&E growth and 69% in Q1, and the CFO put 'roughly three quarters' of Q1 capex in short-lived assets, against the model's 45%.
What carries the story to the next update
Each tripwire states its basis, its tolerance and what the model does with the answer, so the next update applies rules written now. The fiscal second-quarter release is expected in late January 2027 (date not yet confirmed by the company).
- Azure in the December quarter. Basis: the FQ2-27 release, Azure growth at constant currency, against the 49–50% guide. At or above 49%: the base second-half path holds. At 52% or more with second-half commentary of acceleration: the base second half moves to the bull's ($368). Below 46%: it moves to the bear's ($362). A quarter moves little; the length of the stage moves a lot (tripwire 2).
- The length of the Azure stage. Basis: Azure growth in the June 2027 quarter and the fiscal-2027 10-K's RPO split. Azure at 40% or more with the part of RPO beyond twelve months still growing above 50%: the stage runs one year longer ($388). Azure below 30% with that part of RPO growing below 20%: one year shorter ($341). One signal without the other is recorded and changes nothing, in either direction: the stage is set by delivery against contracted demand, so both legs must agree.
- Capex in the December quarter, and calendar 2026. Basis: the FQ2-27 call, capex including finance leases, against "over $50 billion" and the ~$170 billion calendar year. Below $50 billion a second time: the base second half is held level with the first ($370); with Azure at or above the guide the house reads it as the efficiency the CFO describes, with Azure below the guide as a capacity constraint, which also moves Azure's second half under tripwire 1. Above $55 billion: the second half is set 20% above the first ($362).
- The depreciation step. Basis: the FQ2-27 10-Q, depreciation for the quarter. A step of $1.1 billion or more over September's $11.1 billion: the half-absorption margin path holds. A step of $0.8 billion or less again, while the server book adds more than $20 billion: the company is absorbing the bill and the base case moves to full absorption ($372); the building-life benefit and assets not yet in service are the two things to test first. A step of $1.6 billion or more: no absorption ($359); delayed third-party capacity entering service in the quarter is the first thing to test.
- The building clock in the fiscal-2027 10-K. Basis: the policy table and any change-in-estimate note (about late July 2027). Buildings shown at up to 25 years: the fourteenth extension is confirmed in a filed statement, and nothing in the model changes. An effect quantified at $1 billion or more: the house re-mark adds it back to earnings, worth up to $2.73 in the price-to-earnings leg. Policy still "five to 15 years" with no note: the change rests on the call reports and the July outlook as reproduced in the September 8-K, and the house's count of life extensions marks the fourteenth as not confirmed in filed statements.
- Operating leases and the all-in bill. Basis: each 10-Q's lease note, operating-lease assets obtained in the quarter. Above $5 billion again in FQ2-27: the house reports an all-in capex figure (capex including finance leases plus operating-lease assets added) beside the company's in every update, and scores the calendar-2026 trigger on both. Below $3 billion: the September jump is read as a one-off.
- Earnings turning into cash. Basis: free cash flow over net income, quarterly, from the cash-flow statement. Two consecutive quarters at 60% or more: the price-to-earnings weight returns to July's 40% ($379.26 at today's inputs). Two consecutive quarters below 40%: the weight goes to zero ($352.02). Between: 20%, as now.
- The price. Fair value $365.64. A close at or below it resolves the gap. At $274, a 25% margin of safety, the house would look for an entry if the operating tests above hold (this replaces the July note's $400).
What would change our view
The house would move toward the price on any two of the following. Azure at 40% or more in the June 2027 quarter with RPO beyond twelve months still growing above 50% (tripwire 2), which lengthens the stage. Two quarters of free cash flow at 60% or more of net income (tripwire 7), which restores the earnings weight. Two quarters of depreciation steps at or below $0.8 billion while the server book keeps growing (tripwire 4), which would show the bill being absorbed. A disclosed server-fleet life or utilization study supporting five years or more, which would cut the re-mark. Capex below $50 billion again in the December quarter with Azure at or above the guide (tripwires 1 and 3), which would show the efficiency the CFO describes. Or a raw five-year beta one standard error lower, near 0.92 (0.94 adjusted, a 9.3% cost of equity), which alone is worth about $393; or the ten-year Treasury back at July's 4.67%, worth about $387. At July's 8.9% cost of equity the bull case is about $508, still below the price. All of the model-moving items together (that lower beta, a one-year-longer Azure stage, full absorption of the depreciation step and the earnings weight back at 40%) give about $424, still below the price.
The house would move toward the bear case's inputs on any two of the mirror conditions. Azure below 30% in the June 2027 quarter with RPO beyond twelve months growing below 20% (tripwire 2). Two quarters of free cash flow below 40% of net income (tripwire 7). Depreciation steps of $1.6 billion or more (tripwire 4). Capex below $50 billion again in the December quarter with Azure growth below the guide, which would mark a supply limit rather than a timing slip (tripwires 1 and 3). Separately, a raw five-year beta one standard error higher, near 1.26, would on its own lower fair value to about $342, without touching the scenario inputs.
The call
Fair value $365.64 a share (expected value $353.58; bear $228.14, bull $454.90), from $403 on 31 July; the $531.01 close of 28 October is 45% above it, and the $547.30 after-hours print is 50% above it. The quarter confirmed the operating case: Azure is a $33 billion-a-quarter business growing near 50%. It did not shorten the bill: depreciation is growing nearly three times as fast as revenue, and the signed lease pipeline more than tripled in a year. The value fell because the cost of equity we measure rose, on a higher Treasury yield and a higher five-year beta, an estimate whose one-standard-error band runs from about $342 to $393, not because the business weakened. The quarter's facts, the method changes and the OpenAI stake (+$13.51 together) and the halved earnings weight (−$13.62) roughly cancel; the quarter's gains were offset by the cost of building for them. The price asks for the bull case's operating path and a discount rate below July's.
Sources & Method
Company filings. Fiscal first-quarter 2027 results: Form 8-K and Exhibit 99.1, SEC accession 0001193125-26-436122, furnished 28 October 2026 (quarter ended 30 September 2026): revenue, operating income, net income, GAAP and non-GAAP EPS, the OpenAI adjustment, segment revenue, Microsoft Cloud, Azure and RPO growth, capital returned. Form 10-Q for the same quarter, accession 0001193125-26-436133, filed 28 October 2026: income statement, balance sheet and cash-flow statement; Note 1 (accounting policies, estimates, the recast); Note 6 (property and equipment by category, depreciation expense, purchases in accounts payable); Note 11 (RPO, weighted-average duration); Note 12 (leases: right-of-use assets obtained, remaining terms, leases not yet commenced); OpenAI revenue and funding; share repurchases; MD&A critical accounting estimates. Form 8-K, accession 0001193125-26-380280, furnished 2 September 2026, Exhibit 99.1 "FY27 Segments and Investor Metrics": the restated Azure and segment history and the adjusted first-quarter outlook, including the useful-life sentence quoted in §1. Form 8-K, accession 0001193125-26-436149, filed 28 October 2026: fiscal 2024–26 financial statements recast for the new segments (not a restatement), with the auditor's consent. Fiscal-2026 Form 10-K, accession 0001193125-26-323660, filed 29 July 2026: the useful-life policy, the property table, lease disclosures and depreciation for fiscal 2025–26. Fiscal-2026 Forms 10-Q (accessions 0001193125-25-256321, 0001193125-26-027207, 0001193125-26-191507): quarterly depreciation, the gross server book and leases not yet commenced. Fourth-quarter fiscal-2026 8-K, accession 0001193125-26-323632 (29 July 2026). Quarterly cash capex, operating cash flow and net income: SEC XBRL company-concept data (data.sec.gov), fourth quarters derived as fiscal year less nine months. All pulled via EDGAR on 28 October 2026.
Earnings call. Fiscal first-quarter 2027 conference call, 28 October 2026, 5:30 p.m. ET. Every quotation and call figure used is checked word for word against Microsoft's own transcript on its Investor Relations site (retrieved 29 October 2026) and quoted in its wording; two independent transcriptions retrieved on 28 October, The Singju Post and AlphaStreet, agree with it except for hyphenation, capitalization and numerals. All three are archived with the note's data. Call-only figures: capex including finance leases of $45.2 billion and finance leases of $7.4 billion; "roughly three quarters" of capex on short-lived assets; the December-quarter guide (revenue $96.0–97.1 billion, Azure 49–50% at constant currency, cost of revenue $32.7–32.9 billion, operating margin "down slightly"); capex "over $50 billion" in the December quarter and approximately $170 billion for calendar 2026. The July call statements on the building life (15 to 25 years), the lease reclassification and the ~$175 billion calendar-2026 figure are cited from the July note, which sourced them to three articles by two news outlets (Benzinga and CFO Dive).
Market data. MSFT $531.01: the 28 October 2026 settled close on the consolidated tape (SIP), from the brokerage feed. After-hours: $547.30 at 8:40 p.m. ET on the feed's ten-minute extended-session bars, cited as one data point. Beta 1.058: the house rule, a regression of the last 60 monthly log returns (October 2021 to September 2026) against SPY on month-end, split-adjusted closes from the same feed, raw 1.086, adjusted to 0.67 × raw + 0.33; the three-year and peer betas are shown as sensitivities, never used. Ten-year Treasury: FRED series DGS10, 5.05% on 28 October 2026 (the latest print on or before the price date) and 4.67% on 29 July. GPU rent: Silicon Data A100 rental index for neo-cloud providers and GPU marketplaces (SDA100RT), $1.54 per GPU-hour on 3 October 2026, from the house's own capture of the vendor's public page; one vendor, never blended with another index. OpenAI marks: OpenAI's own release of 31 March 2026 ("$122 billion in committed capital at a post money valuation of $852 billion"), confirmed by CNBC and Bloomberg; the recapitalization value from Microsoft's own blog of 28 October 2025; the 24% interest and funding from the 10-Q; quarterly OpenAI effects on net income from the fiscal-2026 10-Qs and 10-K and the September 2026 10-Q (the June 2026 quarter by difference); the pending round from CNBC and TechCrunch, 22 October 2026.
Model. The July engine, reproduced first ($403.12 against the published $403) and then changed one factor at a time; every row of the attribution and every input is in the model workbook (README, Inputs, Scenarios, Summary, Attribution, Stages, ROIC vs WACC, Paths). Explicit years fiscal 2027–2033, mid-year discounting from 28 October 2026 with fiscal 2027's remaining stub at its midpoint and the September quarter's actual free cash flow deducted; terminal value at the end of fiscal 2033 on net income times terminal conversion (84% house clock; 90% bull; 80% bear), growing at the Federal Reserve's longer-run medians of 16 September 2026 (2.0% real GDP, 2.0% PCE inflation; bear one point lower). Cost of equity 9.8% = 5.05% + 1.058 × 4.5%, rounded to a tenth of a point (the house rule, applied at every update). FCFE deducts finance-lease principal ($4.6B in fiscal 2027). Base fair value = 80% FCFE + 20% × 24 × re-marked trailing EPS excluding net OpenAI gains, plus the OpenAI stake a share; bull and bear are FCFE plus the stake; expected value at 25 / 50 / 25. Re-mark: the house v3 vintage depreciation model (published 17 July 2026), unchanged in code, run with fiscal-2026 actual cash capex ($115.9 billion), fiscal 2027–28 at the guided run-rate, and a 62% server share for fiscal-2026 and later vintages (45% before), tax at 20%. Return on capital: the house script, SEC XBRL, twelve quarters, invested capital net of cash. Building-life bound: the June 2026 buildings-and-improvements gross book at a 15-year life against an average two-year-old asset re-lived to 25 years.
Prior TON618 research relied upon. The $116 Billion Question — Microsoft (MSFT) (31 July 2026), which set the $403 fair value, the method, the re-mark and the seven triggers scored in §5. The Depreciation Question (17 July 2026), whose v3 model supplies the re-mark engine. The First Invoice (30 July 2026), the series scoreboard. Meta (META) — The 2027 Bill Is Already Signed (26 October 2026), where the cash-conversion test for the earnings weight was first applied.
Limitations. The building-life benefit is bounded, not measured; the company has not quantified it. The server share is read from the growth of a gross book net of retirements and from one call statement; neither is a disclosed capex split. The fiscal-2027 capex path beyond the December quarter is the house's reading of "accelerate," not a company figure. Beta is window-sensitive; the three-year and peer betas are in the sensitivities. The OpenAI stake is marked at a private round, not a traded price; Microsoft does not disclose its carrying amount or a current fair value, and tax at the 21% federal rate on the gain over funded cost is a simplification. The lease deduction covers leases already on the books; the $365 billion of signed leases not yet started is not split between finance and operating, so its finance share is not in the deduction. Scenario probabilities and the 24 times multiple are judgments, argued in the text.
Disclosures
Information only. TON618 Capital. This report is for information purposes only. Nothing here is an offer to sell or a solicitation of an offer to buy any security, fund interest, or digital asset, and nothing here is personalized investment advice or a recommendation regarding any instrument.
Publisher's exclusion. All research is published solely as general, impersonal information of regular circulation. It is not tailored to the objectives or circumstances of any individual and is not issued in connection with compensation from any client. The Fund has no clients and distributes all research free of charge. On that basis it publishes in reliance on the publisher's exclusion from the definition of "investment adviser" under the Investment Advisers Act of 1940 (§202(a)(11)(D); cf. Lowe v. SEC, 472 U.S. 181 (1985)).
Registration & conflicts. TON618 Capital is not registered as an investment adviser or broker-dealer in any capacity. The Fund is a Bitcoin fund and may hold or transact in the securities or digital assets it discusses; material conflicts are disclosed where they exist. Ownership: the Fund holds no position in MSFT as of the report date, and it is not a Fund holding or a BTC-correlated instrument. Compensation: the Fund received no compensation from any party in connection with this report and charges nothing for it.
Use of AI. Artificial intelligence is used in the creation of this research. All methodology and data integrity are reviewed and approved before publication by TON618 Capital's Chief Investment Officer, Keyth Beck; errors may nonetheless occur, and readers should verify independently.
CFA. This report was prepared to align with CFA Institute analytical standards (methodology only). CFA® and Chartered Financial Analyst® are registered trademarks owned by CFA Institute. That reference describes the analytical framework applied; it does not imply the report was prepared, reviewed, or authored by a CFA charterholder, and the report is not issued, reviewed, endorsed, certified, or approved by — nor affiliated with — CFA Institute.
Risk & feedback. Past performance is not indicative of future results. Digital assets and equities are volatile and may result in total loss of capital. Corrections and feedback are welcome — please direct them to CIO Keyth Beck at keyth@ton618capital.com. Version 1.0.