Research note 02 · Equity valuation · 8 min read
Microsoft: strong business, insufficient margin of safety
A $427 base valuation against a $493.95 close. Our preferred entry is near $340, conditional on the cash-flow thesis holding.
Our view: Microsoft is a strong business, but $493.95 offers too little margin of safety. Our discounted cash-flow model puts base intrinsic value at $427 a share. For a new long-term position, our preferred entry threshold is about $340, conditional on the operating thesis remaining intact.
This is an estimate of business value, not a forecast of where the stock must trade. The comparison uses the 8 September 2026, 4 p.m. ET close, displayed by Microsoft Investor Relations through Refinitiv. Financial evidence comes from the FY26 results and 29 July earnings call; forecasts below are our assumptions. Price and results.
The price is above our base value.
Intrinsic value scenarios · USD per share · same 9% discount rate
01 / The software franchise funds the cloud bet
Productivity and Business Processes generated $84B of FY26 operating profit on $140B of revenue: a 60% margin. Intelligent Cloud added $57B on $138B, growing revenue 30%. These are substantial existing earnings engines. Our positive business view rests on them, not a separate speculative valuation for “AI.” FY26 segment results.
There is evidence of demand beyond one customer. Management reported commercial remaining performance obligations of $678B, up 84%; excluding OpenAI, growth was 25%. Only about 30% of commercial RPO is expected to become revenue within twelve months. Backlog supports visibility, but it is neither cash in the bank nor proof of attractive returns on the infrastructure required to serve it. Earnings call.
Our interpretation: the business merits a premium for recurring demand and high profitability. The unresolved question is how much of that growth investors must reinvest before they can collect it.
02 / A lower capex headline could mislead
FY26 operating cash was $182.9B. Subtracting $115.9B of cash property-and-equipment spending leaves $67.0B. Subtract another $12.4B of stock compensation as an economic cost and $3.1B of finance-lease principal, and this more demanding cash diagnostic falls to $51.5B. It is not our DCF cash-flow definition; it shows why $133.7B of net income cannot simply be treated as distributable cash. 10-K cash flow and lease note.
The forward complication matters more. Microsoft said FY27 changes to estimated building lives would move more future datacenter leases into operating leases, which are excluded from its capex measure. Its revised $175B calendar-2026 capex outlook therefore does not, by itself, establish lower economic investment. Management still expects FY27 capex to grow and FY27 Q1 spending to exceed $50B. Call and outlook.
Our model charges finance-leased investment to capital expenditure and operating rent to operating margins. A change in accounting classification is not an efficiency gain.
03 / What must happen to earn $427 of value
We project the three segments separately over ten forward years. In the base case, Intelligent Cloud growth fades from 28% to 6%, Productivity from 13% to 4%, and Personal Computing recovers from −2% to 3%. Consolidated operating margin declines from 46% to 44.5% as the business mix shifts toward cloud and lease costs persist. The 20% tax assumption follows FY27 guidance; the annual growth paths are ours.
| Base assumptions and outputs | Year 1 | Year 5 | Year 10 |
|---|---|---|---|
| Revenue | $388B | $666B | $981B |
| Operating margin | 46.0% | 44.5% | 44.5% |
| Capital expenditure¹ | $194B | $187B | $196B |
| Depreciation / amortization | $54B | $133B | $137B |
| Cash flow to the firm² | $0.3B | $180B | $288B |
¹ Cash PP&E plus new finance-leased investment; operating rent stays in expenses. ² After operating tax and reinvestment, with stock compensation retained as an expense. Forward annual periods use FY26 as the starting run rate; these are not company fiscal-year forecasts.
The central bet is demanding: annual capital investment stays near $190–205B while revenue almost triples over ten years. Spending remains substantial; its burden relative to sales falls. Installed assets drive higher depreciation, narrowing the gap between cash investment and accounting expense. Faster hardware replacement, higher rent or weaker utilization would undermine that recovery.
The valuation depends on cash recovery.
Base scenario · annual USD billions · all lines are forecasts
Discount those cash flows at 9%, assume 3.5% perpetual growth supported by a 20% terminal return on invested capital, then account for cash, debt, leases and diluted shares: the result is $427. Terminal growth is charged for the reinvestment it requires.
04 / The market is paying for a better outcome
Our base value is 13.6% below the $493.95 close. The downside case gives $265, with slower cloud growth, a 42% mature operating margin and a 15% terminal capital return. The upside case gives $605, with faster growth, a 47% margin and a 25% terminal capital return. None is a price floor or ceiling.
The current price does not require an impossible outcome. Keeping other base assumptions fixed, adding about 3 percentage points to Intelligent Cloud growth in every forecast year brings value to $494. That means roughly 31% initially and 9% in year ten, rather than 28% and 6%. Alternatively, a discount rate of about 8.36% supports the price. This is an implied model discount rate, not a forecast of shareholder returns.
That is the real disagreement: we prefer a higher required return and more room for the efficiency assumptions to disappoint. At an 8% discount rate, the same base operating path is worth $539; at 10%, $349. Readers accepting the lower hurdle can reasonably reach a different conclusion.
05 / The proposal: wait for value, verify the thesis
For a new long-term position, our proposal is to keep MSFT on the watchlist around $494 and consider an entry at or below $340 if the cash-recovery thesis still holds. The threshold applies a subjective 20% margin of safety to base value; it does not come from a backtested trading rule. Even $340 remains above the $265 downside scenario.
Our confidence is moderate in the business case and low to moderate in the valuation. About 71% of base enterprise value comes from cash flows beyond year ten. That dependence, the lease transition and the uncertain replacement cycle prevent a high-confidence price target.
The next full-year evidence should be tested against our roughly 17% total revenue growth, 28% Intelligent Cloud growth, 46% operating margin and $194B investment budget. Persistent weaker growth or greater cash investment/rent requires a lower valuation—not automatically buying because the price fell. Conversely, sustained cloud growth nearer the reverse-DCF path, with controlled reinvestment, would justify raising value. Reassess after the next results and any material change in lease or capital commitments.
The conclusion is concrete: we like the earnings franchise, but require a better entry price to underwrite the cash-flow uncertainty. Intrinsic value changes as evidence changes; the market has no obligation to converge to today's estimate.
Model, sensitivity, sources & limitations
Reproduce and challenge the result
Download the marimo notebook · Reported inputs and all annual assumptions · Annual outputs and sensitivity · Independent numerical check
Run uvx marimo edit microsoft-valuation.py --sandbox to explore the discount-rate, terminal-growth and cloud-growth controls. The notebook is self-contained and writes its inputs, results and charts beside itself. For an independent check, put the two JSON downloads beside the verification script and run python3 verify-microsoft-valuation.py. No account or paid data is required.
Cash-flow and equity bridge
All amounts are USD billions unless per share. FCFF = revenue × operating margin × (1 − tax) + depreciation/amortization − capital expenditure − incremental working capital. Working-capital investment is 5% of incremental revenue. Operating margins include stock compensation; there is no SBC add-back or assumed reduction in shares from future buybacks.
The base present value of ten years' cash flows is $947B and terminal value is $2,290B, giving $3,237B enterprise value. Add $76.843B cash and short-term investments, retain $20B operating liquidity, subtract $46.136B face-value debt and $66.594B finance lease liabilities. Divide the resulting $3,181B equity value by 7.453B FY26 weighted diluted shares. The liquidity reserve and diluted-share proxy are modeling choices; debt face value is conservative relative to the filing's $36.5B estimated market value.
Finance-lease principal is not subtracted again from FCFF. Operating lease liabilities are not separately deducted because operating rent remains in margins. The $329.1B of not-yet-commenced leases must be serviced through projected investment or operating costs; deducting the gross commitment again would double count it. We do not have a contract-level lease forecast. The $26.7B of PP&E purchases in accounts payable must be settled within forward cash capex budgets. A spending path that cannot accommodate these obligations must be revised.
No extra value is added for OpenAI or other strategic investments; related non-operating gains do not enter EBIT. Adding the entire $36.348B equity-and-other-investments balance at book would add $4.88/share before tax and overlap, not close the base valuation gap. Book value is not a realizable sale estimate. No future acquisition program is modeled, so growth requiring acquisitions or new strategic investments would need additional cash deductions.
Terminal value and sensitivity
Terminal cash flow = year-ten after-tax operating profit × (1 + growth) × (1 − growth / terminal ROIC). Divide by discount rate minus growth, then discount ten years. The base terminal cash flow is $298B. The 9% nominal USD discount rate is our selected enterprise valuation hurdle, not an empirically estimated cost of capital. The scenarios share that rate and 3.5% growth; terminal ROIC is 15%, 20% and 25% respectively.
| Base value per share | 2.5% perpetual growth | 3.5% | 4.5% |
|---|---|---|---|
| 8% discount rate | $482 | $539 | $630 |
| 9% discount rate | $393 | $427 | $476 |
| 10% discount rate | $328 | $349 | $378 |
These are assumption sensitivities, not probabilities. The model uses ten forward annual periods and year-end cash flows, with FY26 revenue and the June balance sheet as its latest baseline. It does not separately model the fiscal/calendar stub, intervening balance-sheet movements, a detailed asset replacement schedule or each lease contract. These limitations matter more than cents in the output.
Primary-source register
- FY26 release, 29 July 2026: income, segment results, cash flows and diluted shares; embedded Refinitiv quote timestamped 8 September, 16:00 ET, observed 9 September.
- FY26 Form 10-K: financial statements; Note 6 for PP&E payables and depreciation; Note 10 for debt; Note 13 for leases; revenue-recognition note for RPO. Period ended 30 June; filed 29 July; checked 9 September.
- 29 July earnings call: RPO excluding OpenAI, FY27 revenue/margin/tax/capex guidance and the lease-classification change. Management expectations are attributed, not independently validated demand forecasts.
Research was produced with AI assistance; the reproducible calculations were separately checked. This is general investment research, not advice tailored to a reader's circumstances. The earlier cash-flow note remains in the revision record; its original publication date and this article's URL are preserved.