Ordering a company valuation also supports the project’s development.
Help develop World Market Atlas — an independent project for individual investors. Your support helps us produce analyses and improve our tools.
FY2026 reported revenue, including revenue-sharing; the rest of revenue can also be affected by AI competition. USD billions.
Reported fiscal years ending 30 June. Cash FCF = operating cash flow minus cash property and equipment spending. Finance-leased additions are excluded here, but included in the DCF capex assumptions.
| Metric | FY2025 | FY2026 |
|---|---|---|
| Operating cash flow | 136.162 | 182.935 |
| Cash capex | 64.551 | 115.948 |
| Cash FCF | 71.611 | 66.987 |
Author DCF scenarios, valued at 28 October 2026. Same WACC in the first three cases; stress includes an extra 0.5 percentage point. Bars start at zero. The reference is the 27 October closing price, not a forecast.
Reference price: 531.33 USD per share
| Scenario | Equity value · USD tn | USD per share | Versus our baseline | Versus reference price |
|---|---|---|---|---|
| Baseline | 3.23 | 432.75 | 0.0% | -18.6% |
| Adaptation | 3.02 | 405.03 | -6.4% | -23.8% |
| Central | 2.66 | 357.52 | -17.4% | -32.7% |
| Stress | 1.78 | 239.31 | -44.7% | -55.0% |
Author assumptions. Change the case, WACC and terminal growth; all values update together. Defaults match the article. No scenario probabilities are assigned.
Baseline: 432.75 USD · WACC: 9.0% · Terminal growth: 3.0%
Versus reference price: -32.7% · Terminal share of operating value: 68.2%
Full fiscal year FCFF, USD billions. Only the remaining fraction of FY2027 enters present value; later years are full years.
| PV of forecast cash flows | 843.66 |
|---|---|
| PV of terminal value | 1,809.86 |
| Operating value (EV) | 2,653.51 |
| Cash and short-term investments | 76.84 |
| Debt | -40.29 |
| Finance lease liabilities | -66.59 |
| OpenAI stake proxy | 42.60 |
| Receivables loss | -1.50 |
| Equity value | 2,664.57 |
USD billions unless a percentage is shown. Capex includes newly finance-leased assets. First two years keep baseline dollar capex and D&A.
| Metric | FY2027 | FY2028 | FY2029 | FY2030 | FY2031 | FY2032 | FY2033 | FY2034 | FY2035 | FY2036 |
|---|---|---|---|---|---|---|---|---|---|---|
| Revenue | 375.2 | 418.2 | 457.7 | 501.4 | 545.0 | 586.9 | 633.8 | 678.2 | 718.9 | 754.8 |
| Revenue-share proxy | 0.7 | 0.8 | 0.8 | 0.9 | 0.5 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |
| Core EBIT margin | 44.5% | 44.5% | 44.0% | 44.0% | 44.0% | 44.0% | 44.0% | 44.0% | 44.0% | 44.0% |
| EBIT | 167.4 | 186.5 | 201.8 | 221.1 | 240.0 | 258.2 | 278.9 | 298.4 | 316.3 | 332.1 |
| Depreciation | 53.9 | 70.8 | 82.4 | 95.3 | 106.3 | 114.4 | 120.4 | 122.1 | 122.2 | 120.8 |
| Capex incl. new finance leases | 173.2 | 177.0 | 157.9 | 150.4 | 147.1 | 146.7 | 145.8 | 145.8 | 143.8 | 143.4 |
| Incremental working capital | 0.9 | 0.9 | 0.8 | 0.9 | 0.9 | 0.8 | 0.9 | 0.9 | 0.8 | 0.7 |
| Full year FCFF | 13.7 | 42.2 | 85.2 | 120.9 | 150.3 | 173.5 | 196.8 | 214.1 | 230.7 | 242.3 |
| Core revenue gap vs baseline | 2.0% | 5.0% | 8.0% | 10.0% | 11.0% | 12.0% | 12.0% | 12.0% | 12.0% | 12.0% |
| Cash-flow fraction | 0.671 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 |
| Years to payment | 0.671 | 1.671 | 2.671 | 3.671 | 4.671 | 5.671 | 6.671 | 7.671 | 8.671 | 9.671 |
| Present value | 8.7 | 36.5 | 67.6 | 88.1 | 100.5 | 106.4 | 110.8 | 110.5 | 109.3 | 105.3 |
USD per share. Only WACC and terminal growth change; other central assumptions stay fixed. This is not a confidence interval.
| WACC / Terminal growth | 2.5% | 3.0% | 3.5% |
|---|---|---|---|
| 8.5% | 371 | 395 | 424 |
| 9.0% | 338 | 358 | 381 |
| 9.5% | 310 | 326 | 345 |
Source financials: Microsoft FY2026. Forecasts, stake haircuts and valuation: author calculations. Microsoft FY2026 ↗
Market commentary
My perspective on markets, the economy and events that matter to investors.
What would Microsoft be worth if OpenAI lost to Anthropic?
28 October 2026 · Analysis before FY2027 Q1 results. Financials: FY2026 and 30 June 2026 balance sheet. Reference price: 27 October 2026 close. DCF scenarios are author assumptions.
If OpenAI permanently ceded commercial AI leadership to Anthropic, Microsoft could still be worth about $2.7 trillion — $358 per share in our central case. Successful adaptation gives roughly $405; deeper stress gives $239. These are assumption-dependent outputs of an original DCF, not stock-price forecasts or a claim that the race is settled.
The essential comparison is with our own $433-per-share baseline. The central case reduces that value by 17.4%, or about $561 billion of equity. Against MSFT’s $531.33 closing price on 27 October, the gap is 32.7%. Part of that gap already exists in the baseline; it cannot all be attributed to OpenAI losing. Twelve Data · 27 October 2026 close
1. What does losing mean?
Here, it means Anthropic sustains an advantage for several years in attracting and retaining paying enterprise customers and developers. A single benchmark result is insufficient. OpenAI could remain a large business while losing leadership.
We analyse a conditional outcome without assigning a probability. Google, open models and Microsoft’s own technology also matter. The central case does not assume OpenAI goes bankrupt.
2. OpenAI exposure is only part of Microsoft’s value
The $24.1bn is revenue, not profit. Lost demand affects margins and infrastructure utilisation but can also remove some costs. The report does not split compute sales from partnership payments. The other 92.74% of revenue is not automatically immune to AI: competition can affect Microsoft products and pricing too. Microsoft · 2026 annual report, statements and Note 1
3. Why some Claude upside can stay with Microsoft
Under the 18 November 2025 partnership, Anthropic committed to buy $30bn of Azure capacity, while Microsoft committed to invest up to $5bn. These are total commitments with no disclosed annual revenue schedule. We do not add $30bn to revenue every year. Microsoft · Anthropic and NVIDIA partnership
Claude became generally available in Microsoft Foundry on 29 June 2026, with inference hosted in Azure. A customer switching to Claude can therefore preserve some Microsoft revenue, although not necessarily the same margin. This is a compensation mechanism, not a measurement of its size. Microsoft Azure · Claude in Foundry
The current OpenAI contract also matters. Following the 27 April 2026 amendment, Microsoft’s model and product IP licence is non-exclusive through 2032. OpenAI can serve customers across any cloud. Revenue-sharing payments to Microsoft continue through 2030, subject to an undisclosed total cap. Azure exclusivity cannot underpin today’s valuation. Microsoft · current partnership terms
4. Capital intensity may matter more than model rankings
FY2026 operating cash flow was $182.9bn, against $115.9bn of cash property and equipment spending. The difference was $67.0bn, versus $71.6bn a year earlier. Revenue growth did not produce growth in that definition of free cash flow. The chart shows this simple cash measure, not the FCFF used in valuation. Microsoft · FY2026 cash flow statement
Cash spending is not the whole investment bill: Microsoft also obtains assets through finance leases. Management’s FY2027 outlook calls for higher capex and an approximately 20% effective tax rate. Reclassification between finance and operating leases changes reported capex; it does not remove datacentres’ economic cost. Microsoft · 29 July 2026 earnings call
Our interpretation: the biggest damage would arise if lost demand met long-lived infrastructure commitments. Modelled capex and depreciation therefore stay at baseline dollar levels for the first two years. They do not disappear with revenue.
5. How the valuation works
We value all of Microsoft using FCFF = EBIT × (1 − tax) + depreciation − capex − incremental working capital. Forecasts cover FY2027–FY2036. The first cash flow is prorated from 28 October 2026 to 30 June 2027; subsequent flows are discounted at fiscal year ends. We ignore seasonality within the first partial year.
Baseline core revenue growth fades from 16% to 5%, with operating margin settling at 46%. Capex, including new finance-leased assets, falls from 45% to 20% of revenue; depreciation ends at 16%; incremental working capital absorbs 2% of incremental revenue. These are author assumptions, not analyst consensus or Microsoft guidance. The model table exposes the complete annual path.
WACC is 9.0%, with 3.0% nominal USD terminal growth after 2036. The starting reference is the 4.99% US ten-year Treasury yield on 27 October 2026. Our assumed 0.90 beta, 4.5% equity risk premium, 5.5% debt cost and 2.6% debt weight produce approximately 8.9%, rounded to 9.0%. Except for the Treasury yield, these are assumptions rather than market observations. U.S. Treasury · interest rate data
The actual revenue-sharing component of $24.1bn is unknown. We use $2.4bn in FY2026 as a proxy, subtract it from the core revenue base and forecast it separately: $2.8 / 3.0 / 3.2 / 3.4 / 1.8 / 0bn in FY2027–FY2032. The last partial amount approximates expiry in calendar 2030. Timing, scale and a 100% EBIT margin on these fees are assumptions; no royalties enter terminal value. Changing the starting proxy from $0bn to $5bn moves the central result only from about $360 to $355.
6. Four valuations, three forms of losing
Adaptation: core revenue ultimately falls 4% below baseline, margin is 0.5 percentage points lower and the OpenAI stake retains half its modelled value. Microsoft moves many customers to alternative models and moderates future investment.
Central: the revenue gap reaches 12%, margin falls 2 percentage points and the OpenAI stake retains 25% of modelled value. We also assume a $1.5bn receivables loss. Revenue pressure extends beyond the OpenAI contract to Azure and applications’ competitive position. Investment can eventually fall; its final share of sales is 1 percentage point below baseline.
Stress: the revenue gap reaches 25%, margin is 6 percentage points lower, the OpenAI stake is worth zero and all $6bn of receivables is lost. WACC rises to 9.5%. This combines wider disappointment in AI returns with infrastructure commitments; it is not an automatic consequence of an Anthropic model being better. Every scenario parameter is an author assumption.
7. Where does the $561bn loss come from?
About $431bn of the central case’s reduction comes from lower operating cash-flow value, $128bn from the OpenAI stake and $1.5bn from receivables. We apply no additional discretionary “AI discount” to cash flows already reduced.
The opening stake proxy is 25% × $852bn × 80% = $170.4bn. The $852bn figure is the confirmed post-money valuation from the round announced on 31 March 2026. A 20% haircut approximates illiquidity, tax, preferences and uncertain dilution. This is neither a formal security valuation nor its carrying value. OpenAI · completed funding round
We add cash and short-term investments to operating value and subtract debt and finance lease liabilities. Other equity investments, including Anthropic, are not separately added because sufficiently detailed public allocations are unavailable. Any appreciation there is an omitted benefit. Stock compensation stays an EBIT expense and is not added back. Per-share values use a constant 7.453bn FY2026 diluted shares. The balance sheet remains the latest reported one at 30 June, without pretending it has been updated to October.
8. The greatest uncertainty lies in later years
About 68% of central operating value comes from terminal value. Varying WACC between 8.5–9.5% and long-term growth between 2.5–3.5% gives approximately $310–424 per share for the same scenario. This is sensitivity analysis, not a statistical confidence interval.
The model assumes substantial investment efficiency. Central terminal parameters imply roughly 34.5% returns on reinvested capital, a demanding outcome. Adding 2 percentage points of revenue to terminal capex reduces value from $358 to about $342. Investment cannot be cut indefinitely while preserving growth.
With the baseline forecasts unchanged, $531.33 would require approximately 8.0% WACC. That reverse DCF is not evidence that the market actually uses that discount rate. Higher growth or more durable profitability could also support a higher price. We do not claim access to analyst consensus.
9. What would challenge the scenario?
Milder damage would be supported by customers staying in Azure after switching models, stable application margins and lower investment without lost growth. Deeper damage would be supported by customers leaving Microsoft’s ecosystem, persistently high investment against weaker sales and deteriorating OpenAI collections. These business outcomes matter more than one model ranking.
Watch subsequent Azure and Copilot disclosures, margins, OpenAI receivables and lease commitments. A new funding round can raise the reference price for equity interests without proving improved cash generation.
Conclusion: OpenAI losing leadership could cost Microsoft hundreds of billions of dollars, but the scale depends mainly on customer retention and infrastructure utilisation. Our central configuration gives about $2.7tn of equity value. The estimated cost of the scenario change itself is 17.4% of our baseline valuation.
Data and limitations: analysis frozen before FY2027 first-quarter results, scheduled after the market close on 28 October 2026. Those results are not included. Facts come from company reports and announcements, U.S. Treasury and Twelve Data prices; scenarios and the DCF are author calculations. Full revenue-sharing terms, rights of every share class and future datacentre utilisation are unknown. This is conditional economic analysis, not a buy or sell recommendation. Microsoft · next earnings date
Posts reflect the author’s personal views as of publication. They are not investment recommendations.