Microsoft and Capex: How One DCF Assumption Drives the Whole Valuation

In December, an earlier Stockoscope analysis looked at Microsoft's valuation through the lens of its capital spending, a concern highlighted by Michael Burry. Holding Microsoft's then-elevated capex intensity through the forecast, the model produced a P/E compression scenario of $412-427 and a DCF bear case of $311, against a market price near $490. Here is what that analysis set out at the time:

"Currently, Microsoft trades at approximately $492 per share with a P/E ratio of 34 and earnings per share of $14.11. If earnings were adjusted down by 11-14% to reflect realistic GPU depreciation, the adjusted EPS would fall to $12.13-$12.56. Assuming the P/E ratio remains at 34, the stock price would drop to $412-427, a decline of $65-80 per share, or roughly 13-16%. However, if investors also lose confidence in AI infrastructure returns, the P/E multiple could compress further, potentially amplifying losses beyond the accounting adjustment alone."

"We recalculated our DCF model assuming capex remains at 20% of revenue instead of declining to 15.7%. The additional $15-20 billion in annual capital expenditure reduces free cash flow each year, which lowers the total enterprise value by $547 billion and the per-share intrinsic value by $73.44. Thus, the Burry-adjusted fair value becomes $311.49 per share."

Six months later, Microsoft trades at around $373, after touching as low as $349 on 25 June 2026. Over the same period, Microsoft and its hyperscaler peers reported another step-up in AI capital spending. In the same window, we have also thought hard about how a DCF should treat capex, and refined our model. Both make this a useful moment to walk through how a single assumption, long-run capital intensity, drives the entire valuation, using Microsoft as the worked example.

Microsoft's capex situation

The capital-spending trend that the December analysis focused on has only intensified, and the cash-flow cost is now plain to see:

Table 1. Microsoft capital spending and cash flow, FY2020-FY2025.

Fiscal year	Capex / revenue	Operating cash flow	Free cash flow
FY2020	10.8%	$60.7B	$45.2B
FY2021	12.3%	$76.7B	$56.1B
FY2022	12.0%	$89.0B	$65.1B
FY2023	13.3%	$87.6B	$59.5B
FY2024	18.1%	$118.5B	$74.1B
FY2025	22.9%	$136.2B	$71.6B

Look at the last two columns together. Operating cash flow has more than doubled since FY2020, to $136 billion. But free cash flow has stalled: FY2025 free cash flow ($71.6B) actually came in below FY2024 ($74.1B), even though operating cash flow grew 15%. The entire difference is capex, which has roughly doubled as a share of revenue, from about 11% in FY2020 to 23% in FY2025.

The trend is industry-wide. Across 2026, the four big hyperscalers (Microsoft, Google, Amazon and Meta) have collectively guided to roughly $725 billion of capital spending, up about 77% year over year. This suggests that capital spending has not slowed; it has accelerated.

How we have changed our model

When we published in December, our model treated capex the conventional way: it effectively assumed the company's current capital intensity would persist across the entire forecast horizon and into the terminal value. With hyperscalers now spending so heavily on AI, we went back and questioned that assumption.

Assuming capex stays elevated forever is not realistic. Yes, Microsoft is spending heavily today, but once the build-out is done, a company growing only a few percent a year does not need to keep reinvesting a fifth of its revenue. In a genuine steady state, capital spending converges toward "maintenance" capex, roughly equal to depreciation, which is what replacement of worn-out assets actually costs.

So we changed two things:

1. In the explicit forecast years, capex now glides from the company's recent (elevated) rate down toward maintenance, rather than sitting flat at today's level. For Microsoft that means capex falling from about 20% of revenue today to roughly 10% (maintenance) over the forecast, instead of pretending the AI build-out runs at full intensity forever.
2. In the terminal value, capex is set to maintenance (approximately depreciation), the steady-state level a perpetuity can actually sustain.

This is a more faithful picture than holding a single number to the end of time.

The important consequence: gliding capex down lifts the modeled value, because the model no longer assumes the elevated AI spend is permanent.

What the model says now

Holding everything else constant and flexing only the long-run capex assumption, here is Microsoft's modeled value across the realistic range (price $373):

Table 2. Microsoft intrinsic value under different long-run capex assumptions.

What you assume long-run capex does	Fair value	vs price
Glides down to maintenance (our standard model)	$487	+31%
Glides from the FY2026 30% pace down to maintenance	$464	+24%
Stays elevated permanently (20% of revenue)	$364	-2%
Stays at the FY2026 pace (30%) permanently	$228	-39%

Note: Microsoft's modelled capex path, as a share of revenue: 20%, 19%, 18%, 17%, 16%, 15%, 14%, 13%, 12%, 11%, 10%, then 10% (maintenance) in perpetuity.

Two things stand out. First, the near-term assumption barely matters: gliding from the 30% pace ($464) lands within about $23 of the standard model ($487), because most of a DCF's value sits in the terminal. Second, the terminal assumption is the entire ballgame. Assume the AI build-out normalizes and the model estimate is $487, about 31% above the current price. Assume capital intensity stays permanently elevated and the estimate falls to roughly $364, about where the stock trades today; at the full FY2026 pace it drops to $228.

Because that one assumption matters so much, we have added a control to Microsoft's DCF page, and every other company's: a "Hold Capex Elevated" switch. Off (the default), capex glides to maintenance. On, it stays at the recent rate across the whole forecast and the terminal. You can flip it and watch the modeled value move between roughly $487 and $364 in real time.

Where Microsoft stands now

A few things are measurable rather than matters of opinion.

Since December, Microsoft has fallen about 31% from its October peak of $542 (and as much as 35% at its June low of $349). The move came without an earnings miss; earnings actually rose. What changed was the multiple: the forward P/E compressed from about 34 to roughly 22. In other words, the market re-rated the stock for its capital intensity rather than for any deterioration in profits, the same accounting-versus-cash-flow gap the December analysis walked through.

On the refined standard model, the estimate is $487, roughly 31% above the current price, because that model assumes today's elevated capex normalizes over time. But that is an assumption, not a fact, and it is the one assumption the whole valuation hinges on. Assume Microsoft must keep spending at its current pace indefinitely and the estimate sits around $364, essentially where the market has already put it. The market, in other words, has largely re-priced Microsoft for elevated capex.

Put the range together: Microsoft currently trades close to the estimate the model produces when capex is held elevated (around $364), so on that assumption there is little gap between price and estimate. On the assumption that capex normalizes, the estimate is about $487, a gap of roughly 31%. And if the market were to assume AI capex runs at 30% of revenue in perpetuity, the estimate would fall to around $228. The point is not which of these is correct; it is that the entire range, from $228 to $487, turns on a single assumption about long-run capital intensity, and the market has moved from pricing near the optimistic end of that range to pricing much closer to the elevated-capex end.

This is not a recommendation to buy or sell. The takeaway is methodological: in a DCF for a heavy reinvestor, the long-run capex assumption can be worth more than everything else combined, so it is worth setting deliberately rather than accepting a default. Flip the switch, choose the assumption you can defend, and see where the estimate lands. Microsoft is a candidate for your own research, not a recommendation.

---

All data, prices, and results reflect information available at the time of writing and will move as markets and filings update; the live model on the platform always recomputes from current data. This article is for educational purposes only and is not investment advice. The author(s) and Stockoscope may hold positions in the securities mentioned. Always do your own research.