Microsoft Just Spent $64 Billion on Capex. The Returns Justify Every Dollar

We have heard the bear argument all year. Microsoft is overspending on data centers. Hyperscale capacity is outrunning demand. The depreciation cliff is coming.

The data does not support it. Revenue grew 18.3% on a quarterly basis at the most recent print, the strongest growth rate since calendar 2022. The operating margin sits at 46.3% trailing twelve months, the highest in the company's modern history. Cloud and AI services are the marginal contributor and they are running at structurally higher gross margins than the franchise average.

Capital allocation in tech has historically been judged on three things: incremental ROIC, time-to-revenue lag, and the optionality of the asset base. On all three, the Microsoft data set tells a constructive story.

Incremental ROIC. The four-year operating income build of $58.6 billion against four-year cumulative capex of $181.6 billion implies roughly 32% incremental return on capital. That is well above the company's stated 18% long-term ROIC target.

Time-to-revenue lag. The data center build cycle for AI workloads is roughly 12 to 18 months from groundbreaking to revenue. That puts the FY24-25 capex bulge in revenue contribution range for FY26-27. Bears argue this is too long; the historical AWS build cycles ran 24 to 36 months and the market accepted that lag for a decade.

Optionality. AI capex is the most fungible IT asset Microsoft has ever deployed. Training, inference, traditional compute, and Azure-native workloads all run on the same fleet. The depreciation cliff argument requires assuming the assets cannot be repurposed; they can.

If the AI infrastructure overspend argument were correct, we would already be seeing it in the numbers. Operating margin would be compressing. It is expanding. Cloud gross margin would be deteriorating. Public Azure margin commentary suggests it is holding. Free cash flow would be falling sharply. It dipped 3% on a 45% capex jump, which is the opposite of an FCF crisis.

The historical parallel that bears keep pointing to is 2010-2014 hyperscale capex at AWS, which compressed Amazon's reported margins for several years. That comparison breaks down on the gross margin profile. Microsoft is starting from a 70%-plus gross margin franchise. AWS started from a roughly 20% gross margin retail business. The capex absorption capacity is fundamentally different.

The Bloomberg report from earlier this week on payment delays at one African data center site was an operational issue, not a strategic one. We note it but do not view it as a thesis-mover. Cap allocation in the AI build is a global program; one site running into local payment friction does not change the four-year math.

At $415 per share and FY26 consensus EPS of $19.40, Microsoft trades at 21.4x forward earnings. The five-year average forward multiple sits at 30.2x. Compression of nearly nine multiple turns has already occurred relative to the post-2020 peak.

Enterprise value of $3.06 trillion against TTM revenue of $282 billion implies 9.6x EV/Revenue. EV/EBITDA at 15.4x is a five-year low for the company. The PEG ratio at 1.29 implies fair valuation against an expected mid-teens earnings growth rate.

The analyst rating composite of 4.59 out of 5 is the strongest in the megacap technology cohort. Of 56 analysts polled, 41 rate the stock strong buy. The Wall Street target price of $562 implies 35% upside.

Dividend yield is just 0.85%, but the dividend is not the point. The point is the operating income compounding, which has averaged 16% over the past four years through one of the most aggressive capex cycles in corporate history. That is the capital allocation thesis.