Microsoft shares have fallen 22% over the past twelve months and sit more than 15% below their June high, even as the underlying business kept accelerating. The stock isn’t being punished for what it earned. It’s being punished for what it spent to earn it.

The bull case. Azure grew 40% in constant currency last quarter, the fifth straight quarter of acceleration, and management has repeatedly attributed shortfalls to capacity constraints rather than soft demand, a better problem to have if the supply eventually arrives. Intelligent Cloud revenue rose 30% to $34.7 billion, net income rose 23%, and the AI business now runs at a $37 billion annual pace. Return on invested capital sits at 27.2% against a 10.1% cost of capital, a wide spread that has held for two decades. At 21.3 times forward earnings against 18% expected earnings growth, the stock isn’t demanding much from the future to work.

The bear case. The reason for the discount is capital expenditure. Calendar-2026 capex guidance jumped to roughly $190 billion, about $35 billion above what analysts expected, with roughly $25 billion of that increase coming from component price inflation rather than added capacity. Free cash flow fell to $15.8 billion from $20.3 billion a year earlier even as operating cash flow grew. Depreciation has barely started landing on the income statement, and trailing capex of $97.2 billion against depreciation of just $35.5 billion means that expense rises for years. Copilot, meanwhile, has passed 20 million seats but converts only about 3.3% of the installed base to paid use, and still trails ChatGPT in consumer surveys.

Both stories are true at once, which is why the call here is Buy rather than Strong Buy. A probability-weighted fair value near $449 implies roughly 14% upside, with outcomes ranging from $560 if Azure sustains its growth and spending plateaus, to $300 if a third capex surprise arrives and depreciation compresses margins faster than revenue can offset it. On enterprise value to EBITDA, the shares trade near 16 times against a five-year average closer to 21 times, suggesting the market has already priced in a meaningfully worse capital-efficiency outcome than history would support. Microsoft reports results on July 29, and the number that matters most isn’t Azure growth or earnings, it’s whether 2027 capital spending guidance holds steady or surprises upward again.

A staged entry fits better than a single purchase, with the balance added after the print or on a retest of the low-$350s. Trailing free cash flow near 40 times remains the strongest argument against owning the stock today.

The full valuation model, risk register, and staged trade plan are available in the complete Microsoft (MSFT) report on our Reports page.