Why was Microsoft stock falling?

It fell for most of 2026 because the spending arrived now and the revenue arrived later. Then fourth-quarter results showed Azure past $100bn and the shares rallied about 25% in three days. Here is what caused the fall, what the rebound settles, and what it doesn't.

By Nathan Wickham-Hurd · Founder, Oak Growth · First-class Economics & Finance, MBA · Last reviewed August 2026

The short answer

Microsoft fell for most of 2026 because it was spending enormous sums on AI infrastructure while the revenue from that spending arrived later — free cash flow dropped sharply, Copilot adoption stayed in low single digits, and investors who had priced an immediate payoff repriced for a multi-year cycle. That phase appears to have turned. Fourth-quarter results at the end of July showed Azure growing 43% and passing $100bn of annual revenue for the first time, and the shares rallied roughly 25% in three days.

Context, August 2026. Microsoft spent most of 2026 as the worst performer among the large US technology companies, falling more than 20% from its June peak after a first quarter decline of about 25% — its steepest since 2008. It then reported fiscal 2026 fourth-quarter results: revenue of $90bn, up 18%, with Azure up 43% and annual Azure revenue above $100bn for the first time. The shares jumped around 25% in three sessions and now sit roughly 26% above their June low, though still down on the year. Market conditions change; the framework below is meant to outlast them.

Why it fell — the four reasons

1. Capital spending was enormous and immediate

Data centres, power and chips all get paid for now. Guidance rising toward $190bn against market expectations nearer $155bn was the single biggest shock, and it pushed free cash flow down by over 20%. A discounted cash flow values free cash flow, so a company converting a fifth less of its profit into cash was genuinely worth less while that spending ran — even with revenue growing.

2. Copilot hadn't converted

Paid penetration was reported at roughly 3–4% of more than 450 million commercial Microsoft 365 seats. The bull case assumed a large share of that base upgrading. At that level it looked closer to a pilot programme than a revenue line, and investors ran out of patience waiting.

3. Concentration in one AI customer

The contracted backlog is very large — reported around $627bn, more than two years of revenue — but a substantial portion is tied to OpenAI, a customer actively diversifying its cloud arrangements. That remains a structural dependency rather than a solved problem.

4. Regulatory and legal noise

An FTC investigation into cloud, AI and bundling practices, plus securities class actions covering a defined purchase period and alleging misleading statements about Copilot adoption and Azure AI performance. No judgement has been made, but the headlines and costs persist.

What changed

The fourth-quarter numbers answered the specific question the market was asking. The worry was never whether Microsoft could sell — it was whether the capital spending would ever show up as revenue. Azure growing 43% and crossing $100bn of annual revenue is that spending showing up. Improved cloud margins mattered as much as the growth rate, because margins are what turn revenue into the free cash flow the valuation actually discounts.

Worth being precise about what this does and doesn't settle. It shows the infrastructure build converting to revenue — a real answer to the central bear argument. It does not resolve the OpenAI concentration, the regulatory position, or whether Copilot broadens beyond early adopters. One good quarter changes the evidence, not the whole thesis.

What never changed

This is the part the price action obscured all year. Enterprise switching costs are as high as ever — moving an organisation off Active Directory, Exchange and Office means retraining everyone and rewriting integrations. Azure demand had been constrained by supply, not by lack of customers: data centre capacity, power availability and equipment lead times. A company that cannot build fast enough is in a very different position from one that cannot sell.

Full-year revenue rose about 18% to roughly $332bn with earnings up over 31%. The business was growing throughout the decline. What fell was the market's patience with the timing of cash flows.

The lesson worth taking from it

A fall driven by the timing of cash flows is a different thing from a fall driven by deterioration of a business. The first can create opportunity; the second usually deserves the price. Microsoft in 2026 was the first kind, and the way you could tell at the time — before the rebound — was that revenue kept growing, switching costs were intact, and the constraint on Azure was capacity rather than demand. See the four reasons markets fall.

What to check from here

Capital spending as a share of operating cash flow

The question is when this peaks. Track the ratio each quarter rather than the absolute number.

Azure growth and margin together

Growth alone can be bought. Growth with expanding margin is what converts into free cash flow.

Copilot paid seats, not availability

Penetration of the commercial base is the number. Feature announcements are not.

Backlog concentration

How much of that very large backlog still depends on a single customer.

How to think about the valuation now

Value Microsoft on free cash flow rather than reported earnings, because the gap between them is exactly what was being argued about. Be explicit about when you assume capital spending normalises — that single input moves the answer more than anything else. And note that a 25% move in three days changes the price without changing the business, which cuts both ways: the multiple that looked like a decade low before results does not look like one after a rally of that size.

Value it on cash, not headlines

Oak Growth publishes a discounted cash flow intrinsic value and margin of safety for Microsoft and roughly 1,000 other companies across eight markets.

Explore Oak Growth

Common questions

Why was Microsoft stock falling in 2026?

Because it was spending enormously on AI infrastructure while the revenue arrived later. Capital expenditure guidance rose far above expectations, pushing free cash flow down more than 20%, while paid Copilot adoption stayed at low single-digit percentages of the Microsoft 365 base. Investors who had priced an immediate AI payoff repriced for a multi-year cycle.

Why did Microsoft stock go back up?

Fiscal 2026 fourth-quarter results showed Azure growing 43% and passing $100bn of annual revenue for the first time, with improved cloud margins. That directly answered the central bear argument — whether the capital spending would convert into revenue — and the shares rallied roughly 25% in three sessions.

Is Microsoft stock cheap now?

Its forward multiple compressed to around a decade low during the decline, but a rally of roughly 25% in three days changes the price without changing the business. Whether it is cheap now depends on when capital spending peaks and how much free cash flow recovers, not on the multiple before the rebound.

What is the risk with Microsoft and OpenAI?

A substantial share of Microsoft's very large contracted backlog is tied to OpenAI, a customer actively diversifying its cloud providers. Strong quarterly results do not resolve that concentration — it is a structural dependency to be worked through over several years via contract renewals and broader enterprise demand.

Also see: Microsoft (MSFT) full analysis → · Free cash flow explained → · Why is the stock market falling? →

Also see: Bull market vs bear market →