In June, Microsoft shares fell 21.6% — the worst monthly result since 2008. At one point, the decline reached 20% and threatened to become the worst month since December 2000. A partial recovery at the end of June did not change the overall picture: the company lost over $530 billion in market capitalization.
The official reason is investors' doubts about the return on investment in artificial intelligence. But if you break down the figures, a more specific risk emerges, which analysts discuss in every other report.
Spending more, earning less free cash
In the last quarter, Microsoft's capital expenditures increased by 63% year-over-year — to $38 billion. A direct consequence was a 10% reduction in free cash flow. According to analysts, in fiscal year 2026, the company may spend over $120 billion on infrastructure.
Azure, meanwhile, is growing: revenue increased 39–40% year-over-year, with the AI direction generating $37 billion annually at a growth rate of 123%. Microsoft's total revenue increased 18%. In other words, the business is not broken — the market is repricing not results, but timelines. As Investing.com explains, 2024–2025 investors had priced in immediate and clean AI monetization, but are now shifting to a 3–5 year horizon.
45% of future income — one customer leaving
The structure of Azure's order portfolio reveals concentrated risk. The total Remaining Performance Obligations (RPO) — contracted but not yet recognized revenue — reached $625 billion. 45% of this amount is tied to OpenAI as a single customer.
"Microsoft and Meta are behaving as if they put on winter jackets on the beach in summer"
Dan Ives, analyst at Wedbush Securities, quoted by Wall Street Journal
The problem is not only concentration. According to Investing.com, OpenAI is actively diversifying its cloud footprint — and moving away from Azure. Moreover, Microsoft not only finances OpenAI through its 27% stake: the startup's losses pass directly through Microsoft's financial statements as GAAP expenses. In the first quarter of FY2026, these losses were $3.1 billion — compared to $523 million a year ago.
A separate signal comes from Copilot data. Microsoft 365 Copilot's share among paid subscribers in the US declined from 18.8% to 11.5% between July 2025 and January 2026. GitHub Copilot, meanwhile, is growing — but it's a different product with a narrow use case.
$80 billion in demand that cannot be met
There is also a flip side: Azure has $80 billion in orders in the queue that the company physically cannot fulfill due to capacity constraints — primarily a shortage of electricity for data centers. This means the problem is not lack of demand — but rather that infrastructure cannot keep up with it even despite record spending.
Meanwhile, an atypical buyer has appeared on the market: Michael Burry — the investor famous for betting against the mortgage market before the 2008 crisis — disclosed purchases of call options on Microsoft shares with strikes in the $700 range and expiration in 2028. Following this disclosure, the stock rebounded 5.7% in a single day.
The valuation has also changed: Microsoft is now trading at ~19 times expected earnings, which is below the S&P 500 average (~20x) and significantly below its own ten-year average of 27x.
If OpenAI truly moves a substantial portion of its workload from Azure to alternative providers by the end of 2026 — investors will get confirmation of the worst-case scenario: Microsoft built infrastructure for a customer who is leaving.
