Microsoft's Selloff Mistakes a Supply Problem for a Demand Problem

Microsoft stacks three businesses on one balance sheet: a productivity franchise (Microsoft 365, LinkedIn, Dynamics) that throws off subscription cash, an Intelligent Cloud segment built around Azure that is now the company's growth engine, and a consumer/personal-computing tail (Windows, Xbox, Surface, Bing) that generates headlines disproportionate to its weight. The entire bull-versus-bear argument today reduces to one question inside the middle business: is the roughly $190 billion the company plans to spend on AI infrastructure this year a rational investment against contracted demand, or a margin-destroying arms race with no payback? The market has spent June voting for the second interpretation. I think the evidence points firmly to the first, and that the gap is where the opportunity sits.

The setup: a violent repricing of the same numbers

The stock closed near $373 on June 26 after a sharp single-day bounce, but that masks how brutal the month has been — shares are down more than 10% so far this month. Over the trailing year the picture is similar: Microsoft has delivered roughly a -25% change over the past year, with a 52-week range between $349.20 and $555.45. The all-time high closing price was $538.66 on October 28, 2025 — meaning the stock now trades around 30% below its peak.

What makes this drawdown unusual is that it did not follow a disappointing operating result. It followed two of them that beat. The trigger was capital intensity, not demand. After the prior quarter, Microsoft suffered its worst single-day market cap loss in history, with roughly $357 billion erased despite a headline revenue beat. The most recent quarter then delivered the disclosure that did the real damage: a 2026 calendar capex plan of approximately $190 billion, exceeding analyst consensus of $154.6 billion by roughly $35 billion, with $25 billion of that figure attributed to component price inflation rather than additional capacity. A memory crunch is making the buildout more expensive at the same moment investors decided they were tired of paying for it.

The demand signal is contracted, not hoped-for

Here is what the panic is stepping over. In its fiscal third quarter (ended March), Microsoft posted revenue of $82.9 billion, up 18% year over year, and Microsoft Cloud revenue of $54.5 billion, up 29%, with Azure and other cloud services growing 39% in constant currency. Azure's roughly 40% print exceeded the guided 37%–38% range and consensus near 35%–36% — an acceleration, not a deceleration, in the business absorbing the capex.

More important than any single growth rate is the order book. The AI business surpassed a $37 billion annual revenue run rate, up 123% year on year, and commercial remaining performance obligations grew 99% to $627 billion. That backlog deserves scrutiny rather than awe — a chunk of the jump reflects the OpenAI commitment, and the 26% growth rate excluding OpenAI commitments is the more operationally comparable baseline. But even the deflated figure describes a multi-quarter pipeline of signed contracts, not a marketing projection. For context, the company's total remaining performance obligation stood at $375 billion at the close of fiscal 2025; the commercial slice alone now exceeds that. Contracted revenue of this magnitude tends to convert as capacity becomes available to serve it. That is the crux: the constraint on near-term Azure revenue is supply, not whether customers want the product.

The adoption telemetry agrees. Microsoft 365 Copilot paid seats reached over 20 million, a 33% sequential increase from 15 million in January, with seat additions growing 250% year over year, and customers purchasing more than 50,000 Copilot seats quadrupled over the past twelve months. This is the enterprise flywheel — usage tied to value, sold through distribution Microsoft already owns — doing exactly what the thesis requires.

The margin story is uncomfortable, and fixable

The bears are right about one thing: the economics are visibly compressing right now. Microsoft Cloud gross margin fell to 66%, driven by continued investments in AI infrastructure and growing AI usage, offset in part by efficiency gains in Azure and Microsoft 365 Commercial cloud, and overall gross margin of 67.6% was the narrowest since 2022 as depreciation costs mounted from the data center build-out. When you stand up hundreds of billions in depreciating gear ahead of the revenue it serves, reported margins sag. That is mechanical, and it is the price of building into a supply shortage.

What the selloff underweights is that Microsoft has run this play before — it absorbed margin compression during the first Azure buildout and methodically clawed it back as utilization improved — and this time it holds more levers. The first is silicon. Its Cobalt CPU is now deployed in nearly half of Azure regions, and the Maia 200 accelerator, built on TSMC's 3nm process, is designed to deliver roughly 30% better performance per dollar than the prior generation of Microsoft's own fleet hardware. As inference shifts onto first-party parts, the unit cost of serving AI workloads bends away from merchant GPU pricing — precisely the cost line that is squeezing cloud margins today.

The second lever got almost no airtime amid the capex noise. Microsoft restructured its OpenAI partnership in late April, eliminating outbound revenue-share payments to OpenAI while retaining Azure priority and a non-exclusive model license through 2032 — removing a long-term cost liability that flows straight to the AI margin trajectory. And even the dreaded capex number was less reckless than the headline implied: in the quarter, Microsoft reported $31.9 billion in capital expenditures, while management signaled headcount will decline year over year in calendar 2027. This is a company adding compute aggressively while taking cost out of its operating base — not one spending blindly.

The regulatory drumbeat is real but narrow

The most credible bear case is not capex; it is monetization friction, and the calendar is noisy. On June 25, the EU issued a preliminary report favoring designation of Azure as a gatekeeper under the Digital Markets Act. A day later, Italy's competition authority opened a formal investigation into Microsoft over the integration of Copilot and Designer into Microsoft 365 subscriptions. The consumer-pricing mechanics are squarely in the frame: the AGCM argues Microsoft communicated price hikes in a fragmented manner and placed consumers on a costlier plan by default unless they explicitly withdrew. Layer on ongoing examination from the U.S. FTC, the LinkedIn GDPR appeal still working through the Irish courts, and unresolved IRS transfer-pricing notices seeking $28.9 billion plus penalties and interest, and there is plenty for a jittery market to seize on during headline days.

But size the actual exposure. The Italian probe, headline-grabbing as it is, concerns how the price change was communicated, not whether Microsoft may charge more for a product it has expanded. The plausible remedies — clearer disclosure, an unbundled option, refunds at the margin — dent the slope of consumer ARPU, not the enterprise engine that carries the thesis. A theoretical fine of up to 10% of global turnover sounds existential, but consumer-code cases of this type settle in the millions, not the tens of billions. Microsoft's answer is the one it has deployed for two decades of antitrust: comply on packaging and consent while keeping the economic core — model choice, governance, enterprise data context — intact. The DMA gatekeeper question for Azure is more strategically meaningful, but it is a multi-year process about interoperability obligations, not a near-term revenue event.

The consumer noise, including a More Personal Computing segment that came in at $13.19 billion, down 1% as Xbox hardware weakens and PC dynamics normalize, distracts from where value actually accrues. The heartbeat is Azure, the E5-plus-Copilot motion, Dynamics, GitHub, and the Fabric/Foundry data layer — the B2B businesses where Intelligent Cloud grew 30% and Productivity and Business Processes grew 17%.

Valuation: paying a software-average multiple for above-average growth

The repricing has done something rare to this name. At roughly $373 against trailing earnings, the stock carries a P/E in the low-to-mid 20s — a price-to-earnings ratio near 22.5 on the most recent prints — versus the well-above-30x multiple it commanded near last October's high. The latest quarter delivered $4.27 in EPS against a $4.06 estimate, and the next quarter is modeled near $4.24, which frames a business still compounding earnings in the mid-teens while operating margins expanded to 46.3% from 45.7% a year ago. You are now paying a market-average multiple for a franchise growing its most important segment at 30%, with a contracted backlog that lengthens revenue visibility and a balance sheet — net cash, a $60 billion buyback authorization barely tapped, a rising dividend — that removes financing risk from the capex story entirely.

What to watch and the stance

The swing factor is concrete and dated. Microsoft reports next in late July 2026. The print that flips sentiment is one that shows three things together: Azure holding near 40% in constant currency, evidence that capacity is arriving earlier than feared, and any sequential stabilization in cloud gross margin. The first two are squarely within reach given the backlog and the silicon roadmap; the third is the one investors will pay up for the moment it appears, because it converts the "AI eats margins" narrative into "AI scales margins." Microsoft Cloud gross margin percentage was slightly better than expected at 68%, down 2 points year-over-year from the impact of scaling AI infrastructure, partially offset by continued efficiency gains in Azure and M365 Commercial Cloud.

I land on the bullish side, and not as a hedge. The biggest worries — capex magnitude, margin compression, regulatory headlines — are fully on the table and arguably over-discounted after a 30% derating. The biggest constructive variable, capacity catching up to a contracted order book, is pointed the right way and largely within management's control. When supply meets a $627 billion backlog, revenue should outrun cost of goods, and the multiple that has compressed to the market average has room to re-rate toward the growth Microsoft is actually delivering. The setup into the next two prints is asymmetric, and the asymmetry favors owners.

Disclaimer: Tenzing Analytics, Inc. is not a registered investment adviser, broker-dealer, or financial institution, and all content provided by Tenzing Memo is AI-generated for informational purposes only. It does not constitute investment advice, a recommendation, or an offer or solicitation to buy, sell, or hold any security or financial instrument. Content is general in nature, not tailored to any individual's financial situation or objectives, and may contain errors, omissions, or inaccuracies; it should not be relied upon as a primary basis for any investment decision. Any forward-looking statements are subject to risks and uncertainties and may not reflect actual future results. Users should conduct their own research and consult qualified financial, legal, or tax advisors before making investment decisions, and by accessing this content, you acknowledge and accept these limitations.