Microsoft (MSFT) Deep Dive: AI capex faces an earnings test

Microsoft's investment case now turns on cash returns from AI capex. FY2026 Q3 revenue was $82.886B, up 18%; operating income was $38.398B, up 20%. Net income was $31.778B, up 23%, while diluted EPS was $4.27, up 23%. Microsoft Cloud revenue reached $54.5B, up 29%, and Azure and other cloud services grew 40%. AI business annual recurring revenue (ARR) exceeded $37B, up 123%; Microsoft 365 Copilot paid seats passed 20M. Commercial remaining performance obligations (RPO) reached $627B.

The friction is cash conversion. FY2026 Q3 cash paid for property, plant and equipment (PP&E) was $30.876B and free cash flow (FCF) was $15.803B. Management guided Q4 capex above $40B and calendar-2026 capex near $190B. The stronger cloud and AI demand becomes, the more investors demand proof of unit capital returns.

Price comes first. MSFT closed at $386.74 on 2026-07-06: about +4.9% over one week, -7.2% over one month, +3.9% over three months, -17.8% over six months and -19.7% YTD. QQQ and XLK have been much stronger, which shows that Microsoft has de-crowded inside the AI cloud trade.

The stock is about 30.4% below its 2025-07-31 one-year high of $555.45 and about 10.8% above its 2026-06-25 one-year low of $349.20. The 20D moving average is near $384.9, the 50D near $406.7 and the 200D near $444.8. RSI is around 48. Support sits near $382, $353 and $349. Resistance sits near $403-407, $417-419, $428-436 and $452.

The attribution confidence is medium. AI capex and FCF-yield pressure are central, supported by FY2026 Q3 cash PP&E of $30.876B and FCF of $15.803B. Large-cap technology rotation also matters because QQQ and XLK outperformed MSFT across three-month, six-month and one-year windows. Legal developments add short-term pressure, while workforce actions and Copilot ROI discussion also weigh on sentiment. MSFT lagged QQQ by about 52pp and XLK by about 65pp over one year. The size of that gap points to capex cash-flow discount and factor rotation as the joint drivers.

The dominant driver is the cash return on AI capex. Capacity spending first supports Azure consumption and Copilot use, which then affects cloud gross margin and free cash flow. Those outcomes ultimately determine valuation multiples. Sustained cash returns support Microsoft's premium software-cloud valuation, while weaker margins or free cash flow leave revenue growth offset by FCF-yield pressure.

Capacity intensity is clear. FY2026 Q3 capex was $31.9B, with roughly two-thirds in short-lived GPU and CPU assets and one-third in longer-lived assets that can support monetization for more than 15 years. Management added about 1GW of capacity in the quarter and aims to double AI footprint within two years.

Revenue contribution is also visible. Azure grew 40%, and AI ARR exceeded $37B. Foundry and Fabric adoption is scaling, while M365 Copilot and GitHub Copilot Enterprise metrics continue to rise. Security Copilot adds another enterprise revenue stream. Conversion speed must cover cash PP&E and finance leases; depreciation also affects returns, while power and component inflation raise operating costs. FY2026 Q3 FCF conversion was about 49.7% of net income, far below FY2024's 84.0%.

Microsoft has three major business segments. Productivity and Business Processes generated $35.013B of FY2026 Q3 revenue, up 17%, and $20.973B of operating income, for about 59.9% margin. This segment stabilizes the valuation. Copilot can drive enterprise subscription upgrades, increase M365 Copilot adoption and expand usage credits. It can also improve retention.

Intelligent Cloud generated $34.681B of revenue, up 30%, and $13.753B of operating income, for about 39.7% margin. Azure and other cloud services grew 40%, making them the company's primary growth driver. Segment margin is pressured by AI investment and GitHub Copilot usage, partly offset by Azure efficiency.

More Personal Computing generated $13.192B of revenue, down 1%, and $3.672B of operating income. Windows OEM and Devices declined 2%, Xbox content and services declined 5%, and search advertising ex-TAC grew 12%. The segment can affect sentiment, while consolidated valuation relies more heavily on the first two business segments.

Azure is the infrastructure core. FY2026 Q3 Azure and other cloud services grew 40%, with faster capacity delivery during the quarter. Demand remains above available capacity, so the short-term variables are capacity delivery and inference cost.

Foundry is the model marketplace. The platform supports OpenAI, Anthropic, open-source models and Microsoft's own models. More than 10,000 customers used more than one model, 5,000 used open-source models, and Anthropic plus OpenAI model users doubled sequentially. Claude availability in Microsoft Foundry and Azure strengthens multi-model positioning.

Copilot provides application distribution. M365 Copilot paid seats passed 20M, and customers with more than 50,000 seats increased about fourfold year over year. Accenture deployed 740,000 seats. Bayer and Johnson & Johnson joined Mercedes and Roche in committing more than 90,000 seats. First-party AI-assistant monthly usage is up about 6x year to date, and Copilot queries per user grew about 20% sequentially.

Frontier Company, a $2.5B initiative with about 6,000 industry and engineering experts, is designed to turn enterprise knowledge into AI systems. Copilot Cowork usage-based pricing supports deeper enterprise use, while consumer and enterprise Copilot integration broadens distribution. PitchBook integration with Copilot in Excel adds a specialized workflow, and NHS England adoption extends use in regulated settings.

Public social discussion provides incremental confirmation. Management posts about Frontier Company and data-center efficiency drew the strongest business attention. Copilot Cowork and NHS Copilot posts highlighted enterprise adoption, while in-house MAI models focused attention on tokens per dollar. This evidence carries lower valuation weight than filings. It explains the shift in investor focus from seat counts toward usage and inference cost, with changes in customers' operating adoption and usage as the next test. Risk-focused discussion centers on Copilot ROI and layoffs; resource pressure from the AI investment cycle aligns with litigation headlines in the news flow.

Microsoft is a downstream AI platform. It carries GPU and CPU investment, along with data-center power and lease costs. Enterprise customer relationships and application distribution are central advantages, while identity and security strengthen retention. The developer channel adds another source of demand. Nvidia has greater upside to upstream scarcity, while Google and Meta have a larger valuation cushion from free cash flow yield. Microsoft's mature software businesses provide more capacity to absorb AI capex than Amazon and Oracle.

Infrastructure competition centers on AI workloads, with Azure and AWS facing Google Cloud and Oracle Cloud. Enterprise application competition includes Microsoft 365 Copilot and Salesforce; Google Workspace, Adobe and ServiceNow also compete for knowledge work. Developer platform competition is more fragmented. GitHub Copilot and Foundry rely on Azure integration against Vertex and Bedrock, while Databricks, Snowflake and OpenAI API compete for model calls. Microsoft's advantage spans identity and data, with Office and Teams reinforcing distribution through GitHub and Azure.

The long-run trajectory still shows profit expansion. FY2023 revenue was $211.915B, operating income $88.523B, net income $72.361B, OCF $87.582B, cash PP&E $28.107B and FCF $59.475B. FY2024 revenue was $245.122B and FCF $74.071B. FY2025 revenue was $281.724B and FCF $71.611B.

FY2026 Q3 amplifies the tension. Revenue and profit remain strong, while cash PP&E reached $30.876B and FCF was $15.803B. Capex as a share of revenue rose from 13.3% in FY2023 to 22.9% in FY2025 and 37.3% in FY2026 Q3. Capex/OCF rose from 32.1% in FY2023 to 47.4% in FY2025 and 66.1% in FY2026 Q3.

Commercial RPO reached $627B in FY2026 Q3, up 99%. About 25% will convert over the next 12 months, with the longer-duration portion up 138%. Excluding OpenAI impact, commercial RPO grew about 26%, which keeps core enterprise demand positive while highlighting OpenAI duration and capacity matching as key variables.

FY2026 Q4 guidance gives hard thresholds. Total revenue is guided to $86.7B-$87.8B, while Productivity and Business Processes is guided to $37.0B-$37.3B. Intelligent Cloud guidance is $37.95B-$38.25B, with Azure constant currency growth at 39%-40%. More Personal Computing guidance is $11.75B-$12.25B. Microsoft Cloud gross margin is expected near 64%, and Q4 capex is expected above $40B.

For FY2027, management expects revenue and operating income to grow double digits, opex to grow mid to high single digits, and headcount to decline year over year. If delivered, the market can treat high capex as demand-backed investment. If revenue growth, cloud margin and FCF weaken together, the valuation can shift toward a heavier infrastructure cloud multiple.

We use peer multiples, DCF and scenarios. At $386.74, Microsoft trades near $2.87T market cap, about 23x P/E, 20x-21x forward P/E, 9x P/S and 39x P/FCF. Profit quality supports a premium multiple, while roughly 2.5% FCF yield limits upside.

The peer table places Microsoft in the middle: stronger margins than most software and cloud peers, with weaker FCF yield because capex is heavy. Mature software earnings provide more cash-flow protection than Oracle and Amazon. Infrastructure exposure is heavier than at CRM and Adobe, while Nvidia offers greater scarcity upside.

Multiples imply $380-455. Forward EPS of $18-19 at 21x-24x P/E points to $420-440. Forward revenue of about $350B-$365B at 7.5x-8.5x P/S points to $365-415. EV/EBITDA at 15x-18x points to $380-465.

DCF is a sensitivity check. The base case starts with about $65B of FCF, assumes five-year FCF CAGR near 14%, fades toward 5%, uses 3% terminal growth and 8.0% WACC, plus about $38B of net cash. The base DCF is about $430. A faster FCF repair and 7.5% WACC can move value above $500; a longer high-capex phase and 8.5% WACC can pull value near $350.

A stricter 10-year DCF starts with TTM revenue of $318.3B, TTM FCF of $72.9B and 22.9% FCF margin, then fades revenue growth from 17% toward 4.5%, lifts FCF margin from 23.5% to 27.0%, applies 8.3% WACC and 3.0% terminal growth. That produces about $376 per share, close to spot. Upside depends mainly on cloud AI execution and multiple repair.

Scenarios are: bull 25% at $540, base 50% at $430, bear 25% at $320. The weighted center is $430. The wide span from $320 to $540 keeps confidence at medium.

Management is emphasizing capital efficiency. Faster capacity delivery and GPU lifetime optimization improve asset use, while higher inference throughput supports revenue generation. MAIA 200 tokens per dollar and Cobalt deployment show progress in custom silicon. FY2027 opex discipline adds a cost constraint. AI spend must translate into revenue and margin.

Capital return continues. Microsoft returned $10.2B to shareholders in FY2026 Q3. Through the first nine months of FY2026, buybacks were about $17.7B and dividends about $19.7B. With nine-month FCF near $47.3B, capital return absorbs a large share of FCF. If AI capex stays high, buyback flexibility can compress while dividend durability remains strong.

Insider disclosures are a governance input. Form 4 activity at a megacap often includes RSU vesting and tax withholding. Preset plans and open-market trades require separate interpretation. Operating data and cash flow carry much higher valuation weight, with guidance providing another core input.

FY2026 Q4 is the nearest catalyst. Azure 39%-40% cc growth and Microsoft Cloud GM near 64% would confirm cloud resilience, while Copilot seats and usage credits would test application monetization. Foundry multi-model growth can support Azure inference demand. Frontier Company customer deployments add an enterprise adoption signal, and FY2027 double-digit revenue and operating income guidance would extend the investment horizon.

Recent events split into positive commercialization signals and negative discount factors.

AI capex/FCF pressure is the main risk, tracked through capex/OCF and cash PP&E, with FCF/net income showing the conversion effect. Cloud gross-margin compression would reinforce that pressure. OpenAI commitment duration raises the importance of the RPO share converting over the next 12 months. Copilot ROI and litigation headlines increase the burden of proof for paid seats and queries per user; usage credits provide another operating test. Regulation and cloud platform rules create policy risk. Data-center power and water affect expansion costs, while cyber security and AI liability remain standing valuation discounts.

The thesis can be tested through thresholds. If FY2026 Q4 revenue falls below $86.7B, or Azure cc falls below 39% while capex exceeds $40B, capacity conversion is weak. If FY2026 Q4 COGS exceeds $29.6B and opex exceeds $19.4B after one-time retirement costs, expense discipline is weaker. If FY2027 Q1 Azure cc falls out of the high-30s range and Microsoft Cloud GM falls below 65%, the second-half acceleration and efficiency offset assumptions are impaired. If CY2026 capex exceeds $210B while Azure cc is below 40% and Cloud GM is below 65%, the bear weight should rise.