Microsoft (MSFT): After Azure Growth, How Can the Platform Advantage Withstand the Cash Investment?

Company earnings update｜Data updated through November 8, 2026｜Amounts are in US dollars except per-share figures. The financial reporting period is FY2027 Q1, ended September 30, 2026. The company release, formal 10-Q, FY2026 annual report, and official earnings call transcript have been checked.

Research Conclusion

There is still evidence of improvement in the platform’s ability to monetize, but revenue growth and shareholder cash flow are moving at different speeds. Azure and Microsoft 365 cloud drove revenue, while operating expenses grew more slowly than gross profit, allowing operating income to grow faster than revenue. At the same time, the increase in cash equipment payments exceeded the increase in operating cash flow, and free cash flow declined. This shifts the research focus from “whether anyone is using AI” to “whether customer payments can cover equipment, leases, and supply chain funding and leave cash per share after replacement.”

Our view is that Microsoft plays two roles at once. At the platform layer, it controls the enterprise entry point, developer tools, and governance; the more intense competition among large language models becomes, the more models resemble interchangeable components and the more value concentrates at the platform. At the compute layer, it is a capital-intensive “utility,” bearing pressure from equipment, leases, and falling prices. This quarter’s gross margin change reflects the revenue mix shifting from the former toward the latter.

1. Why Are Enterprises Willing to Keep Paying?

Enterprise AI adoption requires compute, data, permissions, security, and applications that get work done. Microsoft sells cloud services, development tools, and enterprise software, allowing it to extend existing customer relationships to more uses. When customers integrate data governance, deployment, and workflows into these products, changing suppliers involves engineering, staff training, and security costs. This is the economic mechanism behind platform cross-selling and renewals.

But this mechanism does not automatically translate into high incremental cash returns. As software subscriptions expand, the cost of serving more users may be relatively limited; expanding GPU compute requires equipment, power, and data centers. If the revenue mix shifts toward compute, the company’s overall capital intensity will rise even if customer stickiness improves. The following sections therefore examine revenue sources, gross profit and expenses, cash equipment spending and long-term contracts in order, before returning to the platform role.

2. Where Is Growth Coming From?

This quarter’s results first changed the way the business must be viewed. The company moved to two segments, Agents and Infra and Devices and Consumer, and recast the prior-year comparison, so growth cannot be measured by directly comparing the new Agents and Infra with last quarter’s Intelligent Cloud. The new Azure definition includes GPU and non-GPU consumption services; Microsoft 365 commercial cloud includes GitHub cloud, Copilot, and some security products. The new classification is closer to management’s current product mix, but requires researchers to reconstruct a consistent series.Company segment descriptions and 10-Q metric definitions

Source: the product and service revenue table in the company’s FY2027 Q1 release. The prior-year figures were recast by the company under the new classification. Each share of the revenue increase is the item’s change divided by the total revenue increase of US$14.606 billion; because some items declined, positive contributions total more than 100%.

Approximately three quarters of the revenue increase came from Azure, with Microsoft 365 cloud contributing approximately 28%; together they accounted for approximately 66% of total revenue. Traditional licenses, Windows, and XBOX all declined. Microsoft’s growth engine is highly concentrated in cloud and enterprise subscriptions, while consumer and one-time licensing businesses no longer drive growth. The company also disclosed in its release that adjusted Microsoft 365 commercial cloud revenue grew 18% year over year, consumer cloud grew 17%, and search and advertising excluding traffic acquisition costs grew 12%.

Agents and Infra contributed approximately 88.3% of company-wide operating income, with its margin falling slightly, consistent with a cloud mix shifting toward compute. Devices and Consumer revenue declined, but costs and expenses fell more, so operating income still rose slightly; it is a shrinking source of cash rather than a source of growth. Segment operating income does not allocate all capital costs and cannot be treated directly as each business’s return on investment.

Product-level adoption signals point in the same direction. Management said on the earnings call that Foundry had more than 120,000 customers and revenue was more than triple the prior-year level; Copilot usage and the number of consumption-billed customers also rose. These are signs of actual commercial adoption, but the company did not disclose absolute Foundry revenue, an update on paid Copilot seats this quarter, or standalone profits for individual products, so their returns on investment cannot be calculated. Management also acknowledged weaker sales execution for M365 E3 and some individual products; the platform advantage still needs execution to be realized.Official earnings call transcript

3. How Did Core Operating Profit Improve?

Table amounts are in US$100 million, except EPS, which is in US dollars. Prior-year figures use the basis presented by the company this quarter.

Source: the company’s FY2027 Q1 release. Growth rates are calculated from the original figures and differ slightly from the company’s rounding to whole numbers.

Gross profit rose approximately US$8.929 billion and operating income rose approximately US$7.954 billion, leaving only approximately US$975 million in additional operating expenses between them. This is the main financial evidence of improved operating leverage this quarter. Consolidated gross margin fell from approximately 69.0% to 67.8%, while operating margin rose from approximately 48.9% to 49.8%; both can occur because expenses fell as a share of revenue.

The company’s 10-Q explains that Microsoft Cloud gross margin fell to 66%, partly because the mix shifted toward Azure, offset in part by product-level efficiency improvements. This is not evidence that gross margin deteriorated for every AI service, nor can the group’s absorption of expenses be treated as an improvement in GPU unit returns. The research conclusion should remain at the observable level: the platform as a whole can absorb more revenue, but the full economics of individual products have not been separately disclosed.

4. How Much of EPS Growth Reflects the Investment Comparison Base?

The OpenAI investment produced an after-tax gain of US$207 million and EPS of US$0.03 this quarter; a year earlier it produced an after-tax loss of US$3.086 billion and negative EPS of US$0.41. GAAP EPS consequently grew 32% year over year, well above the company’s 18% growth excluding this item. This difference cannot determine whether customer demand for OpenAI is real; it only shows that changes in financial investment gains and losses caused GAAP growth to diverge from a more comparable operating trend.Company non-GAAP reconciliation

Excluding investment gains and losses does not eliminate the company’s economic relationship with OpenAI. The 10-Q disclosed US$10.0 billion in quarterly revenue from commercial arrangements with OpenAI, including revenue share; receivables at quarter-end were US$7.2 billion, up from US$6.0 billion at the end of June. US$10.0 billion represents approximately 10.8% of quarterly group revenue, but cannot all be treated as Azure compute revenue because the disclosure explicitly includes different commercial arrangements. An increase in receivables alone does not prove they are overdue; payment periods, collections, and transaction terms are still needed.10-Q Note 1, OpenAI partnership disclosure

The structure of the relationship is also changing, but Microsoft has not reduced its investment. Microsoft’s commitment to invest in OpenAI remained US$13.0 billion, with the amount invested rising from US$11.9 billion at the end of June to US$12.0 billion at the end of September; its post-conversion stake fell from approximately 25% to approximately 24%. The FY2026 annual report explains that the decline resulted from the October 2025 restructuring and other OpenAI fundraising, and Microsoft consequently recognized a dilution gain rather than selling its stake. After the parties revised their agreement in April 2026, Microsoft’s license to OpenAI models and products continued through 2032 but became non-exclusive; OpenAI may sell products through any cloud while continuing to deploy on Azure first; Microsoft no longer pays revenue share to OpenAI, while OpenAI’s revenue share to Microsoft continues through 2030 at the same rate but with an aggregate cap.FY2026 annual report; OpenAI announcement of the revised agreement

Platform breadth and customer concentration risk therefore need to be considered together. A dispersed customer base does not mean funding sources for all major AI contracts are equally dispersed. Major partner customers’ payment, performance, and financing capacity should be checked quarterly to avoid treating supplier revenue and customer investment commitments as independent proof of end demand.

5. Why Is Free Cash Flow Lagging Operating Income?

Source: 10-Q cash flow statement and Note 12. CFO less cash equipment spending is recalculated by the analyst from the statements; lease asset acquisitions are not equal cash payments in the period and cannot simply be added as cash outflow.

CFO increased approximately US$12.943 billion, but cash equipment payments increased approximately US$19.973 billion, so simplified free cash flow fell approximately US$7.03 billion, down 27.4% year over year. This does not mean the company cannot generate cash. On the contrary, core operating cash flow is strong; the issue is that incremental cash still does not fully offset the increase in expansion payments. Over the same period, depreciation, amortization and other rose from US$8.147 billion to US$11.744 billion, an increase of approximately 44.2%, well above revenue growth; past expansion has begun to appear in the income statement.

Leases make consistent measurement even more important. New operating lease assets rose and new finance lease assets fell, meaning that looking at only one capital spending measure misses some future resource commitments. The amount the company obtains by subtracting cash equipment spending from CFO still does not deduct finance lease principal; deducting this quarter’s US$1.0 billion in principal produces a mechanical difference of US$17.633 billion. This is not another official company FCF measure, nor is it cash distributable to shareholders after all future obligations. Operating lease payments are already in CFO and should not be deducted again.

Supply chain funding cannot be ignored either. The 10-Q classifies some payments to assist purchases of server components as other receivables and investing activities rather than cash PP&E; these current other receivables rose from US$27.8 billion at the end of June to US$35.7 billion, with another US$12.6 billion of investments restricted under supplier agreements. This means looking only at equipment spending understates total infrastructure funding deployed in some periods. However, these payments are not equivalent to permanent consumption; whether they can be recovered, converted into equipment, or create other rights depends on the contracts.10-Q Note 1 and liquidity analysis

Capital Spending Guidance: Strong Demand, but Delivery and Recovery Mismatches Still Require Attention

First-quarter capital spending discussed on the earnings call was US$45.2 billion, including items such as asset acquisitions and amounts recognized at lease commencement; cash purchases of PP&E were US$39.367 billion. Management separately mentioned US$7.4 billion in finance leases for the quarter. Because recognition timing and measurement differ, US$45.2 billion and US$39.367 billion cannot be added, nor can their difference be used directly to calculate US$7.4 billion. Newly acquired finance lease right-of-use assets of US$5.192 billion in the 10-Q cash flow note and the US$7.4 billion on the earnings call also use different bases and cannot be treated as the same metric without a company reconciliation.Official earnings call; 10-Q

Management said approximately three quarters of earnings-call capital spending was on shorter-lived assets such as CPUs and GPUs, and second-quarter spending was expected to exceed US$50.0 billion; calendar 2026 was expected to be approximately US$170.0 billion, which is not the same period as the full FY2027 year. Third-party capacity delivery delays shifted some investment to the second half of the fiscal year. The lower calendar-year estimate therefore includes timing deferrals and cannot directly be interpreted as lower ultimate investment demand. The high share of short-lived equipment also means replacement and technology iteration must be considered after initial construction; returns cannot be assessed on the assumption that equipment is purchased only once.

Management expects FY2027 free cash flow to remain positive and said it will use other funding sources when conditions are favorable. This is an outlook, not a guarantee. First-quarter operating cash flow of US$58.0 billion less cash PP&E was US$18.633 billion; after further deducting US$6.757 billion in cash dividends and US$4.379 billion in share repurchases from the cash flow statement, approximately US$7.497 billion remained, before finance lease principal and other investments. This single-quarter calculation shows a cushion remains, but cannot prove that future expansion and buybacks can always be funded by internally generated cash; the US$10.2 billion in shareholder returns cited on the earnings call should not be directly conflated with cash flow statement repurchases as though measured on the same basis.

6. How Do Contracts and Assets Create Long-Term Tradeoffs?

Commercial RPO was US$677 billion, up 73% year over year; approximately 30% is expected to be recognized over the next twelve months, and the weighted average contract duration is approximately 2.3 years. RPO demonstrates contract visibility, not current cash or future profit. The company still must deliver compute, bear costs, and collect from customers. Conversely, asset and lease investments occur first; if customers delay consumption or renegotiate, the timing between revenue and funding outlays will stretch.

Lease commitments not yet commenced at the end of September were approximately US$364.6 billion, mainly for data centers, some conditional, and expected to commence gradually from FY2027 through FY2034. They cannot be treated as borrowing this quarter, debt immediately due, or capital spending this year. Their research significance is that the company has made substantial arrangements for future capacity, and whether revenue converts as expected will affect future flexibility.10-Q Notes 11 and 12

Cash and short-term investments totaled US$64.509 billion in the quarter, down from the end of June. The company still has ongoing operating cash flow and access to capital markets, and this article found no evidence on which to conclude there is an immediate funding crisis. But “access to financing” and “good investment returns” are separate questions, and the former cannot substitute for the latter.

7. The Platform Role: Where Does Value Flow as Model Competition Intensifies?

The financial figures above answer “how much cash is being invested”; this section answers another question: where Microsoft makes money in an industry where models are rapidly becoming commodities.

Demand and Pricing: Usage Growth Must Offset Lower Unit Prices

Demand for AI infrastructure consists of model training, inference, and enterprise applications, but the payers, duration, and equipment required differ among these three categories. Training projects may lead large customers to concentrate purchases over a short period; inference depends on whether end users return to use services and whether enterprises are willing to pay for results. If model efficiency improves, compute required for each task falls, and demand may also expand as prices fall. The resulting revenue depends on whether usage growth can offset lower unit prices, not merely on the number of model calls.

Microsoft’s competitive advantage is its simultaneous control of the enterprise entry point, data and security tools, applications, and cloud infrastructure. Customers may lower integration costs by entering through Microsoft 365, building applications with Foundry, and running them on Azure; however, customers can also use other clouds, models, or in-house technology. The better the platform supports multiple models, the more likely it is to retain customer workflows; at the same time, model providers and customers can more easily compare prices, so Microsoft must prove its added value through service quality, trust, and cost efficiency.

Reverse Engineering and Open Source: The Cost of Copying Software and Models Is Falling

As AI coding capabilities improve, the cost of rewriting software based on public features, file formats, or usage behavior is falling. This is not merely an old issue for the open-source community; the speed of copying has changed. Microsoft’s 10-Q risk factors state directly that some competitors modify and distribute open-source software at very low or zero cost to users, or use open models and earn money through advertising or integrated services; these firms do not bear all the research and development costs of open-source products, and some open-source products also imitate features of Microsoft products. The company believes this could lead to lower sales, price reductions, or higher expenses.10-Q risk factors

At the application layer, features of Office, Windows, and server licenses are easier to imitate; but enterprise switching costs come mainly from data migration, permissions and compliance settings, compatibility with existing systems, and employee training, not just the code itself. This quarter, productivity and server license revenue fell approximately 3.1% year over year and Windows OEM and devices fell approximately 22.0%, but the financial statements did not attribute these changes to open source or imitation, so they cannot prove substitution has occurred.

Pressure is more apparent at the model layer. Management said Microsoft’s own MAI models aim to achieve comparable capabilities within three to six months of leading model releases at a fraction of the training cost; the CEO also said some customers use execution records accumulated with frontier models to train their own models or switch to open-weight models. The ability to catch up quickly in model capabilities gives the platform an opportunity to pass cost advantages to customers, but also means the premium period for a single model is shortening; the value of Microsoft’s OpenAI intellectual property and revenue share is likewise affected by this speed, and the revenue share already has an aggregate cap. The latter point is our inference; the company has not quantified it.Official earnings call transcript

Microsoft’s Response: Actively Making Models Interchangeable Components

Microsoft is not betting on a single model; it makes models compete while controlling routing and governance. The CEO said on the earnings call that developers using coding agents will want Anthropic and OpenAI to compete for workloads on every task, not once a quarter, which is precisely the value of routing and automatic dispatch. Specific actions include:

• GitHub Copilot Fusion: Officially launched October 28, 2026, it does not require users to choose a model; depending on the task, the system uses a single model, starts with a cheaper model and escalates based on checks, or has a second model review and revise the work. GitHub has not disclosed the complete model pool; official documentation says only that available models will change over time.GitHub changelog
• Foundry multi-model: The earnings call disclosed that revenue from open-source models on Foundry grew 10 times, and customers building agents with multiple models increased 5 times. Media reported that Microsoft announced in November 2025 a commitment to invest up to US$5.0 billion in Anthropic, while Anthropic committed to purchase US$30.0 billion of Azure compute, and Claude models were made available on both Foundry and Copilot.TechRadar
• MAI proprietary models: MAI Code Flash had 3 million users on GitHub Copilot last month; MAI Cyber was responsible for more than 40% of scans by Microsoft’s own agentic code-scanning tool, at an 18% lower cost per scan.
• Open-source cost reduction tools: Microsoft Research, Microsoft Security, and UC Berkeley open-sourced TokenEconomizer, which can be attached to agents including Claude Code, Codex, and Copilot CLI to reduce tokens by remembering stable project facts, removing unused tools, and intercepting failure loops. In the authors’ own SWE-Bench tests, average costs across five agent and model combinations fell 18.4% to 30.9%; these figures have not been independently verified.GitHub project page

GitHub itself is also a major platform for open source and code collaboration. Management said PRs opened by AI agents increased sixfold over six months and already account for more than one third of all PRs on the platform. Open source and imitation shift value from “owning code” toward “execution, governance, cybersecurity, and compute,” the areas where Azure, GitHub, and security products charge customers. The cost is that as software and models become easier to replace, customers can also compare cloud providers’ prices more easily. Reverse engineering and open source exert continuing pressure on individual licensed products, while shifting the points where a platform can charge: whether Microsoft benefits depends on whether revenue can move smoothly from license fees to charges for usage, governance, and compute, and compute revenue requires far more capital than software licenses.

Compute as a Utility: Savings Improve Capital Intensity and Gross Margin

Viewing compute as a utility, efficiency gains can produce two improvements. The first is capital intensity: the CFO said on the earnings call that as infrastructure efficiency improves, each dollar of capital spending may support more revenue and higher returns. But what is more likely to fall is “capital spending required per dollar of revenue,” not the absolute amount of capital spending; guidance on the same call still called for more than US$50.0 billion in the second quarter and acceleration in the second half, and the US$364.6 billion in signed leases not yet commenced will not be withdrawn because efficiency improves. If efficiency advances faster than demand, idle capacity could instead emerge.

The second is gross margin. When services move to outcome-based pricing and the platform decides which model to use behind the scenes, customers pay a price for completed work, while the platform can lower costs through routing, proprietary models, and token savings and retain the difference. This mechanism is our inference; the company has not disclosed gross margin by product. A preliminary sign in the financial results is that Copilot customers actually consuming credits through usage billing increased more than 75% in one quarter, and management described such spending as directly tied to the value delivered by agents. The limitation is that outcomes must be measurable, and the cost of failed tasks also falls on the platform; OpenAI’s Codex and Anthropic’s Claude Code also sell “results” directly and compete head-on with GitHub Copilot.

Three Layers, Different Degrees of Benefit

My judgment is that the thesis “the more intense model competition becomes, the more the platform benefits” holds at Microsoft’s platform layer and already has financial evidence; but Microsoft is also a compute supplier, and this layer bears falling prices and capital burdens. Microsoft Cloud gross margin falling to 66% is a structural result of the revenue mix shifting from the high-margin platform layer toward the capital-intensive compute layer. Continued platform benefits depend on Microsoft retaining the customer entry point; after the contract revision OpenAI can sell on any cloud, and model providers continue to extend into the application layer, making this the threat the platform thesis most needs to monitor.

8. Bull and Bear Cross-Examination: Is Growth Sufficient to Support Investment?

Bull question: Azure revenue grew 48% year over year, CFO grew, and gross profit and operating income increased together; why still question the investment? The answer is that these figures prove strong demand and current profitability, but do not separately reveal the full revenue, depreciation, maintenance, electricity, and replacement costs of the new generation of equipment. The investment may be reasonable, but cash recovery must be tracked over a longer period; revenue growth alone cannot establish payback.

Bear question: Do a 27.4% year-over-year decline in FCF, large leases, and exposure to one major customer mean the company is overexpanding? That conclusion cannot yet be drawn directly. Investment preceding collection is normal timing for infrastructure construction, and Microsoft still has strong subscription revenue, cash generation, and enterprise customers; but if future utilization falls short of expectations, fixed commitments and depreciation will remain, and operating leverage may work in reverse.

Bull follow-up on customer concentration: Has US$10.0 billion in commercial revenue from OpenAI proved ultimate enterprise demand? Not entirely. The amount covers multiple commercial arrangements and revenue sharing, not just compute sales; receivables and financing capacity should also be considered. This is not a judgment of default, but a reminder to distinguish contract revenue, end-user payments, and cash collection.

Bear follow-up on efficiency: Could improvements in model and chip efficiency make existing equipment obsolete early? Possibly; but as discussed in section 7, efficiency may also increase the workload served by the same equipment, improve gross margin, and expand usage. Both effects coexist. Equipment utilization, unit compute costs, prices, and actual incremental demand should be monitored; the company does not fully disclose these metrics, so precise incremental ROIC cannot be calculated for it.

My judgment: Microsoft currently has real customer demand, platform distribution capabilities, and strong cash flow to support continued investment; at the same time, short-lived equipment, leases not yet commenced, and large-customer concentration mean “strong growth” is insufficient to prove “strong investment returns.” I am positive on the company’s competitiveness and conditionally cautious about AI capital recovery. This is an operating judgment, not trading advice.

9. Medium- and Long-Term Concerns: Faster Growth Makes Replacement and Commitments More Important

Taken together, the preceding sections show no financial crisis this quarter, but there are five medium- and long-term concerns, ranked by importance. They share a common feature: revenue growth is strong, but each dollar of growth requires increasing capital, and a significant portion consists of commitments that are difficult to withdraw.

First, capital intensity is rising and cash conversion is falling. Simplified free cash flow fell from US$25.663 billion to US$18.633 billion, and depreciation, amortization and other increased approximately 44.2%, while operating income increased over the same period. Approximately three quarters of capital spending is on short-lived assets that must be replaced and repurchased after several years, so payback cannot be assessed as if the infrastructure only needs to be built once.

Second, fixed commitments are large and difficult to exit when demand slows. US$364.6 billion in leases not yet commenced will start gradually through FY2034; other receivables related to server components are US$35.7 billion, and another US$12.6 billion in investments is restricted under supplier agreements, tying approximately US$48.3 billion in total to supply chain arrangements. If AI usage falls short of expectations, revenue will slow first, while rent, depreciation, and component arrangements will not decline at the same time, and operating leverage may work in reverse.

Third, OpenAI concentration links the customer’s financial capacity to Microsoft’s revenue. Quarterly revenue of US$10.0 billion accounts for approximately 10.8% of group revenue, receivables continue to rise, and the stake is approximately 24%. If OpenAI’s operating cash or external financing is insufficient to support its compute purchases, Microsoft’s revenue, receivables, and investment value would be affected simultaneously; the post-revision revenue share cap also limits upside. This is not a default judgment, but a reminder that the two parties’ financial performance is not independent.

Fourth, the revenue mix is shifting toward compute, potentially diluting pricing power. Microsoft Cloud gross margin fell to 66%, and group gross margin fell from approximately 69.0% to 67.8%. As discussed in section 7, compute is easier to compare on price than software subscriptions, and open-source models and rapid catch-up also make it easier for customers to switch models and platforms; the company has not disclosed enough unit-level data to verify whether usage growth can offset lower prices.

Fifth, traditional per-seat businesses are showing signs of weakness. Management acknowledged weaker sales execution for M365 E3 and individual products, and Windows OEM and devices revenue fell approximately 22% year over year. If AI agents reduce the number of users enterprises need, per-person pricing will face long-term pressure; the company is shifting revenue toward premium seats and consumption billing, but the new model’s margins and stability remain to be seen. This point is our inference; the quarter’s financial results did not show a decline in seats.

The forces offsetting these concerns are also real: quarterly operating cash flow was US$58.0 billion, up approximately 29% year over year, commercial RPO was US$677 billion, up 73%, and operating margin was still rising. A more accurate description at present is therefore a timing issue in which “investment precedes recovery.” The real turning point is whether free cash flow per share can resume growth when the first wave of AI equipment enters its replacement cycle.

10. How Can This Assessment Be Tested Going Forward?

The positive scenario is continued payment for Azure and enterprise applications, customer consumption and payments keeping pace after capacity investment, cloud efficiency offsetting price competition, and cash per share returning to growth after equipment replacement. The cross-product platform advantage would then have evidence in both revenue and cash.

The neutral scenario is continued revenue and operating income growth, but higher funding requirements for leases and equipment and a longer cash recovery period. The company remains competitive, but valuation requires assumptions of lower cash conversion and a longer payback period.

The negative scenario is weaker financing or consumption by major customers, leases commencing on schedule, and depreciation and rent continuing to rise. Large RPO figures may not prevent earnings revisions downward, because delivery costs and customers’ ability to pay still determine the economic outcome.

The next quarterly report should track Azure, Microsoft 365 cloud, cloud gross profit, OpenAI transactions and receivables, CFO, cash equipment spending, and leases under the new segment and product definitions. To test the platform thesis in section 7, three additional indicators can be monitored: whether capital spending as a share of revenue begins to decline, whether Microsoft Cloud gross margin can stop falling, and the revenue share from usage-based and outcome-based billing. Official guidance calls for FY2027 Q2 total revenue of US$96.0 billion to US$97.1 billion and Azure constant-currency growth of 49% to 50%, but this is management’s outlook, not actual results. This article does not include the latest price or consensus expectations and does not infer whether the stock price reflects the gap versus expectations.