Microsoft (Minutes): 1Q capital expenditure $30 billion, full-year expenditure higher in the first half and lower in the second half

The following is the earnings call minutes for$Microsoft(MSFT.US) FY25Q4 organized by Dolphin Research. For earnings interpretation, please refer to "Azure's Unlimited Growth? Microsoft Deserves the AI Leadership"

I. Review of Core Financial Information

1. Supplementary Information

Capital Expenditure: Capital expenditure of $24.2 billion (including $6.5 billion in finance leases), with more than half supporting long-term assets over 15 years, and the rest for servers and GPUs.

Cash Flow: Operating cash flow of $42.6 billion (up 15%), free cash flow of $25.6 billion.

Shareholder Returns: Over $37 billion returned to shareholders for the year (including $9.4 billion this quarter, comprising dividends and buybacks), with an effective tax rate of approximately 17%.

2. 1Q26 Guidance:

a. Foreign Exchange: Foreign exchange is expected to boost revenue growth by 2 percentage points. The Productivity and Business Processes segment is expected to grow by about 3 percentage points due to foreign exchange, while the Intelligent Cloud and More Personal Computing segments are each expected to be boosted by about 1 percentage point. Foreign exchange is expected to increase the growth rate of cost of goods sold and operating expenses by about 1 percentage point.

b. Others:

- New Commercial Contracts: Healthy growth is expected as the base of expirations increases;

- Microsoft Cloud Gross Margin: Expected to decline year-over-year to around 67% due to continued expansion of AI infrastructure.

- Capital Expenditure: Driven by strong demand, first-quarter capital expenditure is expected to exceed $30 billion; the timing of cloud infrastructure expansion and finance lease deliveries may cause quarterly expenditure fluctuations.

c. Productivity and Business Processes: Revenue is expected to be €32.2-32.5 billion, growing 14%-15% (including a 3 percentage point contribution from foreign exchange).

- M365 Commercial Cloud: Constant currency growth of 13%-14%, with stable business trends and ARPU growth driven by E5 and M365 Copilot.

- M365 Commercial Products: Revenue growth in the mid to high single digits (including Windows Commercial on-premises components and Office transactional purchases, affected by in-period revenue recognition dynamics).

- M365 Consumer Cloud: Revenue growth is expected to be around 20%, driven by price increases in January.

- LinkedIn: Revenue is expected to grow in the high single digits;

- Dynamics 365: Growth of around 10%-20%, with continued growth in all workload revenues.

d. Intelligent Cloud: Revenue is expected to be $30.1-30.4 billion, growing 25%-26% (including a 1 percentage point contribution from foreign exchange), driven by Azure.

- Azure: Constant currency growth of about 37%, driven by strong demand for the service portfolio; year-over-year growth rates may fluctuate quarterly due to capacity delivery and contract mix impacts.

- On-premises Server Business: Revenue is expected to decline in the mid to low single digits as customers shift to cloud products.

e. More Personal Computing: Revenue is expected to be $12.4-12.9 billion.

- Windows OEM and Devices: Revenue is expected to decline in the mid to high single digits, with high inventory levels at the end of the fourth quarter continuing to decline (to a greater extent than normal), leading to a decline in device revenue.

- Search and News Advertising (ex-TAC): Growth of around 15% (down sequentially), driven by search volume and revenue per search from Edge and Bing, with overall growth in the low double digits.

- Gaming: Revenue is expected to decline in the mid to high single digits; Xbox Content and Services: Revenue is expected to decline in the mid-single digits due to a high base in the prior year.

f. Overall Performance Guidance:

- Cost of Goods Sold: $24.3-24.5 billion, growing 21%-22%.

- Operating Expenses: $15.7-15.8 billion, growing 5%-6%.

- Other Income and Expenses: Expected to be -$1.3 billion (mainly due to equity method investment losses, with no recognition of mark-to-market gains or losses).

- Effective Tax Rate: 19%-20%.

3. FY26 Guidance:

a. Revenue and operating profit are expected to maintain double-digit growth, with a slowdown in capital expenditure growth and operating profit margin remaining stable.

b. The proportion of short-term assets will increase. As new capacity, including large finance lease sites, will be delivered in the first half of the year, Capex growth in the first half is expected to be higher than in the second half.

c. Foreign exchange is expected to boost revenue growth by 2 percentage points, increase operating expense growth by 1 percentage point, and maintain an effective tax rate of 19%-20%.

II. Detailed Content of the Earnings Call

2.1 Key Information from Executive Statements

1. Azure Cloud:

a. New data centers have been opened on six continents, with over 400 data centers in 70 regions, leading the industry in number. Over the past 12 months, more than 2GW of installed capacity has been added, expanding faster than competitors. All Azure regions have achieved "AI-first," supporting liquid cooling to enhance fleet replaceability and flexibility.

b. Driving compound S-curves across chip systems and models, enhancing GPU efficiency through software optimization (e.g., increased inference token volume for the GPD 4.0 series models). Launched "Microsoft Sovereign Cloud," covering both public and private cloud deployments to meet customer data residency and sovereignty needs.

c. Nestlé migrated over 200 SAP instances, more than 10,000 servers, and 1.2PB of data to Azure, achieving near-zero business disruption, marking one of the largest and most successful migrations in commercial history.

2. Quantum Computing: Partnered with Atom Computing to deploy the world's first secondary quantum computer, with a long-term investment and technological breakthrough over a decade.

3. AI Platform Ecosystem:

a. Microsoft Fabric: As a comprehensive data and analytics platform for the AI era, covering SQL, NoSQL, and analytics workloads. Revenue grew 55% year-over-year, with over 25,000 customers, making it the fastest-growing database product in the company's history. Fabric 1.0 expanded to all databases and clouds, integrating Power BI semantic models to become the core knowledge base for AI applications.

b. Azure AI Foundry: Helps customers design, customize, and manage AI applications and agents at scale, with built-in tools, management, observability, and trusted AI controls. Supports multi-model deployment, covering OpenAI, DeepSeek, Meta, XAI's Grok, etc., with plans to integrate Black Forest Labs and Mistral AI. Agent services adopted by 14,000 customers, with 80% of the Fortune 500 using them.

4. AI Application Layer:

a. Copilot: Over 100 million monthly active users in commercial and consumer sectors, with over 800 million monthly active users for all product AI features.

- Microsoft 365 Co-Pilot: Launched the largest update, integrating chat, search, and other features, with customer adoption leading, and record quarterly seat purchases (e.g., Barclays deploying for 100,000 employees, UBS expanding to all employees); launched deep reasoning and group-level agents, with hundreds of partners building third-party agent integrations.

- Copilot Studio: Customers created 3 million seats, supporting fine-tuning and customizing agent tone, language, etc., based on company data.

- GitHub Copilot: 20 million users, with a new agent mode; enterprise customers increased by 75% quarter-over-quarter, with 90% of the Fortune 100 using it; driving GitHub AI projects to more than double, with millions of code review agents per month.

- Dragon Copilot recorded over 13 million doctor-patient interactions, up nearly 7 times year-over-year (e.g., Mercy Health System's 1,000+ doctors using it, saving over 100,000 hours).

b. Security and Business Applications: Launched the first security agent, adding over 100 security features, with Microsoft Sentinel integrating over 350 third-party connectors; Entra, Defender, and Purview enhancing AI security protection, with nearly 1.5 million security customers.

c. Dynamics 365 gained market share, with industry customers like Verizon and Domino's.

5. To-C Business:

a. Browser and Windows: Launched Copilot mode and Edge mode, integrating Copilot Composer, Chat, and other features to create the next-generation browser. Copilot will be available on all Windows 11 PCs, providing real-time assistance through Copilot Vision; after Windows 10 support ends, Windows 11 and Copilot Plus PCs will have an advantage due to security and AI value.

b. LinkedIn: With 1.2 billion members, achieving double-digit growth for four consecutive years, with comment volume growing over 30% and video uploads growing over 20%. Continues to integrate AI, introducing agents in recruitment and sales.

c. Xbox and Gaming:

- 500 million monthly active users across all platforms, becoming the largest publisher for Xbox and PlayStation this quarter (e.g., "Forza Horizon 5," "Remnant: From the Ashes" success).

- "Call of Duty: Black Ops 6" reached 50 million players, with total playtime exceeding 2 billion hours; "Minecraft" set records for monthly active users and revenue (thanks to movie success).

- Cloud gaming hours exceeded 500 million, with Game Pass annual revenue approaching $5 billion for the first time, and nearly 40 games in development.

2.2 Q&A Session

Q: What is the best way for software companies to monetize AI through SaaS? Is there a difference in monetization between horizontal general applications like M365 Copilot and targeted functions on the agent side? How should we view the long-term trajectory of SaaS AI profit margins?

A: If we don't limit ourselves to SaaS, the transition from servers to the cloud is essentially an expansion of server usage—previously, the deployment of servers had high barriers and limited market size; cloud services are flexible and have low barriers, leading to exponential growth, and the current AI field is undergoing a similar transformation.

If we agree that "intelligence is the logarithm of computing," the amount of computing will grow, and it needs to be used efficiently to continuously create intelligence. Classic categories such as infrastructure (including data layers, application servers) will remain, but the scale will expand by 1-2 orders of magnitude, and the ratio of storage and computing required per GPU is an indicator of its exponential growth.

In the application layer, SaaS applications are integrating intelligent agents, chat interfaces, and autonomous agents, with other coding tools also adding more GitHub repositories. GitHub's monetization revolves around enterprise editions and features like Copilot, consistent with Microsoft 365 and Dynamics 365, where opening up data layers, business logic layers, and UI layers can increase usage, which is reflected in performance.

The monetization tools for this transformation are similar, including user-tiered models (some linked to consumption) and pure consumption models, which will continue to merge as AI model capabilities improve, and teams will optimize model usage to match work needs, which is the foundation for the future.

Q: This is the second consecutive quarter of significant growth. Can you explain two or three specific catalysts driving customer migration? How do you view the sustainability of this growth trend?

A: 1. Customer migration continues to be healthy, such as Nestlé's SAP instances, large amounts of data, and server migration, which is still in the middle game and far from the end.

2. Cloud-native applications are expanding significantly, including typical cloud-native enterprises, with some customers coming for AI but staying beyond AI business, continuing to stay on Azure.

3. A large number of new AI workloads have been added.

Q: This quarter's performance was strong across the board. Is there anything you originally thought wouldn't happen but actually did? What are your thoughts on this unexpected upside performance?

A: Nothing really surprised us, but when developing AI applications, we noticed that the platform is no longer just a model and an API call. A year ago, this might have been the most advanced technology, but now there are very stateful application patterns that even require a lot of rethinking of the application stack.

For example, the storage layer, its complexity involves how many indexes to build or preprocess, which affects the quality of prompt words or context engineering. Today, products like Azure Search, Fabric, Cosmos DB, and related frameworks are becoming more powerful, supporting the construction of important applications.

What is gratifying is that the learning curve for the technology stack is spreading rapidly both inside and outside the company, and the speed of building applications has also increased significantly. Just like when relational databases appeared, it took some time for people to build ERP systems, and now we are building very complex applications at a very fast pace, based on the maturity of emerging markets.

Q: From your interactions with customers, what is their understanding of "Copilot is just the beginning, and it will expand more broadly in the future"?

A: Even within Copilot, there are already analysts, researchers, and various third-party agents, far beyond "request-response," but generating applications that run autonomously and then return results. But the user interface (UI) is still important, even for asynchronous work—commands, monitoring asynchronous work all require a UI (which may not be a chat interface), and it also needs to be able to check the content of asynchronous work.

For example, with GitHub, even if you don't use GitHub Copilot to create core code submissions or [merge] requests, even if you use other tools to write code, the usage of its code review agent has still increased significantly. All systems are experiencing similar situations.

Therefore, everything starts with a UI focused on chat, but it will soon go beyond this scope, which can be seen in M365, Dynamics 365, and GitHub.

Q: You previously mentioned being able to accelerate Azure growth while slowing capital expenditure, and you have done so. What is the outlook for the future? The next quarter's capital expenditure guidance is a positive signal for the cloud services business, combined with Satya's mention that the AI technology stack consumes more infrastructure. How should we understand the relationship between the capital expenditure curve and Azure growth rate in the coming years? Are we at a stage where we need to continue investing while expecting breakthroughs in inference and applications to form a richer gross margin structure?

A: First, to clarify, the full-year capital expenditure and Q1 guidance of over $30 billion need to be based on our $368 billion contract backlog to be delivered (covering the entire Microsoft Cloud, not just Azure). I am satisfied with the return on investment of the expenditure and its relevance to the business, as these expenditures are directly related to the contract business that needs to be delivered, requiring the team to efficiently deploy capacity. The year-over-year growth rate of capital expenditure will decline, but short-term asset investments in servers, GPUs, etc., are closely related to the backlog and demand curve. In January, we hoped for an improvement in supply and demand by June, and now we hope for better by December—this is not because we are slowing down capital expenditure, even if we accelerate investment, demand is still growing.

I do not focus on predicting the intersection of revenue and capital expenditure growth, but rather on accumulating orders, expanding business, and delivering capacity, with a good return on investment at present. I do not want everyone to focus too much on the turning point, as setting a time point during the expansion phase can be too conservative in terms of market share and business expansion, and my energy is more focused on this.

Another point I mentioned in the last earnings report: the difference between hosting providers and hyperscale service providers lies in software, and this applies here as well. The GP40 example I gave relies entirely on software, even the optimization over the past year is the same. We can use software skills to enhance the performance of any hardware several times, which is where the benefits come from.

But as mentioned earlier, when expanding "factories," progress should not be slowed down, but rather pursued on multiple fronts, which will have a compounding effect over time. When talking about the software layer, it is important to mention its association with the overlapping S-curves. It is important to know that we have seen this situation in previous cloud transformations, and we did the same at that time. Now, the same skills and logic will drive this transformation at a faster pace.

Q: The full-year operating profit margin is expected to remain flat. With the business mix shifting towards Azure and more AI-related products, can you elaborate on how to balance these trade-offs and offset the impact of the mix change? Are there any highlights in terms of productivity improvements brought by internal use of AI? Are there any other factors supporting the full-year expectations?

A: When it comes to profit margins, the focus is not only on cost control but also on launching competitive, innovative, and market-share-grabbing quality products to drive revenue growth—revenue and revenue growth are the sustainable ways to improve profit margins, forming a self-reinforcing cycle.

Secondly, as previously discussed with Kash and mentioned by Satya, we are using all skills to improve efficiency, and the compounding effect of the S-curve will manifest at every layer of the technology stack. We are advancing this work while also expanding, so even with continued investment, efficiency will improve.

Additionally, we need to continuously attract top talent and focus on products and opportunities with large market sizes and high success rates. By doing these three things well, with clear direction and focus, we can ensure that profit margins meet expectations. But the core is always the product and the ability to deliver products to customers—this is our focus.

<End of Text>

Risk Disclosure and Statement of this Article:Dolphin Research Disclaimer and General Disclosure