Microsoft conference call: DeepSeek is a positive factor, AI business annual revenue run rate has now exceeded $13 billion, a year-on-year increase of 175%

On January 29th, Eastern Time, after the US stock market closed, Microsoft announced its financial data for the second quarter of fiscal year 2025 (hereinafter referred to as the fourth quarter) for the year ending December 31, 2024 financial data.

In the fourth quarter, Microsoft's commercial cloud total revenue grew by 21% year-on-year, with analysts expecting growth to remain flat at 22% compared to the third quarter; Azure and other cloud services revenue grew by 31%, also slowing from 33% in the third quarter, with AI products contributing 13 percentage points to Azure's growth, exceeding expectations; capital expenditures nearly doubled.

In the conference call, Microsoft CEO Satya Nadella stated that deepseek indeed has some real innovations, noting that when token prices drop and inference computing prices decrease, it means people can consume more, leading to more applications being developed.

Microsoft CFO Amy Hood revealed that future orders grew by 67%, or 75% when calculated at fixed exchange rates, with most of the expected growth likely coming from OpenAI, which Microsoft continues to host exclusively on Azure. Nadella also informed investors that the current revenue run rate for the company's AI business has exceeded $13 billion, a 175% increase compared to the same period last year.

The following are key points from the conference call:

Satya Nadella (CEO) Insights

• AI Scaling Law: "The AI scaling law continues to compound in terms of computation during pre-training and inference."

• AI Efficiency Improvement: "In inference, due to software optimization, we typically see a cost-performance improvement of over 2 times for each generation of hardware and over 10 times for each generation of models."

• Data Center Expansion: "Over the past three years, our overall data center capacity has more than doubled, and the capacity added last year was more than in any other year in our history."

• AI-Driven Data Growth: "If you look at the underlying technologies of ChatGPT, Copilot, or enterprise AI applications, you will find that as these workloads scale, raw storage, database services, and application platform services are all growing."

• Success of GitHub Copilot: "GitHub Copilot is increasingly becoming the preferred tool for digital natives like ASOS and Spotify, as well as the largest global enterprises like HP, HSBC, and KPMG."

• Widespread Adoption of Copilot: "The number of daily users of Copilot has again more than doubled month-on-month."

• Popularity of Copilot Studio: "Over 160,000 organizations have used Copilot Studio, creating over 400,000 custom agents in total over the past three months, more than doubling month-on-month."

• Combination of Open Source and Closed Source AI Models: "Our support for OpenAI's leading models and the best open-source models and SLM puts us in a favorable position." ”

• Recognition of DeepSeek: “DeepSeek's R1 was released today through Foundry and the model catalog on GitHub, featuring automatic red teaming, content safety integration, and security scanning capabilities.”

Amy Hood (CFO) Insights

• Strong AI business revenue: “As you heard from Satya, our AI business has an annual revenue run rate exceeding $13 billion, which is above expectations.”

• Significant growth in commercial bookings: “Commercial bookings grew by 67% and 75% (at constant currency), significantly exceeding expectations due to OpenAI's commitment to Azure.”

• Growth in Microsoft cloud revenue: “Microsoft cloud revenue was $40.9 billion, growing by 21%.”

• Reasons for gross margin decline: “The gross margin for Microsoft cloud was 70%, in line with expectations, but declined by two percentage points year-over-year due to the scaling of AI infrastructure.”

• Contribution of Copilot: “With M365 Copilot, we continue to see growth in adoption, expansion, and usage, with ARPU growth again driven by E5 and M365 Copilot.”

• Azure AI growth: “Azure's growth includes 13 percentage points from AI services, growing 157% year-over-year, and exceeding expectations, even as demand remains above our available capacity.”

• Adjustments for scaled sales: “In non-AI areas, the real challenge lies in what we call scaled sales. So think about it, these are primarily customers we reach through partners and more indirect sales methods.”

• Capital expenditures: “More than half of our cloud and AI-related spending is on long-term assets that will support monetization over the next 15 years and beyond.”

• Future capital expenditure plans: “In fiscal year 26, we expect to continue investing to respond to strong demand signals, including the backlog of customer contracts we need to deliver across Microsoft cloud. However, the growth rate will be lower than in fiscal year 25, and NYX spending will begin to shift back to short-term assets more related to revenue growth.”

• Outlook for the full fiscal year 2025: “For the full fiscal year, we continue to expect revenue and operating income to achieve double-digit growth as we focus on improving efficiency in sales costs and operating expenses.”

The following is a transcript of the conference call, translated by AI:

Host:

Hello everyone, welcome to Microsoft's Q2 earnings conference call for fiscal year 2025. All participants are currently in listen-only mode. There will be a Q&A session following the formal presentation. (Host's note) A reminder that this meeting will be recorded.

I am now honored to introduce your host, Vice President of Investor Relations Brett Iversen. Please go ahead.

Brett Iversen, Vice President of Investor Relations:

Good afternoon, thank you all for joining today. Joining me on the call are Chairman and CEO Satya Nadella, CFO Amy Hood, Chief Accounting Officer Alice Jolla, and Company Secretary and Vice President of Legal Affairs Keith Dolliver

On Microsoft's investor relations website, you can find our earnings press release and financial summary slides, which are intended to complement the prepared remarks for our meeting today and provide a reconciliation of the differences between GAAP and non-GAAP financial metrics. More detailed outlook slides will be available on Microsoft's investor relations website when we provide outlook comments during today's conference call.

In this conference call, we will discuss certain non-GAAP items. The non-GAAP financial metrics provided should not be viewed as a substitute for or superior to the financial performance metrics prepared in accordance with GAAP. These items are provided as additional explanatory items to help investors further understand the company's performance in the second quarter and the impact of these items and events on financial results.

All year-over-year growth mentioned in today's conference call is related to the same period last year, unless otherwise noted. When available, we will also provide constant currency growth rates as a framework for assessing the performance of the underlying business, excluding the effects of currency fluctuations. When the constant currency growth rate is the same as the actual growth rate, we will only mention the growth rate.

We will post the prepared remarks on the website immediately after the meeting concludes, until a complete transcript is available. Today's conference call is being webcast live and recorded. If you ask a question, your question will be included in the live broadcast, transcript, and any future recordings. You can replay the conference call and view the transcript on Microsoft's investor relations website. During this conference call, we will make forward-looking statements, which are predictions, estimates, or other descriptions of future events. These statements are based on current expectations and assumptions and are subject to risks and uncertainties. Actual results may differ materially due to factors discussed in today's earnings press release, comments made during this conference call, and the risk factors section of our 10-K, 10-Q, and other reports and documents filed with the Securities and Exchange Commission. We do not undertake any obligation to update any forward-looking statements. With that said, I will now turn the time over to Satya.

Satya Nadella, Chairman and CEO:

Thank you, Brett. This quarter, we saw Microsoft Cloud continue to experience strong growth, with revenue surpassing $40 billion for the first time, a year-over-year increase of 21%. Enterprises are moving from the proof-of-concept phase to full enterprise-wide deployment to unlock the full return on investment in AI. Our AI business annual revenue run rate has now exceeded $13 billion, a year-over-year increase of 175%. Before diving into the quarter's details, I want to talk about how we manage our infrastructure fleet and the core principles of how we allocate computing resources.

The laws of AI scaling continue to compound in both pre-training and inference compute. Over the years, we have seen significant efficiency gains in both training and inference. In inference, we typically see price-performance improvements of over 2x with each generation of hardware, while each generation of models can achieve over 10x improvements due to software optimizations. As AI becomes more efficient and accessible, we will see exponential growth in demand

Therefore, just like what we did with the commercial cloud, we focus on continuously expanding our global infrastructure scale and maintaining an appropriate balance between training and inference as well as geographical distribution. From now on, this is a more sustainable cycle driven by revenue growth and capability growth, thanks to the software-driven AI scaling laws and the cumulative effects of Moore's Law. Next, I will introduce the progress we have made at various levels of our technology stack.

Azure is the infrastructure layer for AI. We continue to expand data center capacity based on recent and long-term demand signals. Over the past three years, our overall data center capacity has more than doubled, and last year we added more capacity than in any previous year. Our data centers, networks, racks, and chips are integrated as a complete system, driving new efficiencies to support today's cloud workloads and next-generation AI workloads. We continue to leverage Moore's Law to update our infrastructure, which is reflected in our support for the latest products from AMD, Intel, and NVIDIA, as well as our self-developed innovations in Maya, Cobalt, Boost, and HSM chips.

In terms of cloud migration, we continue to see customers like UBS migrating workloads to Azure. UBS alone has migrated a mainframe workload containing nearly 400 billion records and 2PB of data. For critical Oracle, SAP, and VMware applications of our customers, we remain the preferred cloud platform.

On the data front, we see Microsoft Fabric making breakthrough progress. We now have over 19,000 paying customers, from Hitachi to Johnson Controls to Schaeffler. Fabric is now our fastest-growing analytics product in history. Power BI is also deeply integrated with Fabric, with over 30 million monthly active users, a 40% increase from last year.

In addition to Fabric, we are also seeing new AI-driven data patterns emerge. If you look at the underlying technology of ChatGPT, Copilot, or enterprise AI applications, you will see growth in raw storage, database services, and application platform services as these workloads scale.

The number of Azure OpenAI applications running on Azure databases and Azure application services has more than doubled year-over-year, driving significant adoption growth of SQL Hyperscale and Cosmos DB. Now let's talk about AI platforms and tools. As we shared last week, we are excited about OpenAI's new large-scale commitment to Azure. Through our strategic partnership, we continue to mutually benefit from each other's growth. Since OpenAI's API runs entirely on Azure, customers can rely on us for world-leading models.

OpenAI will soon have more new content, so stay tuned. Azure AI Foundry provides top-notch tools and runtimes to build agents, multi-agent applications, AI operations, and API access to thousands of models. Since its launch two months ago, we have had over 200,000 monthly active users By simultaneously supporting OpenAI's leading models and the best open-source models and SLM selection, we are in a favorable position. DeepSeek's R1 was released today through the model catalog on Foundry and GitHub, including automated red team testing, content safety integration, and security scanning.

Our five series of SLMs have now been downloaded over 20 million times. We also have over 30 models from partners such as Bayer, PageAI, Rockwell Automation, and Siemens, aimed at solving specific industry use cases. As AI evolves, the way we build, deploy, and maintain code is undergoing fundamental changes. GitHub Copilot is increasingly becoming the preferred tool for digital-native enterprises like ASOS and Spotify, as well as some of the world's largest companies like HP, HSBC, and KPMG. We are delighted by the early response to GitHub Copilot in VS Code, with over 1 million registered users in just the first week of its launch.

Overall, GitHub now has 150 million developers, growing by 50% over the past two years.

Now, let's talk about the future of work. Microsoft 365 Copilot is the user interface for artificial intelligence that greatly enhances employee productivity and provides a range of intelligent agents to streamline employee workflows. We see that customers are adopting it faster than ever, whether for large or small orders. We are not only winning new Microsoft 365 Copilot customers, but most of our existing enterprise customers are also coming back to purchase more licensed seats.

Among the customers who purchased during the first quarter of Copilot's launch, their total number of licensed seats has increased more than tenfold over the past 18 months. For example, Novartis has added thousands of licensed seats each quarter over the past year and now has 40,000 seats. Barclays, Carrier Global, Pearson, and the University of Miami all purchased 10,000 or more seats this quarter. Overall, the number of people using Copilot has once again doubled quarter-over-quarter.

Employees are also using Copilot more than ever before. The intensity of use has increased by over 60% quarter-over-quarter, and we have expanded our collaboration with Copilot Chat, which was announced earlier this month. Copilot Chat, along with Copilot Studio, is now available to every employee so they can directly use agents in the workplace. With Copilot Studio, we have made creating agents as simple as creating an Excel spreadsheet.

Over 160,000 organizations have already used Copilot Studio, collectively creating over 400,000 custom agents in the past three months, with a quarter-over-quarter growth of over 2 times. We have also launched our own first-party agents to assist with meetings, project management, resolving common HR and IT queries, and accessing SharePoint data We continue to see partners like Adobe, SAP, ServiceNow, and Workday building their third-party agents and integrating with Copilot. The driving force behind Copilot as an AI user interface and our momentum in agents is our rich data cloud, which is the largest source of organizational knowledge globally.

Every day, billions of emails, documents, and chat records are added, along with millions of Teams meetings and millions of SharePoint sites. This is the enterprise knowledge cloud. It is growing rapidly, with a year-on-year growth of over 25%. More broadly, we see Copilot plus agents disrupting business applications.

We are actively investing in this. With Dynamics 365, we are gaining market share as organizations like Ecolab, Lenovo, RTX, Total Energy, and Vizent shift from traditional vendors to our AI-driven applications. In the healthcare sector, the monthly doctor-patient interactions of DAX Copilot exceed 2 million, with a quarter-on-quarter growth of 54%. It is used by top healthcare providers like Brigham and Women's Hospital, Michigan Medicine, and Vanderbilt University Medical Center to enhance physician productivity.

Speaking of Windows, as support for Windows 10 is coming to an end, we see momentum increasing. Customers are choosing the latest Windows 11 devices because they offer enhanced security and advanced AI capabilities. This holiday season, 15% of high-end laptops in the U.S. are Copilot plus PCs, and we expect that most PCs sold in the coming years will be Copilot plus PCs. We are also seeing more and more developers, from Adobe and CapCut to WhatsApp, building applications that leverage the built-in NPU.

They will soon be able to run DeepSeek's R1 distilled model natively on Copilot plus PCs, along with the vast GPU ecosystem available on Windows. Besides Copilot plus PCs, the most powerful AI workstations for local development are Windows PCs running WSL 2, powered by NVIDIA RTX GPUs. Now, let's talk about security issues. We continue to make progress on our security future plans and put what we've learned into practice, launching over 80 new product features in the past year.

In terms of Security Copilot, private and public sector organizations like the City of Johannesburg, Eastman, Intesa, National Australia Bank, and NTT have been able to improve incident resolution speed by 30%. Data governance is becoming increasingly important, and customers are now using Microsoft Purview to audit over 2 billion Copilot interactions to ensure secure and compliant usage. Now, let's talk about our consumer business, starting with LinkedIn. More and more professionals are having high-value conversations on LinkedIn, with comment volume increasing by 37% year-on-year

Short videos continue to grow on this platform, with the growth rate of video creation being twice that of other post formats. We are also innovating through agencies to help recruiters and small businesses find qualified candidates faster, and our recruitment business has regained market share in subscriptions. LinkedIn Premium achieved over $2 billion in annual revenue for the first time this quarter. Subscription user growth has nearly increased by 50% over the past two years, with nearly 40% of subscribers using our AI features to enhance their profiles. LinkedIn Marketing Solutions remains the leader in B2B advertising.

Now let's talk about search ads and news. We have regained market share in Bing and Edge. Edge has over 30% market share on Windows systems in the U.S. and has gained market share for 15 consecutive quarters.

Our investments to improve ad rates are paying off, as advertisers increasingly view our network as an important platform for optimizing return on investment. Our Copilot consumer application is seeing increased engagement and retention due to its improved speed, unique personality, and innovative features like Copilot Vision. Just today, we made the Think Deeper feature powered by o1 available for free to all Copilot users worldwide. Now let's talk about gaming. We are focused on improving the profitability of the business to prepare for long-term growth, primarily driven by high-margin content and platform services.

We are realizing this plan. "Call of Duty 6" became the best-selling game on Xbox and PlayStation this quarter, with player numbers in its release quarter exceeding any other paid release in the series' history. We also saw enthusiastic reviews for "Indiana Jones and the Great Circle," which has been played by over four million people. We continue to see strong momentum in Xbox Cloud Gaming, with a record 140 million hours streamed this quarter. Overall, Game Pass set a new quarterly revenue record, and its PC subscriber base grew by over 30% as we focus on driving the growth of fully paid subscriptions across all endpoints. In summary, we continue to innovate across the entire tech stack to help our customers succeed in this AI era, and I am excited about the many opportunities ahead. Now, I will hand the microphone over to Amy.

Amy Hood, Executive Vice President and Chief Financial Officer:

Thank you, Satya, and good afternoon, everyone.

This quarter, our revenue was $69.6 billion, an increase of 12%. Gross profit grew by 13%, or 12% at constant currency, while operating income grew by 17%, or 16% at constant currency. Our earnings per share were $3.23, an increase of 10%. We achieved another quarter of double-digit top-line and bottom-line growth.

These results were driven by strong demand for our cloud and artificial intelligence products and services, while we also improved our operational leverage through better-than-expected operating profit growth. As you heard from Satya, our AI business has an annual revenue run rate exceeding $13 billion, which has surpassed expectations. Commercial bookings grew by 67%, and by 75% at constant currency, far exceeding expectations, primarily due to the Azure commitment from OpenAI. In our core annuity sales actions, the number of contracts over $100 million for Azure and Microsoft 365 both increased. The commercial remaining performance obligations rose to $298 billion, an increase of 34%, and 36% at constant currency.

About 40% will be recognized as revenue in the next 12 months, representing a year-on-year growth of 21%. The remaining portion will be recognized after the next 12 months, growing by 45%. In this quarter, our annuity mix ratio was 97%. Foreign exchange had no significant impact on our results and was generally in line with our expectations for total company revenue, more personal computing revenue, total company cost of sales, and operating expenses. The impact of foreign exchange on revenue in our commercial segment exceeded expectations. Microsoft Cloud revenue was $40.9 billion, a growth of 21%. Microsoft Cloud gross margin was 70%, in line with expectations, and decreased by two percentage points year-on-year, primarily due to our expansion of AI infrastructure. The company's gross margin increased by one percentage point year-on-year to 69%, mainly due to a shift in the sales mix towards higher-margin businesses, as well as improvements in gaming and search, partially offsetting the impact of expanding our AI infrastructure. Operating expenses grew by 5%, below expectations. Operating margin increased by two percentage points year-on-year to 45%. The better-than-expected margin expansion was achieved through efficiencies realized in our business while we invested in expanding AI infrastructure and building AI applications. At the company level, as of the end of December, the number of employees grew by 2% compared to the same period last year and remained essentially unchanged from the previous quarter.

Now, let's take a look at our segment results. Revenue from the Productivity and Business Processes segment was $29.4 billion, a growth of 14%, and 13% at constant currency, despite the unfavorable foreign exchange impact mentioned earlier. The results exceeded expectations, primarily due to Microsoft 365 Commercial. Microsoft 365 Commercial Cloud revenue grew by 16%, and by 15% at constant currency, slightly above expectations, mainly due to the better-than-expected performance of E5 and Microsoft 365 Copilot. For M365 Copilot, we continue to see growth in adoption, expansion, and usage. The growth in average revenue per user (ARPU) was again driven by E5 and M365 Copilot. Paid M365 Commercial seats grew by 7% year-on-year, primarily due to our expanded install base across all customer segments, although mainly in our small and medium-sized businesses and frontline worker products. M365 Commercial product revenue grew by 13%, mainly due to better-than-expected transactional purchases at the launch of Office 2024, as well as the better-than-expected performance of the M365 suite M365 consumer cloud revenue grew by 8%, slightly above expectations.

We saw continued growth in M365 consumer subscriptions, which increased by 10% to reach $863 million, while transitioning to the mixed version of M365 Basic. LinkedIn revenue grew by 9%, mainly due to sustained growth across all business lines. In our talent solutions business, the results were slightly below expectations, primarily due to a continued weak hiring market in key verticals. Dynamics 365 revenue grew by 19%, with a growth of 18% at constant currency, slightly above expectations, mainly driven by growth across all workloads.

Segment gross profit grew by 13%, with a growth of 12% at constant currency, while the gross margin declined year-on-year, mainly due to our expansion of AI infrastructure. Operating expenses grew by 6%, while operating profit grew by 16%, with a growth of 15% at constant currency. Next is the Intelligent Cloud segment. Revenue was $25.5 billion, growing by 19%, with foreign exchange impacts being more adverse than expected.

Excluding foreign exchange impacts, the results for Azure non-AI services, on-premises servers, and enterprise and partner services were slightly below expectations, partially offset by the better-than-expected performance of Azure AI services. Azure and other cloud services revenue grew by 31%. Azure's growth included a 13 percentage point contribution from AI services, which grew by 157% year-on-year and exceeded expectations, although demand continues to exceed our available capacity. Our non-AI services growth was slightly below expectations, mainly due to challenges in market execution, particularly for customers we primarily engage through our scale actions.

As we balance driving non-AI consumption and AI growth. In our on-premises server business, revenue declined by 3%, slightly below expectations, mainly due to slower-than-expected purchasing around the launch of Windows Server 2025. Enterprise and partner services revenue declined by 1%, below expectations, mainly due to underperformance in enterprise support services and industry solutions. Segment gross profit grew by 12%, with a growth of 13% at constant currency, while the gross margin declined by four percentage points year-on-year, mainly due to our expansion of AI infrastructure.

Operating expenses grew by 10%, while operating profit grew by 14%. Now let's look at the more personal computing segment. Revenue was $14.7 billion, essentially flat compared to the same period last year, mainly benefiting from Windows OEM pre-installations, third-party partnerships in search, and the launch performance of Call of Duty in gaming. Windows OEM and device revenue grew by 4% year-on-year, exceeding expectations, mainly due to commercial inventory building in response to the end of Windows 10 support and uncertainties surrounding tariffs. Search and news advertising revenue (excluding technology costs) grew by 21%, with a growth of 20% at constant currency, exceeding expectations, mainly due to the use of third-party partnerships. Growth continues to be driven by the rate expansion of Edge and Bing and healthy growth in advertising volume

In terms of gaming, revenue declined by 7%, or 8% when adjusted for constant currency, as growth in content and services continued to be offset by a decline in hardware. Xbox content and services revenue grew by 2%, exceeding expectations, primarily due to the better-than-expected performance of Blizzard and Activision content, including Call of Duty. The department's gross profit increased by 13%, or 12% when adjusted for constant currency. The gross profit margin increased by six percentage points year-over-year, mainly due to a shift in the sales mix towards higher-margin businesses and strong margin improvement execution in gaming and search. Operating expenses decreased by 1%. Operating profit grew by 32%, or 30% when adjusted for constant currency, primarily due to continued prioritization of high-margin opportunities. Now, back to the overall company results. Capital expenditures, including finance leases, were $22.6 billion, in line with expectations, while cash expenditures for PP&E were $15.8 billion.

More than half of our cloud and AI-related spending is on long-term assets that will be monetized over the next 15 years or more. The remaining cloud and AI spending is primarily on servers, including CPUs and GPUs, to serve customers based on demand signals, including our customer contract backlog. Operating cash flow was $22.3 billion, an 18% increase, primarily driven by strong cloud billing and collections, partially offset by higher payments to suppliers, employees, and taxes. Free cash flow was $6.5 billion, a 29% year-over-year decline, reflecting the capital expenditures mentioned earlier. In this quarter, other income and expenses were negative $2.3 billion, lower than our guidance in October, primarily due to impairment charges on our cruise investments. Our effective tax rate was slightly below expectations at 18%. Finally, we returned $9.7 billion to shareholders through dividends and stock buybacks. Now, let's look at our outlook for the third quarter, unless otherwise noted, all figures are in U.S. dollars.

First, foreign exchange. Since October, with the strengthening of the dollar, we now expect foreign exchange to reduce total revenue growth by two percentage points. Across various segments, we expect foreign exchange to reduce revenue growth in Productivity and Business Processes and Intelligent Cloud by two percentage points, and by about one percentage point in More Personal Computing. Compared to our guidance assumptions for the foreign exchange impact in the third quarter in October, this will reduce total revenue by about $1 billion. We expect foreign exchange to reduce the growth of cost of sales by about two percentage points and operating expenses by about one percentage point.

Many trends in our outlook continue the trends we saw in the second quarter into the third quarter. Demand for our differentiated cloud and AI products and services in Microsoft Cloud should drive another strong growth quarter. In commercial bookings, we expect year-over-year growth to be roughly flat due to a relatively flat expiration base and strong performance in large Azure contracts last year. We expect to maintain consistent execution in our core annuity sales actions and our long-term commitment to our platform.

A reminder that the timing of large long-term Azure contracts can be unpredictable, which may lead to quarterly fluctuations in our bookings growth rate. Microsoft Cloud's gross profit margin should be around 69%, down year-over-year, primarily due to the impact of expanding our AI infrastructure Next, we have departmental guidance. In terms of productivity and business processes, we expect revenue to grow at a fixed exchange rate between 11% and 12%, or between $29.4 billion and $29.7 billion.

Microsoft 365 commercial cloud revenue growth should be between 14% and 15% at a fixed exchange rate, relatively stable compared to our better-than-expected second-quarter results. We expect to continue achieving ARPU growth through E5 and M365 Copilot, and we again anticipate that seat growth will slow due to the scale of the installed base. For M365 commercial products, we expect revenue to be roughly flat compared to the same period last year. Just a reminder, M365 commercial products include the Windows commercial on-premises components of the M365 suite.

Therefore, our quarterly revenue growth may vary due to the contract mix, primarily depending on revenue recognition during the period. M365 consumer cloud revenue growth should be in the mid to high single digits, driven by M365 subscriptions. For LinkedIn, we expect revenue growth to be in the low to mid-single digits, although we anticipate growth across all businesses, the trends from the second quarter and talent solutions will continue to pose headwinds to growth in the third quarter. In terms of Dynamics 365, we expect revenue growth to be in the mid-teens, primarily driven by continued growth across all workloads.

For Intelligent Cloud, we expect revenue to grow at a fixed exchange rate between 19% and 20%, or between $25.9 billion and $26.2 billion. Revenue will continue to be driven by Azure, and just a reminder, Azure may experience quarterly fluctuations due to the contract mix, primarily depending on revenue recognition during the period. We expect third-quarter revenue growth to be between 31% and 32% at a fixed exchange rate, primarily due to strong demand for our service portfolio. As we shared in October, with more AI capacity coming online, the contribution from our AI services will increase.

In terms of non-AI services, healthy growth continues. Although we expect to continue facing the execution challenges mentioned earlier in the second half of the year. While we anticipate being constrained by AI capacity in the third quarter, by the end of this fiscal year, we should be able to roughly meet recent demand, primarily due to our significant capital investments. In our on-premises server business, we expect revenue to decline in the mid-single digits, primarily due to a decrease in transactional purchases.

In enterprise and partner services, we expect revenue growth to be in the low to mid-single digits. In more personal computing, we expect revenue to be between $12.4 billion and $12.8 billion, continuing to prioritize high-margin opportunities. Windows OEM and device revenue should decline in the mid-single digits. We expect Windows OEM revenue to be roughly flat compared to the same period last year, as our outlook assumes inventory levels will normalize. Actual results may vary due to current tariff uncertainties. Device revenue will decline. Search and news advertising (excluding technology costs) revenue growth should be in the mid-teens, primarily driven by continued growth in ad volume and revenue per search for Edge and Bing, gaining market share across all areas We expect growth to slow down from the previous quarter, mainly due to additional foreign exchange impacts and the normalization of the use of third-party partnerships mentioned earlier. Excluding technology costs, search growth will be higher than the overall growth in search and news advertising revenue, and we expect overall search and news advertising revenue growth to be in the mid to high single digits. In terms of gaming, we expect revenue growth to be in the low single digits. We anticipate Xbox content services revenue growth to be in the low to mid single digits, primarily driven by first-party content and Xbox Game Pass. Hardware revenue is expected to decline year-over-year. Now back to company guidance. We expect Cox to grow at a fixed exchange rate of 19% to 20%, or between $21.65 billion and $21.85 billion, with operating expenses growing at a fixed exchange rate of 5% to 6%, or between $16.4 billion and $16.5 billion. Other income and expenses are expected to be approximately negative $1 billion, primarily driven by equity method investments. Just a reminder, we do not confirm gains or losses on equity method investments at market value. Finally, we expect an effective tax rate of about 18% for the third quarter. Now, let's look at some additional thoughts for the remainder of this fiscal year and beyond. First, foreign exchange. Since October, with the strengthening of the dollar, we now expect foreign exchange to reduce revenue and cost of sales growth by more than one percentage point in the fourth quarter, and to reduce operating expense growth by about one percentage point.

Next, capital expenditures. We expect quarterly spending in the third and fourth quarters to remain at similar levels to our spending in the second quarter. In fiscal year 2026, we expect to continue investing based on strong demand signals, including the backlog of customer contracts we need to fulfill, covering our entire Microsoft Cloud. However, the growth rate will be lower than in fiscal year 2025, and our spending will begin to shift back toward short-term assets that are more closely correlated with revenue growth. Just a reminder, our long-term infrastructure investments are interchangeable, allowing us to meet customer demand globally, including AI workloads, while remaining agile. As always, there may be quarterly spending variances due to the timing of cloud infrastructure builds and financing lease deliveries. For the entire fiscal year, we still expect revenue and operating profit to achieve double-digit growth as we focus on achieving efficiencies in cost of sales and operating expenses. Given the operating leverage we have achieved throughout the year, including efficiencies in expanding AI infrastructure and utilizing our own AI solutions, we now expect operating profit margins for fiscal year 2025 to increase slightly year-over-year. Finally, we expect the effective tax rate for the full fiscal year 2025 to be between 18% and 19%.

To summarize, we are committed to providing real-world AI solutions that help customers achieve growth and improve their outcomes. We are confident in our leadership position as our customers grow. Before we enter the Q&A session, I have a special thank you. Brett Iversen will transition to a new role as the head of our Americas sales finance team

On behalf of the company, I would like to thank you for your tremendous impact on investor relations over the past four years, as well as your collaboration with Satya and me. I would also like to welcome Jonathan Neilson, who was previously the financial head of our security products, and he will be returning to the investor relations department to lead this team. Now, let's move into the Q&A session, Brett.

Brett Iversen, Vice President of Investor Relations:

Thank you, Amy.

We will now move into the Q&A session. Out of respect for others on the call, we ask participants to ask only one question. Operator, could you please repeat your instructions?

Operator:

Q&A session.

Thank you for your questions. Our first question comes from Keith Weiss at Morgan Stanley. Please go ahead.

Keith Weiss:

Thank you for taking my question. I would also like to echo Amy's comments, Brett, congratulations on your new role. It has been a pleasure working with you, and I wish you all the best in your new position. Looking at this quarter's performance, commercial bookings have been another very strong quarter.

However, once again, we are slightly disappointed with Azure, which is at the low end of the guidance range. Amy, I hope you can delve into what some of the execution issues might be and what the solutions to those issues are. Are we still confident about accelerating in the second half after the June quarter and last quarter? Thank you very much.

Amy Hood, Executive Vice President and Chief Financial Officer: Thank you, Keith.

Let me take some time to talk about what we saw in the second quarter and provide you with some additional context regarding the execution issues we mentioned in the short term. First, let me be very specific. These issues arose in the non-AI ACR segment. Our Azure AI results were better than we expected, primarily due to some excellent work by the operations team, who moved up some delivery dates, even by a few weeks.

When you are in a capacity-constrained situation, a few weeks can make a significant difference, and the team's excellent execution is reflected in the revenue results. In the non-AI area, the real challenge lies in what we call scaling actions. So think primarily about these customers, whom we reach through partners and more indirect sales methods. The real issue is that when these customers are approached in this way, they are trying to balance how to handle AI workloads while continuing with migrations and other foundational work. Then we changed our sales actions in the summer, really trying to balance both.

When you do this, you learn with customers and partners how to find a balance in investments, marketing funding, and critical human resources to help customers make these transitions. I believe we will make some adjustments to ensure we achieve that balance because when you make these changes in the summer, you can see the impact and whether the balance has been achieved as those changes take effect in the system. So the team is working hard to address these issues, and they are already making adjustments

I expect we will see some impacts in the second half of the year because when you deal with scale actions, it takes some time to adjust. I am satisfied with the team's understanding and addressing these issues. So I hope this information is helpful to you. Then we talked a bit about the situation in the third quarter. We discussed 31 to 32, and we announced 31 this quarter. We are satisfied with the AI results, and we talked about our ability to generate revenue. So again, in the third quarter, we are in a fairly constrained capacity position, which is no different from what I expected when I spoke with you last October. When I talk about being in a capacity-constrained state, it requires two things. You must have space, which I usually refer to as long-term assets, right? That is infrastructure and land, and then you must have equipment. We continue to do this, and you have seen our spending direction change; we have been investing long-term. We have been lacking in power and space. So as the investments we have made over the past three years come to fruition, we will be close to that balance by the end of this year.

So confidence in AI remains, in terms of sales, utilization, and being, I think, the ability inspired by signals. What we are seeing is waiting to see how non-AI ACR plays out in the second half through scale actions. But overall, the only real change is the scale actions, from my perspective. Keith, I hope this is helpful to you.

Unidentified Speaker:

Yes, I think, Amy, I just want to add one thing, Keith, regarding your question. As Amy mentioned, the AI growth rate is actually better than we expected, and we have dealt with some supply issues. More importantly, some workloads are scaling well. When you dive into these AI workloads, another good aspect is that even ordinary storage, data services, application services, etc., as we mentioned.

So whether it's ChatGPT or Copilot, or even the emerging AI workloads in enterprises, this is all good. Enterprise workloads, whether it's SAP or VMware migrations, are also in good shape. It's just a subtle issue, as Amy mentioned, about how to really adjust incentives during the platform transition period. You really need to ensure that you are inclined towards new design victories. You are not just doing things from the last generation. This is the art that Amy mentioned, to ensure you find the right balance.

But let me put it this way. You would rather win new victories than just protect past achievements. This is definitely something we will always lean towards.

Unidentified Speaker:

Thank you, Keith.

Operator, the next question.

Operator:

The next question comes from Mark Moerdler of Bernstein Research. Please go ahead.

Mark Moerdler:

Thank you very much for taking my question.

Can you provide more details on the factors driving Microsoft's AI revenue growth that far exceeded expectations? We have talked about the Azure AI segment. But can you provide more details on this aspect? Our estimate is that the scale of Copilot is much larger than we expected, and its growth rate is also faster. Any additional information regarding Microsoft's AI exceeding expectations would be very helpful. Thank you.

Amy Hood, Executive Vice President and Chief Financial Officer: Thank you, Mark, your question is valuable. Yes, as we discussed, the results are better than expected. There are several aspects, as you rightly pointed out. The first is the Azure segment, which we just talked about.

The second part, you are correct, the performance of Microsoft Copilot has also exceeded expectations. Importantly, we have seen strong performance in terms of seats, whether new seats or expanded seats, as Satya mentioned, and usage, which, while it does not directly impact revenue, certainly does so indirectly, as people derive more value from it. Additionally, the price per seat is actually quite good, indicating value. So Mark, these are the main parts that exceeded our expectations.

Analyst:

Thank you, Mark.

Brett Iversen, Vice President of Investor Relations:

Operator, the next question.

Operator:

The next question comes from Brent Thill of Jefferies. Please go ahead.

Brent Thill:

Thank you, Satya. You mentioned DeepSeek several times in your opening remarks. I think everyone would be interested in your thoughts. Are we seeing AI scale at a lower cost? Have we reached a milestone where you can see this, or do we still need some time to get there? Thank you for your thoughts.

Satya Nadella, Chairman and Chief Executive Officer: Yes.

Thank you, Brent. Yes, in my remarks, I talked about the development of AI, which is no different from the development of conventional computing cycles. It's always about bending the curve and then adding more points on that curve. So there is Moore's Law running at super speed.

Then, on top of that, there is the AI scaling law, whether it's pre-training or inference time computation, which compounds, and these are all software. You should understand that what I mentioned in my remarks is that we have observed for some time that the improvements in each cycle are tenfold, simply because of all the software optimizations in inference. So that's what you're seeing. And on top of that, I think DeepSeek indeed has some real innovations.

These are even some things that OpenAI discovered in o1. So, of course, all of this will now be commoditized and widely used. And the biggest beneficiaries in any software cycle are the customers, right? Because ultimately, if you think about it, the shift from client-server to cloud computing means more people are buying servers; it's just called cloud computing now. So when token prices drop and inference computing prices drop, it means people can consume more, and there will be more applications being written

Interestingly, when I mention these quite powerful models, it's hard to imagine that we are at the beginning of 2025, and you can run a model that requires quite a large-scale cloud infrastructure on a PC. So this optimization means that AI will become more ubiquitous, and this is good news for large-scale cloud service providers like us and PC platform providers like us.

Analyst:

Thank you.

Unidentified Speaker:

Thank you, Brett.

Operator, the next question.

Operator:

The next question comes from Karl Keirstead of UBS. Please go ahead.

Karl Keirstead:

Thank you. Perhaps this question can also be answered by Satya. It's not about numbers, but Satya, I want to ask you about the news regarding Stargate and the changes in the relationship with OpenAI announced last week.

It seems that most investors interpret this as Microsoft certainly still being very committed to the success of OpenAI, but choosing to play a more supporting role in terms of future capital expenditure needs for OpenAI's training. I hope you can elaborate on Stargate and Amy, whether investors should draw any conclusions about your thoughts on capital expenditure needs in the coming years from this decision. Thank you.

Satya Nadella, Chairman and CEO:

Yes, thank you for your question.

We are still very satisfied with our partnership with OpenAI, and as you can see, they have committed to using Azure in a significant way, and even in the bookings, what we have confirmed is just the first payment, so you will see that, given the role we play in this, there will be more benefits in the future. Of course, their success is our success, as all the other business arrangements we have detailed in the blog are consistent with this announcement. But regarding the overall question you mentioned, I want to say that we are building a fairly flexible fleet, right? We are ensuring the right balance between training and inference. It is geographically distributed.

We are working very hard on all the software optimizations, right? I mean, not just some software optimizations done by DeepSeek, but all the work we have done, such as working with OpenAI over the years to reduce the cost of GPT models. In fact, we have done a lot of work on inference optimization, which has always been a key driver. A key point in AI is that you not only launch cutting-edge models, but if they are too expensive to serve, that’s not good, right? It won’t generate any demand. So you need to do this optimization to reduce inference costs and make it widely usable. So that’s the physics of the fleet we are managing. And remember, you don’t want to buy too much at once because Moore's Law gives you a 2x performance improvement every year, and your optimization gives you a 10x performance improvement. You want to continuously upgrade the fleet, modernize the fleet, refresh the fleet, and ultimately have the right monetization ratio with what you think the training expenditure should be So, I am very satisfied with the investments we are making; they are flexible and allow us to better scale the business in the long term.

Amy Hood, Executive Vice President and Chief Financial Officer:

Perhaps, Karl, let me reiterate some of my comments on capital expenditures, as I think it helps further elaborate on what Satya mentioned about what a flexible fleet means. We have, I believe we have discussed, close to $300 billion in RPO. This is the committed contracts that need to be delivered to customers, and the faster and more efficiently we complete these deliveries, the better we will be—not just in the OpenAI partnership, which is just part of it, but for our entire platform that we need to deliver for customers. I think one point that is sometimes overlooked is that when we talk about flexibility, we are not just referring to primary use cases; we have been talking about inference, and then there is post-training, which is a key component, and then there is running the commercial cloud, which is necessary at every layer of modern AI applications.

It needs to be distributed and it needs to be global. All of this is important because it means you are the most efficient. So the capital expenditure investments you see, as you mentioned, at the front end are this infrastructure build-out that allows us to really catch up on the AI infrastructure we need, but it can also be seen as the buildings themselves, the data centers, and some of the things we need to catch up on in the commercial cloud.

Then you will see a shift towards more CPUs and GPUs. This shift will be more directly related to revenue and will be related to the OpenAI partnership or other partnerships you asked about. So what I want everyone to understand is that capital expenditure growth is undergoing this cyclical shift that is closely tied to customer contract deliveries, regardless of who the end customer is.

Karl Keirstead:

Thank you.

Brett Iversen, Vice President of Investor Relations:

Operator, the next question.

Operator:

The next question comes from Brad Zelnick of Deutsche Bank. Please go ahead.

Brad Zelnick:

Thank you very much for taking my question, and I also want to congratulate and thank Brett.

Satya, as we consider Microsoft's rich Copilot product portfolio, which has been in the market for over a year now, with product accuracy continuously improving and inference costs declining, how do you view the journey from here and perhaps the ability to bring products to market to meet the broadest customer and client needs? Thank you.

Satya Nadella, Chairman and Chief Executive Officer: Thank you, Brad, that is a valuable question. In fact, you have recently seen two announcements we made. One is regarding M365 Copilot.

We now have Copilot Chat. So this will effectively be widely deployed across the installed base, as you can enable it immediately through IT, and everyone can start using the web-based chat with all enterprise controls right away It has built-in Copilot Studio, which means they can start building agents. So we think this is a great combination with the full Copilot, and I believe this will actually accelerate aspects like seat usage and agent building.

So that's one aspect. The other aspect is that even on the consumer side, we just launched o1 yesterday, which now powers the "deep thinking" feature on Copilot. It has now been rolled out globally, right? So you can see the benefits of reasoning optimization, and the reduction in costs means you can drive broader adoption of these features that were once high-level functionalities. We will definitely do this across the entire product portfolio.

The same will happen in GitHub Copilot and in Security Copilot as well. So you will see us doing this across the length and breadth of our product portfolio.

Brad Zelnick:

Thank you.

Unidentified Speaker:

Thank you, Brad.

Brett Iversen, Vice President of Investor Relations:

Operator, next question.

Operator:

The next question comes from Brad Reback of Stifel. Please go ahead.

Brad Reback:

Great.

Thank you very much. Satya, if you look a few years out, is there any way to know how much of the reasoning done on Azure will use proprietary models versus how much will use open-source models? That said, does this ultimately matter to Microsoft? Thank you.

Satya Nadella, Chairman and CEO: Yes, that's a great question. Because to some extent, what you see is a large number of models being used in any application, right? When you dive into products like Copilot or GitHub Copilot, you have already seen many different models.

You build models, you fine-tune models, you distill models. Some of these models are distilled into open-source models. So there will be such a combination. So we have always believed that having cutting-edge models is always good.

You always want to build your applications with the highest ambitions using the best models available at the time, and then optimize from there. So that's another aspect. There is a sense of timeliness, right? The given COGS profile you start with doesn't have to be final, because you can continuously optimize latency and COGS and introduce different models. And in fact, all this complexity, by the way, needs to be managed by a new application service, a new application server.

So we are making significant investments in Foundry because from the perspective of application developers, you want to keep up with the tide of these emerging models. You want to have a way to keep your applications evergreen to benefit from all these innovations without bearing all the development costs, or what people call DevOps costs, or what people call AIOps costs So we are also heavily investing in application servers for any workload, so that we can benefit from these different models, whether they are open-source, closed-source, or of different weight classes. At the same time, from an operational perspective, it is faster and easier.

Analyst:

Great, thank you.

Brett Iversen, Vice President of Investor Relations:

Thank you, Brad. Operator, the next question.

Operator:

The next question comes from Brad Filz of Bank of America. Please go ahead.

Analyst:

Oh, great, thank you very much. I'm glad to hear about the strong performance of Copilot. I would love to understand the source of this strong performance.

Is it department-level deals, with customers moving from proof of concept to department-level deals, or are there multiple department-level deals within enterprises? You mentioned some rising usage trends. I would love to understand the common use cases you are seeing that give you confidence it will translate into future monetization. Thank you.

Satya Nadella, Chairman and CEO: Yes, I think the initial batch of seats was in places where there was more belief in immediate productivity, like sales teams, finance departments, or supply chain departments, where there is a lot of, for example, SharePoint data that you want to combine with web data to produce beneficial results.

But what happened next is very much like what we saw in these previous generations of productivity products, where people collaborate across functions and roles, right? For example, even in my daily habits, I use chat, I use work tags, I get results, and then I immediately use pages to share with colleagues. I somewhat call it thinking with AI and working with people. This pattern then requires you to scale it more broadly in the enterprise. You can achieve this through teams and various means. What we have done to make it easier is to start with adding Copilot Chat. So this gives enterprise customers more flexibility to have something more universal.

Analyst:

Great, thank you very much.

Brett Iversen, Vice President of Investor Relations:

Thank you, Brad. Operator, we have time for one last question.

Operator:

The last question will come from Brent Bracelin of Piper Sandler.

Please go ahead.

Brent Bracelin:

Thank you for taking my question. Good afternoon. I want to go back to the commercial bookings and commercial RPO question; I believe it increased by $39 billion, which is the highest level we have seen on a quarterly basis.

Commercial bookings grew by 75%, at constant currency. This is double what we have seen over the past decade. I know this metric has some volatility, but this quarter seems to be related to the momentum of backlog and bookings. Can you talk about the breadth of this growth? Is it widespread? Or are there a few large deals? Any details would be very helpful. Thank you

Amy Hood, Executive Vice President and Chief Financial Officer: Thank you, Brent. That's a great question. We've talked about one of the main drivers, which is the Azure commitment made by OpenAI.

What I want to say is that while this is obviously a significant component, you will continue to see OpenAI making commitments. So I don't want to separate this concept from a one-time thing; this is a sustained relationship that will continue to grow, and it absolutely will. Regarding your question, what other factors are involved in this number? First, our core actions are performing very well. Our core actions include, as you mentioned, the renewals of our existing contracts, the additions to those contracts, and upselling, such as Copilot or GitHub Copilot or other processes, which I think is important.

We also had a great E5 quarter, and when we talk about a lot of the M365 Copilot content, I sometimes forget to also talk about the momentum of the suite. We saw that this quarter as well, and we are very pleased with it. The last component is the large Azure commitments. As you mentioned, these look just as we expected; this is a good signal of platform commitment. As you said, this is a very broad number. These Azure commitments come in two forms. One is existing customers who have fulfilled their commitments and are making larger commitments, which is a good signal of platform commitment. Then you have new customers making commitments. We saw that this quarter as well. So as you said, sometimes a large number like this can make you feel like it's just one thing. I think you're right. Part of it is one thing, but a lot of it is our good consistent execution on workloads.

Analyst:

Thank you.

Satya Nadella, Chairman and Chief Executive Officer:

Thank you, Brent. That concludes the Q&A session of our earnings call today. Thank you all for joining us today.

We look forward to speaking with all of you again soon.

Amy Hood, Executive Vice President and Chief Financial Officer:

Thank you, everyone.

Operator:

The meeting has now concluded. You may disconnect your lines. Have a great day.

Event has ended