The Zhitong Finance App learned that Microsoft (MSFT.US) announced financial results for the first quarter of the 2027 fiscal year on Wednesday. According to the data, the company's revenue for the quarter reached US$92.279 billion, up 19% year on year; GAAP net profit was US$36.629 billion, up 32% from US$27.747 billion in the same period last year; GAAP diluted earnings per share were US$4.92, up from US$3.72 in the same period last year, up 32% year on year.
Non-GAAP adjusted net profit reached $36.422 billion, up 18% year over year; adjusted earnings per share were $4.89, up from $4.13 in the same period last year, up 18% year over year. Microsoft said that non-GAAP results excluded the impact on OpenAI investments.
In terms of profitability, Microsoft's operating profit for the first fiscal quarter reached 45.915 billion US dollars, up 21% year on year and 22% at a fixed exchange rate; the operating profit margin was about 49.8%, up from 48.9% in the same period last year. The gross profit for the quarter was US$62,559 million, up about 17% year on year, and gross margin was about 67.8%.
The following is the full text of the Microsoft performance conference call (artificial intelligence tool assisted translation):
I. Executive speeches
Satya Nadella, Chairman and CEO
The fiscal year started off quickly. Microsoft's cloud revenue surpassed $63 billion, and the growth rate accelerated to 30% over the same period last year. The company focuses on three things: designing AI that empowers humans, ensuring that each organization has sovereignty over its own learning, and building trust through controls that ensure the safety and security of smart devices.
Next, let's talk about how companies are building new technology stacks around these goals, starting with Copilot. This quarter, the company released its biggest Copilot update to date, integrating chat, collaboration, autopilot (autopilot), and code into a unified experience. This is an integrated product experience covering business and consumer, with a full suite of Office built in. The company sees Copilot as a new operating system for work.
Customers are increasingly using Copilot to get real work done. Automated task execution has nearly doubled since July; daily active users increased by nearly 50% month-on-month and more than fivefold; user satisfaction increased by nearly 20% month-on-month, reaching a record high; 90% of Copilot queries use an automatic (auto) mode to dynamically route tasks to different models to optimize between quality, speed, and cost. These improvements are being translated into larger purchases. The number of customers with more than 50,000 Copilot seats increased more than eightfold year over year, including Capgemini (Capgemini), HCL Tech, Hitachi Group (Hitachi Group), Truist Financial, and the UK Revenue and Customs Department (UK Revenue and Customs Department).
Just last week, Volkswagen (Volkswagen) announced plans to deploy Copilot to more than 330,000 employees, one of the largest deployments in Europe to date. The E7 kit (integrating Copilot E5, Entra and Agent 365) has strong momentum and has been purchased by thousands of customers in various industries so far, including Amadeus, BlackRock (BlackRock), Fiserv, and the NSW Department of Health. The company's business model is expanding from a subscription per seat to an inclusive consumption model, increasing total addressable market (TAM) and average revenue per user (ARPU). Subscriptions bring increasingly powerful everyday intelligence to every user, and are usually funded by IT departments to integrate and extend cutting-edge capabilities to the entire organization; when agents take on higher-value work, the consumption model opens up the demand for funding from business units, and expenses are directly linked to the value delivered by these intelligence devices.
The number of Copilot customers actively spending credits through pay-as-you-go billing increased by more than 75% this quarter. In the field of software development, all major coding agents have landed on GitHub. In the past six months, pull requests (PR) generated by agents have increased 6 times. Currently, more than one-third of PR on the platform are initiated by agents. The company is focusing on improving reliability and coping with the scaling up of intelligent activity on GitHub. At the GitHub Universe conference this morning, the company showed how to expand computing power and restructure the core Git architecture for the wave of smart devices; at the same time, it announced the full launch of Copilot Fusion, introducing multi-model orchestration into software development, which can accelerate by up to 40% and reduce costs by 77%. Since pay-per-use billing came into effect, credit consumption for smart features on GitHub has more than doubled. The company is also helping developers ensure code security and transform AI-generated code into high-quality software.
GitHub Advanced Security's revenue grew by more than 50% year over year, and customers used its AI capabilities to scan the entire code base and fix bugs in real time. In the field of commercial applications, the company continues the vision of “Copilot as a new operating system for work” and directly incorporates the business logic and capabilities of Dynamics 365 into Copilot to help each organization build its own “infinite SaaS factory”. This quarter, the company launched 30 new Copilot skills in Dynamics 365, enabling customers to directly invoke CRM and ERP capabilities in their workflows.
In the field of recruitment, over 10 million LinkedIn members use AI-driven job searches every day, and 25,000 customers use the company's smart recruitment solutions. Recruiters using recruitment assistants found interview-quality candidates 33% faster, while the number of resumes reviewed to find qualified candidates decreased by 83%.
In the medical field, the company automated 30 million patient visits this quarter, an increase of 70% over the previous year, helping clinicians reduce administrative work and spend more time serving patients. Ten healthcare providers, including Advocate Health, Mass General Brigham (Mass General Brigham), Mercy, and Intermountain Health, have each recorded more than 1 million visits using the company's solutions since the beginning of the calendar year.
In the field of research, the company's Frontier Deployment Engineer (FDE) team collaborates with customers to accelerate scientific progress using Discovery. A typical example is Saint-Gobain (Saint-Gobain), which uses Discovery to accelerate the development of low-carbon building materials.
Under these first-party smart experiences, it's a unified platform. First is Foundry, which is increasingly becoming an important driver of growth. Foundry helps organizations build learning cycles rooted in their business context and securely deploy agents at scale. Currently, the company has more than 120,000 Foundry customers, and revenue has increased more than 3 times over the same period last year. Foundry's model selection and feature set, combined with its deep integration with the company's technology stack, is differentiating. Azure OpenAI directly benefited from significant advancements in the OpenAI model last quarter; combined with Foundry's built-in model selection capabilities, customers were able to build solutions that optimized quality, cost, and performance. In fact, the number of customers using multiple models to build intelligent solutions increased 5 times year over year, and revenue from the open source model on Foundry increased 10 times. The company's first-party model also continues to advance rapidly. The company aims to advance the “cost-to-outcome frontier” (cost-to-outcome frontier), that is, to deliver comparable performance at a cost far below the cost of training within three to six months of the launch of the leading model, thereby transferring these cost benefits to customers. This is reflected in today's products: mai-cyber now supports over 40% of MDash code scanners, reducing the cost of a single scan by 18%; mai-transcribe is the most accurate and cheapest transcription model in the world, supporting Dragon Copilot's multi-language workflow; mai-code-flash had 3 million users on GitHub alone last month, helping developers use AI credits more efficiently; Microsoft-Field1 is the company's structured The new decision-making model, which is superior to LLM and other decision-making models in terms of latency and quality, was used by more than 4,000 customers in the first month of release, including Atos Group (Atos Group), Norwegian National Petroleum (Equinor), and Austrian Olympic International Bank (RBI). The company recently completed MAI-base-2 pre-training. In testing, its training computing power efficiency was 4 to 12 times higher than other open weight models. This basic model will be used for subsequent inference, code, and cybersecurity MAI models.
In addition to the model, the company also opens up the framework that supports its own Copilot experience to Foundry customers, providing Microsoft-hosted methods to build agents that can reason, call tools, and complete complex and long-term tasks, while retaining the customer's model choices. Another key layer is data. The company now has 44,000 paid Fabric customers, an increase of more than 50% over the previous year. The company is integrating Foundry more closely with Fabric to enable customers to build their own intelligence next to data. Currently, over 19,000 customers use Foundry and Fabric at the same time, and enterprises can connect and operate agents based on real-time operations, analysis, and unstructured data. Customers are also rapidly deploying the company's AI-optimized databases, including PostgreSQL, SQL DB Hyperscale, and Cosmos DB, to provide agents with fast and secure data and contextual access. PostgreSQL's growth accelerated for the fourth consecutive quarter, while the number of PostgreSQL customers using Foundry increased 90% year over year.
The company also has one of the most important databases for any business—the knowledge base under Microsoft 365, covering people, their relationships, documents, emails, meetings, chats, projects, and other assets. Work IQ provides this context to every agent workload. Using Microsoft IQ, the company integrated Work IQ, Fabric IQ, and Foundry IQ into a unified context for intelligent systems. Nearly 60,000 customers are accessing contexts above Microsoft IQ, up 25% from month to month.
As enterprises deploy more agents, they also need to be governed at scale. Through Agent 365, the company provides an AI observability platform that enables customers to manage, protect, and optimize the performance and cost of their agents. Currently, there are more than 50 million agents in the registry, an increase of more than 30% over the previous month. Finally, as the access rights of agents to enterprise data and systems expand, security becomes more critical. The company's approach is to separate “intelligence” from “authority” and enforce identity, authority, monitoring, and containment outside of the model. Last week, the company expanded Perception's new capabilities to help customers protect and curb agents within agents, including agent discovery in local and cloud environments, dynamic permissions to perform minimal access to every task, and ongoing risk assessments. The company also launched the Guardian Smart Device Team to continuously monitor smart activity and prevent dangerous behavior before it occurs. Perception received positive feedback from preview customers, including KPMG (KPMG), Nationwide Building Society (Nationwide Building Society), Northern Trust (Northern Trust), and Salentis. More broadly, the company is helping customers protect and govern the use of AI across the enterprise. To date, customers have used Microsoft Purview to audit more than 7.6 billion agent activities, a fivefold increase over the previous month.
What underpins all of this is infrastructure. The company is building fungible (fungible) computing power pools for its growing business to dynamically allocate production capacity to the highest value opportunities in computing, data, Foundry, and Copilot. Accelerated demand caused the company to add 24 new data centers on five continents this quarter and speed up the launch of production capacity. The “doctor lifetimes” (doctor lifetimes, or delivery cycle) of the new GB300 in the largest regions was shortened by about 20% month-on-month. These regions form the main part of GPU deployment. At the same time, the company is extracting more value from existing infrastructure by optimizing data center designs, chip systems, software, and models. For example, the throughput of Copilot workloads increased by more than 50% month-on-month. The company continues to expand its computing power pool with the latest innovations from NVIDIA (NVIDIA), AMD and its own MAIA systems, and has installed the first batch of Nvidia Vera Rubin GPUs and is very excited about this. The Maia 200 has achieved up to 1.7 times the token output per watt in Copilot's AI experience and team, and the Maia 300 will further drive the gain, increasing the token output per watt by 40% compared to leading GPU systems.
The growth didn't stop with GPUs. The company's self-developed cloud-native Cobalt CPU continues to be strongly adopted, improving the cost performance ratio for general cloud workloads. Azure's overall growth spans a wide range of businesses and small to medium customers, as well as between core cloud services and AI infrastructure. Finally, there is the equipment and consumer business. On the Windows side, the company is implementing the vision of “bringing hybrid intelligence to every desk and every home, making every PC a secure smart container”. As more powerful models run locally, there is an opportunity to “almost unmetered” (almost unmetered) intelligence to a greater extent. The company is collaborating with the broader hardware ecosystem to bring these capabilities to the next generation of PCs based on the AMD Ryzen AI 400 series, Intel Core Ultra series 3, and Qualcomm Snapdragon X2 series. Earlier this month, the company also announced new PCs equipped with Nvidia RTX Spark and DGX Station for Windows. With these new devices, users can run last year's cutting-edge models on their laptops and last quarter's models on their desktops. The company integrated Windows with Agent 365 to enable agents to operate on devices with secure boundaries. The company launched Microsoft Execution Containers (MXC) — a policy-driven execution layer to govern and contain agents on Windows and other platforms, and integrates with Agent 365 to achieve governance and observability. Agents such as OpenAI Codex, Anthropic Claude Code, and Meta Muse are using MXC, and customers such as Meta have begun deploying in their enterprise environments.
When it comes to search and advertising, Bing and Edge are once again gaining share. On the LinkedIn side, the engagement of 1.4 billion members continued to deepen, and the length of time spent viewing streaming content increased 15% year over year. In terms of games, the company focuses on creating high-quality games and covers platforms chosen by various players such as console, PC, mobile and cloud. Minecraft delivered a record quarter with record levels of engagement and revenue; World of Warcraft (World of Warcraft) hosted the largest test in the series's history. The company entered the holiday season with the strongest Formula 13 lineup in recent years, including “Modern Warfare IV,” which went on sale last week and received the best reviews for “Call of Duty” in over 10 years, and “Grand Theft Auto 6” (GTA 6), which will hit Xbox next month.
Overall, management has never been more confident about the long-term return on investment, based on the needs seen, the strength of the product, and the efficiency driven by the company across technology stacks. Now leave the meeting to Amy.
Amy Hood, Executive Vice President and CFO
The results exceeded expectations in terms of revenue, operating profit, and earnings per share, and were driven by strong demand from Microsoft Cloud. Revenue for the quarter was $92.3 billion, up 19% year over year. Gross profit in dollars increased 17%, and gross margin was 68%. Operating profit increased 21% and increased 22% at the fixed exchange rate. Earnings per share were $4.89, up 18%, and 19% at a fixed exchange rate and adjusted for the impact of OpenAI investments. The exchange rate effect is generally consistent with the guidelines.
Operating expenses increased by 6%, driven by continued investment in R&D computing power and talent. Operating margin rose slightly to 50% year over year, driven by strong revenue growth, improved platform efficiency, and focused investment in the investment portfolio, partly offset by the shift to Azure sales structure. After adjusting the impact of OpenAI's investment, other revenue and expenses were negative $234 million, slightly lower than expected. Capital expenditure of US$45.2 billion was delayed until the second fiscal quarter due to delays in the delivery of some third party capacity. Approximately three-quarters of capital expenditure is spent on short-lived assets (mainly CPUs and GPUs, including higher memory and storage costs) to continue building the infrastructure customers need to build and expand AI applications; the rest is used for long-life assets. Financial leases for the quarter were $7.4 billion. As a reminder, the company confirms the full principal value of financial leases and capital expenses when the lease takes effect. The cash used to purchase plant equipment for the property was US$39.4 billion. Operating cash flow was $58 billion, up 29%, driven by strong cloud billing and payback. Free cash flow was $18.6 billion, reflecting higher capital expenditure. Finally, the company returned $10.2 billion to shareholders through dividends and share repurchases.
The following is the department's performance, presented according to the new departmental structure announced in September. Agents and Infra (Agents and Infra) revenue was $76.8 billion, an increase of 25%, reflecting the strong performance of all connected layers of the platform. Commercial bookings fell 12% (fixed exchange rate 11%), which was affected by large-scale Azure commitments in the same period last year, partially offset by healthy growth in commercial sales momentum this quarter (including a major Azure commitment this quarter). Commercial Residual Performance Obligations (RPO) increased 73% to $677 billion, and the month-on-month decline was less than historical seasonality due to large Azure commitments this quarter. The RPO weighted average period is 2.3 years, and about 30% will recognize revenue over the next 12 months, up 29% year over year; the rest will increase 98% for the portion confirmed after the next 12 months.
On the Microsoft Cloud side, revenue was 63.9 billion US dollars, and the growth rate accelerated to 30% (31% fixed exchange rate). The increase in US dollars from month to month almost doubled year on year, reflecting strong demand from Azure and M365 commercial clouds. Azure's revenue exceeded expectations and grew 48% despite the aforementioned third-party capacity delays, driven by improvements in Foundry and CPU monetization. Foundry's revenue increased more than threefold year over year, partly due to improved GPU computing power pool efficiency, which enabled the company to allocate more production capacity to customers and the higher single-token revenue it benefited from this quarter. In terms of CPU computing power pools, engineering work enabled the company to identify and monetize underutilized production capacity this quarter, helping the company use existing production capacity to serve more growing demand. The timing of such engineering efficiency, as well as ARPU changes brought about by third-party GPU leasing and delivery, the company's own production capacity construction, and model structure, may cause quarterly fluctuations in the year-on-year growth rate. M365 commercial cloud revenue increased 18% (fixed exchange rate 19%, adjusted after 1 percentage point of current revenue confirmation for the same period of the previous year); reporting volume increased 17% (fixed exchange rate 18%). Performance exceeded expectations, driven by increased adoption and usage of GitHub Copilot, partially offset by weak sales execution of M365 E3 and standalone SKUs. The company continues to optimize sales operations, balancing a more traditional per-seat SKU rhythm with updated high-end E7, M365 Copilot, and production-based productivity products. The year-on-year increase was driven by ARPU and seats. Among them, ARPU growth was led by high-end per-seat products such as M365 Copilot, e5, and e7 — e7 paid seats have reached millions — and pay-per-use products, mainly GitHub Copilot, contributed. Copilot's revenue grew at an accelerated pace, and adoption of M365, GitHub, and security continued. M365 commercial paid seats increased by 5%, partly affected by the weak sales execution described above. The growth was driven by the expansion of stock in all customer segments, mainly from SME and frontline employee products. M365 consumer cloud revenue increased 17% (fixed exchange rate 16%), declining month-on-month, as it began to face a high base brought about by price increases in the same period last year. Productivity and server license revenue declined by 3% (fixed exchange rate of 2%), mainly affected by a decrease in transactional purchases due to the high base brought about by product launches in the same period last year, partly offset by an increase in the share of contracts with higher current revenue recognition. Industry solutions cloud revenue grew 10%, and Dynamics 365 and LinkedIn talent solutions grew, but the booking trend continued to slow. Frontier and Support Services (Frontier and Support Services) revenue growth of 4% (fixed exchange rate of 5%) was driven by the support business.
Microsoft Cloud's gross margin fell to 66% year over year, driven by a shift to the sales structure to Azure and partly offset by improved margins at the product level, mainly Azure. The sector's gross profit in dollars increased by 20% (fixed exchange rate 21%). Operating expenses increased by 10%, driven by the aforementioned R&D investment. Operating profit increased by 24% (fixed exchange rate of 25%), and operating margin fell slightly to 53% year over year, offset by strong revenue growth, improved platform efficiency, and focused investment.
In terms of devices and consumers, revenue was $15.5 billion, down 3%. Search and advertising revenue (after deducting TAC traffic acquisition costs) increased by 12%, exceeding expectations, and was driven by a sharp rise in the volume and price of LinkedIn, Edge, and Bing. Xbox revenue fell 2%, and Xbox content and service revenue fell 2% (1% fixed exchange rate), driven by declines in third-party content and Xbox Game Pass, partially offset by first-party content growth. Windows OEM and device revenue fell 22%. Windows OEM revenue fell below expectations, down 27% (fixed exchange rate 28%). Inventory levels are still high, driven by demand brought about by the end of support for Windows 10 in the same period last year, inventory removal from OEMs and channel partners, and weak demand in the PC market. The department's gross profit in US dollars was basically flat year over year, and operating expenses fell 3%. It was driven by Xbox, partially offset by the aforementioned R&D investment. Operating profit increased 2% (3% fixed exchange rate), and operating margin rose to 35% year over year, driven by reduced amortization of intangible assets related to the Activision acquisition, partly offset by a decline in the share of Windows OEM revenue.
The following is the outlook for the second fiscal quarter, all in US dollar terms except as specified. Based on the current exchange rate, the company expects that the exchange rate will have no significant impact on total revenue, sales costs, operating expenses growth, and revenue growth in the two divisions. In terms of sector guidance: Smart devices and infrastructure revenue is expected to be US$80.1 billion to US$80.7 billion, an increase of 24% to 25%. In terms of commercial reservation volume, commercial sales momentum is expected to grow healthily, but the results will be affected by major Azure commitments signed in the same period last year; the relevant timing of such major Azure promises will cause significant fluctuations in reservation volume, making the quarterly growth rate less reflective of basic sales execution. On the Azure side, revenue is expected to increase by 49% to 50% under the fixed exchange rate, and continued efficiency improvements in CPU and GPU computing power pools will partially offset the delay of third-party GPU-as-a-service contracts in the second half of the year; customer demand continues to exceed available supply. Looking ahead to the second fiscal quarter, the year-on-year growth rate will continue to reflect the timing and scale of project efficiency delivered during the quarter, new production capacity, and a high base of accelerated growth in the second half of the same period last year. On the M365 commercial cloud side, it is expected to increase by about 19% under the fixed exchange rate. The exchange rate is expected to reduce the reported growth rate by less than 1 percentage point; the momentum of high-end seat-based products and pay-per-use products is expected to continue, and implementation of E3 and independent SKUs is expected to improve; growth will be mainly driven by ARPU, and seats will also contribute, and revenue growth is expected to continue to accelerate throughout the year. M365 consumer cloud revenue is expected to grow at a low double-digit rate, falling month-on-month due to the high base of the price increase described above. Productivity and server license revenue are expected to drop by a high single digit, declining month-on-month, as the share of contracts confirmed by high current revenue is expected to decline; the quarterly growth rate may fluctuate significantly due to contract structure, product launch timing, and other transactional procurement trends. Industry Solution Cloud expects medium to high single-digit growth, which is still affected by the booking trends described above. Frontier and support services are expected to grow by as little to mid-single digits. Microsoft Cloud's gross margin is expected to be around 65%, a year-on-year decline, as continued improvements in product-level profit margins are offset by the structural shift to Azure.
In terms of devices and consumers, the estimated revenue is between $15.9 billion and $16.4 billion. Search and ad revenue (after TAC) is expected to grow at a low double digit rate, driven by consistent trends in volume and price, and overall search and ad revenue is expected to grow at a similar rate. Xbox content and services are expected to grow in the middle to high single digits, with a month-on-month increase, driven by the time the third party content was published; given the pace of content publishing throughout the year, growth is expected to be strongest in the second fiscal quarter, slowed down in the third fiscal quarter, and further slowed in the fourth fiscal quarter. Hardware revenue for the second fiscal quarter is expected to decline year over year. Windows OEM and device revenue is expected to drop by around 30% (mid-30s). Growth will still be affected by high inventory levels, PC market demand, and the high base of Windows 10 discontinuation of support in the same period last year. As in previous quarters, the likely outcome range is still wider than usual. The company expects these trends to continue throughout the rest of the fiscal year.
As a result, at the company's overall level, revenue is expected to be between $96 billion and $97.1 billion, an increase of 18% to 19%. Sales costs are estimated to be US$32.7 billion to US$32.9 billion, an increase of 26% to 27%; operating expenses are US$18.5 billion to US$18.6 billion, an increase of 9%, driven by continued investment in R&D computing power, production capacity and talent. The cost of sales and operating expenses guidelines both include estimates of recently announced voluntary employee separation plans. Operating profit margins are expected to decline slightly year over year. Excluding the impact of OpenAI's investment, other revenue and expenses are expected to be about negative 300 million US dollars, as interest income will be offset by interest expenses (including interest related to data center financial leasing). The effective tax rate for the second fiscal quarter is expected to be around 20%.
In terms of capital expenditure, the second fiscal quarter is expected to exceed $50 billion. As a result, annual calendar expenditure is expected to be around $170 billion due to delays in the delivery of production capacity to some of the aforementioned third parties. Capital expenditure is expected to accelerate in the second half of this fiscal year, mainly due to CPUs (including higher memory and storage costs) and GPUs, as well as delays in third-party capacity delivery until the second half of the year. As infrastructure efficiency continues to improve, every dollar of capital expenditure is expected to support more revenue and higher returns. In view of strong business performance, the company expects FY27 to maintain positive free cash flow throughout the year. To supplement the cash generated from operations and maintain financial flexibility, the company will use additional sources of financing when the time is deemed appropriate.
All in all, FY27 had a strong start, and Microsoft Cloud's growth accelerated. Product improvements and efficiency improvements in data center infrastructure, first-party applications, and agents are increasing the value the company delivers to customers and improving results. Delivering value to customers lends the company a foundation for long-lasting long-term growth.
II. Q&A session
Morgan Stanley analyst Adam Wood: Drivers of AI infrastructure returns
Q: Along with lower-than-expected capital expenditure and better cloud gross margin performance, as new AI production capacity is launched one after another, the more critical driver of returns over the next few years is an improvement in utilization and unit cost during the year, or the ability to monetize each workload in more layers on top of Fabric, Foundry, and Copilot.
A (Satya Nadella): All of the above factors. The company's idea is to establish a pool of computing power and make it as efficient and interchangeable as possible—this involves not only the full physical aspects of data center construction, the operational rigor of putting production capacity online, but also how to run optimized workloads on that infrastructure. Once this is done, it is only necessary to distribute according to demand signals. This can be seen even under the new divisional disclosure standards: the stateful database under Copilot is Work IQ, which is used by Foundry — customers build supply chain multi-agent systems on Foundry that span unstructured and structured data, and of course use PostgreSQL databases in Azure. Additionally, all agent jobs are registered in Agent 365 and their security is managed by Perception. Therefore, the company sees these as the same set of infrastructure and only presents them in the form of different products. The core question is where profit and leverage come from — not from any single link, but from every link, and when the company integrates them, the benefits grow compounded.
A (Amy Hood): Sorry Adam, if I missed the first half of the question, I've at least answered the latter part.
UBS analyst Carl Kilstead: Drivers of Azure's 49%-50% growth guide
Q: Let's further break down one of the key metrics of this conference — the Azure fixed exchange rate growth guide of 49% to 50% in December. As you can imagine, there are many variables: production capacity currently in production, distribution to third-party customers, and probably OpenAI's strength, but given the excellent performance of this figure, please discuss some of the key drivers.
A (Amy Hood): Talking about the first fiscal quarter first helps to deduce the second fiscal quarter and even the future. In fact, this is also related to Adam's question. In the first fiscal quarter, the engineering team delivered excellent work on improving the efficiency of the platform layer — not just GPU efficiency, but the efficiency of every layer: how applications use the platform, how Foundry uses the platform, and the Azure platform itself. These efficiencies accumulate little by little every week. Some ideas work, some don't, and require continuous investment, so it is not necessarily possible to accurately predict the impact of a single season. The first fiscal quarter performed well, and some of the results were reflected in the company being able to allocate more production capacity to Foundry during the quarter. This is a good thing because, as mentioned, demand still exceeds supply; whenever the engineering team delivers this kind of efficiency, the company tries to use it for Foundry because, as Satya said, it's critical for customers to build new stacks on a wide range of occasions. Second, token pricing will cause some fluctuations. This quarter's token pricing was higher than expected — the good profit margins brought about by the high-quality model did benefit the company, which happened in the first fiscal quarter. Furthermore, in the data for the first fiscal quarter, the above efficiency is not only applicable to GPU computing power pools; people may have underestimated the integrated workload of GPU, CPU, storage, and memory in Satya's example — it involves all aspects. As a result, the company was also able to create additional production capacity in the existing computing power pool, which quickly turned into results in the first fiscal quarter. In the second fiscal quarter, these efficiency gains continued, and the company still needs to explore more efficiency — we are confident, because the team's work in the first fiscal quarter is expected to continue into the second fiscal quarter, and it also needs some help from this work, because it was expected that part of the supply for the second fiscal quarter has been delayed until the second half of the year. Additionally, token pricing is uncertain during the quarter. Overall, I'm satisfied with the efficiency gains I've seen, and I'm happy to give guidance based on them; however, other aspects are harder to predict.
Bernstein analyst Mark Modler: Azure AI-powered source
Q: Focus on the driving sources of Azure AI's strength: enterprise customers or general digital native/AI native customers; how should we view the current share of enterprise customer workloads in the overall workload as inference begins to expand and Microsoft becomes the first choice for enterprises; and the drivers for the strong growth of Azure AI platform products — whether they are coming back from enterprises or digitally native customers (obviously a mix of the two, but details that reflect enterprise volume are more valuable).
A (Satya Nadella): In Europe last week, the quality of intelligent systems seen was impressive, and they are being widely deployed. For example, a delivery platform's fully automated smart contactless system, covering automatic order placement, transaction dispute management, etc.; Rolls-Royce (Rolls-Royce) deployed a system to manage its maintenance — its jet engine involved 30,000 components, which were quite complex supply chains. The planning, execution, and inventory management systems were all built on Foundry and implemented in the form of an intelligent body; Austria's Olympic International Bank (RBI) showed more than 20 end-to-end business processes, whether credit or others. Its breadth, quality, and multi-model characteristics — they started with cutting-edge models, and the Unilever (Unilever) case even directly used the MAI model trained by the company Traces. The maturity of model optimization, workload optimization, and token optimization is encouraging. The breadth is opening up.
Ultimately, enterprise workloads give people the feeling that enterprises built database applications in the past, and now they are happening again in the field of AI. In fact, some “chasing novelty” things are fading away, and what companies are more concerned about is whether they can build a system, whether they can have weight, whether they can manage costs, and whether they can have reliability and safety. As a result, an enterprise-wide movement is beginning. But like everything, the company's business is not quarterly, but a ten-year dimension — this is exactly what the company was waiting for, not just four customers purchasing large amounts of production capacity and showing growth for a few quarters (this isn't a business). The company pursues breadth. The company has a good card position in terms of the breadth of customers and is open to business: as long as the price is right, part of the production capacity can be sold; otherwise, it will patiently build a broad business.
Goldman Sachs Analyst Gabriela Borges: Token Combinations and Token Value Evolution
Q: On the “breadth” topic, management mentioned in the statement that MAI can now match the lead model within three to six months after launch, while OpenAI accounted for 11% of total revenue for the quarter. Please take a step back and talk about the latest thinking: how the token combination and token value will evolve between cutting-edge models, MAI, and proprietary AI strategies of some mature enterprise customers.
A (Satya Nadella): The toolbox for any workload can be understood as three things. First, when the new model is released and the performance is excellent, customers can evaluate the results they value and adopt it — as was the case with some OpenAI models from the previous quarter. They are excellent. The company is looking forward to OpenAI's progress and future at Astra, and has also seen good demand. Second, the customer will then ask: Can multiple cutting-edge models be used at the same time, because their prices vary greatly. In fact, in the real benchmark given by Artificial Analysis this morning — using the company's Fusion coding intelligence — the company ranked second or third, which is very exciting: now you can get cutting-edge performance, a 45% increase in speed, and a 75% price reduction, so why not do it. For coding smart users, they want Anthropic and OpenAI to compete for every task rather than once a quarter. This is the value of Routing and Fusion Autorouting. Third, and importantly, many customers — such as Unilever — have their own systems, and in the end, even though they have done a lot of work with cutting-edge models, they still want to train their own models, either open weights or the MAI model in this example. Therefore, all of these should be available options, and customers will choose based on pricing and needs. There is timing in this: it's not about starting and stopping there, but maybe starting with a cutting-edge model, then moving to open weights, moving to the MAI model, and then returning to the cutting edge a few months when the next model is released — the workload is just that dynamic. Therefore, on Foundry, Fabric, CoPilot, and GitHub Copilot, the company wants to build a model system that continuously delivers value to customers rather than locking them into one item.
Jefferies Analyst Brent Thiel: Maintaining 50% Revenue Growth and 50% Operating Profit Margin
Q: Azure's growth rate is close to 50%, and the operating profit margin is close to 50%. The market often asks how to maintain this growth at the level of both first-line growth and bottom-line profit margins of 50%. Please talk about ongoing operational improvements to maintain a high bottom line.
A (Amy Hood): First, the best solution is revenue growth, so you should always start at the top of the income statement and move downwards — even though many people think the company is moving up from the bottom of the profit sheet. The core of what the company is doing actually runs through all the reviews: the restructuring of departments is due to changes in the way products and services are being delivered — delivering new, more integrated products with higher value. As top-end revenue accelerates, it naturally contributes to positive profit margins and gross profit dollars in a healthy way. Second, the company is really focused on efficiency. When it comes to “efficiency,” people often think of cost, but although the results are related to costs, the real benefits of efficiency are stability, safety, and reliability — that is, delivering the highest quality units at the lowest cost, which is operational excellence. Therefore, on top of product quality (which performed well in the last quarter), continuous improvements at the operational level are combined with a focus on operating expenses to ensure that it invests in building new things or providing safety and quality. The company made a very proactive choice to focus on these two things, which are particularly important in the current environment. Good results can be seen when this is achieved in a good market with a wide range of customers and global business. This is especially true this quarter: the quarter delivered significant efficiency gains, achieved revenue growth without additional costs, and naturally saw benefits. Management is pleased with what underpins this figure — a focus on proper leverage.
Stifel analyst Brad Rybak: M365 consumption model unlocked
Q: When the company switched GitHub from a per-seat model to a consumption model, it unlocked significant revenue growth. Whether the tools that core M365 customers now have in place on the consumer side can they be unlocked just as fast and require no sales investment, or will it take quite a bit of effort to turn these “meters” around.
A (Satya Nadella): On the coding side, it will be faster because the market is used to paying for what you use. In fact, products such as Fusion announced today on GitHub Universe provide a more attractive way — buying credits and hoping to use them most efficiently, which is what GitHub Copilot Fusion promises. The company is optimistic about this. As far as Copilot itself is concerned, in a sense, the company is injecting more and more value into subscriptions with every model release; in many cases, IT invests in integrated promotion; in many cases, IT is funding; in many cases, it wants standardized configurations throughout the organization. At the same time, pay-as-you-go or consumption meters are also kept for scenarios where there is a need. Both options are open, and the company is trying to make customers understand both and use them in the best way, rather than overbuying one of them and getting bogged down in token maximization, which is bad for everyone. The company wants customers to understand the value of a subscription — it's continuously improving and not static — while also being able to exercise consumption meters. This is where sales actions and customer understanding are becoming more mature. The company is satisfied with the progress, but there is still a lot of work to be done.
A (Amy Hood): This is why the company invests in cutting-edge companies and is cooperating with the aforementioned changes in sales actions — the two are indispensable, because as Satya said, they want customers not only to make full use of the Copilot license, but also to understand when consumption meters bring value to them. Investments in Frontier Deployment Engineer (FDE) organizations or Frontier collaborations, as well as new moves in modern offices and Microsoft 365 sales teams, are all aimed at achieving this. As Satya mentioned, the migration outside of coding will be slower, but the company is still seeing encouraging signs and is positive about it.
Bank of America Analyst Tal Liani: Capital Strategy
Q: Regarding capital commitments/capital strategies, management stated in the statement that future financing will be carried out if required. Please discuss strategies between debt, equity, and other sources, and how financing will be carried out if needed; please also confirm whether Copilot-related figures have been disclosed this quarter.
A (Amy Hood): What management said before is: if the financing conditions are attractive and reasonable, the company will use it. The company has a great deal of financing flexibility. It was previously mentioned that there is a form of financing called leasing, which companies have been using. Furthermore, the company is fortunate to have access to the debt market, which has not been — nor needed — to be used, given the company's ratings. Of course there are other options. The company is satisfied with positive free cash flow and is confident in the return on every dollar invested in capital expenditure, so it has plenty of options to use the financing market when appropriate and attractive. As for the immediate question — no specific Copilot figures were disclosed for this quarter; only revenue and usage momentum were discussed.