SAP SE (SAP) Earnings Call Transcript & Summary

Ladies and gentlemen, thank you for standing by. Welcome, and thank you for joining the SAP Q3 2026 Financial Results Conference Call. [Operator Instructions] I would now like to turn the conference over to Alexandra Steiger, Global Head of Investor Relations. Please go ahead.

Good evening, everyone, and welcome. Thank you for joining us. With me today are CEO, Christian Klein; and our CFO, Dominik Asam. On this call, we will discuss SAP's third quarter 2026 results. You can find the deck supplementing this call as well as our quarterly statement on our Investor Relations website. During this call, we will make forward-looking statements, which are predictions, projections or other statements about future events. These statements are based on current expectations and assumptions that are subject to risks and uncertainties that could cause actual results and outcomes to differ materially. Additional information regarding these risks and uncertainties may be found in our filings with the SEC, including but not limited to the Risk Factors section of our annual report on Form 20-F for 2025. Unless otherwise stated, all numbers on this call are non-IFRS and growth rates and percentage point changes are non-IFRS year-on-year at constant currencies. The non-IFRS financial measures we provide should not be considered as a substitute for or superior to the measures of financial performance prepared in accordance with IFRS. And with that, over to you, Christian.

Yes. Thank you, Alexandra, and welcome to everyone joining us. Q3 was an outstanding quarter. We brought the autonomous enterprise to life. Customers are now getting their hands on our latest AI capabilities and the response has been incredible. Our strategy is not just resonating. It's being adopted. Here's what that looks like in the Q3 numbers. Current cloud backlog reached EUR 24.1 billion, up 26%. Cloud revenue was EUR 6.7 billion, up 25% in constant currencies. The strong growth in cloud revenue was significantly driven by our business AI, data and platform business. Continuing the trend from recent quarters, 9 out of 10 of our 50 largest deals included AI and Business Data Cloud. Total revenue for the quarter was EUR 10.3 billion, up 12%. This is the first time in SAP's history that quarterly revenue crossed EUR 10 billion. Operating profit grew 18% to more than EUR 3 billion. And we are extending our leading position by winning market share at scale. According to IDC, we are growing about 11 percentage points faster than the market. In Q3, AI was the main driver behind our customers move to the cloud. There are 3 key reasons. First, AI accelerates the [indiscernible] cloud migration. Second, AI makes the business case for our suite more attractive. And finally, our customers employ our business AI platform with Verizon grow. They know it's essential for building the data and process foundation required for high-quality enterprise-grade AI. Wise and business AI are winning across every sector. We added new customers in financial services with [indiscernible] Bank of Canada, technology and research with Infineon and Fraunhofer-Gesellschaft, manufacturing with NTN Corporation and logistics with [indiscernible] and the U.S. Defense Logistics Agency. On the growth side, AI is also a key differentiator. For example, [indiscernible]. South America's largest railway manufacturer plans to use Joule to boost real-time operational visibility. [indiscernible] wins included Petroplus, A5 X and [indiscernible]. Industry [ AI ] is becoming a critical differentiator, especially in large enterprises. We saw wins such as Astrazenca, where we are actively codeveloping AI solutions for the life science industry. [ Decade long ] we'll be leveraging our AI to build a foundation for a more intelligent, automated and customer-focused business. Meanwhile, Danish retailer selling group uses industry AI to manage suppliers, classic for new products and trigger supply chain actions. Time to shelf improved by up to 66% and Manual category management time was cut by up to 70%. Further AI wins came from [ Man ] Industries Limited, building what they call a once in a generation digital foundation to unify operations. Hengel chose our business AI platform for more efficient finance processes and improve decision-making. Austrian Post, [indiscernible] selected BDC. In solution areas and LOBs, [ Tower ] Corporation went with SAP Signavio. Entity data with our procurement solutions and [ Helvetia ] are holding with success factors. IDC estimates that SAP is growing more than 1.7x faster than the overall HCM market year-over-year. Many customers went live with Wise in Q3, including Mondelez, Panasonic, Roche, Infosys and [indiscernible]. We also saw you HUGOBOSS with SuccessFactors [indiscernible] Talia Boucher with SAP warehouse solutions and Battle Smart Recruiters. In Q3, we already saw the great impact of the autonomous enterprise strategy on our business and in our numbers. Let's start with what customers see. Joule Work. This is where enterprise AI finally becomes usable. Our 400 million SAP users can now work with SAP in natural language and Joule gets the chop down. It is based on 3 key capabilities: first, conversations. The way users get work done. Here, users ask questions and delegate tasks in natural language. Joule handles every work directly. But when a task requires more than just a query, the user can use Joule to call a specialized agent. Take a supply chain planner. They can ask Joule to summarize the supplier contract then delegate the resolution of a supplier shortfall to an agent that identifies an alternative source and triggers the production adjustment directly in SAP. Second, spaces. The live view across system, Joule spaces brings data together into a single always up-to-date view, so users don't have to build PowerPoints and Excel spreadsheets. For our supply chain planner, they have a space combining live IBP and S/4HANA data with a real-time supply chain risk summary. They can see current demand, supply and the risk all in one place. Third, Joule Studio, the way to extend our agents or build your own from no code to pro code and with seamless integration of at end with Joule Studio customers extend SAP's agents or create new ones for their specific needs. Here, the planners team can use the supplier resolution agent and tailor it to their own sourcing rules without starting from scratch. As the examples showed, Joule Work serves every business user. It is much more than an AI chatbot. It is the workbench for the autonomous enterprise. This is true for users who need a presentation, a dashboard, an e-mail or solutions for even the most complex enterprise challenges, thanks to our [indiscernible] systems and 300 agents across the autonomous suite. While I'm pleased with the progress we made, the focus is quality over quantity. One well-designed agent can outperform several fine grained ones. That is how we build. With Joule Work desktop, companies like Reinhausen see the potential to permanently transform how their teams interact with business data. The same shift is happening in finance at Accenture. In their words, their Joule journey. It's not about deploying another AI tool. It is about working differently. And Danforth shared that the sourcing assistant has significantly improved the efficiency and quality of RFP creation. What makes Joule Work and the autonomous suite different is what lies behind this simple experience. Our SAP business AI platform, the place to build contextualize and govern AI. Let's look at this in more detail, starting with build. Joule Studio gives every developer and domain expert, the tools to create, extend and deploy agents on top of SAP's platform. And when building custom solutions, customers can select the best model for each use case. SAP offers access to more than 70 models, both Frontier and open weight. Unlike some of our peers, SAP is not committing to a preferred model. We give our customers choice, and we will always make sure that they have access to the best options available. Moving on to context and reason. Joule draws from the deepest business contacts in the industry through our knowledge graph, anchored in BDC, mapping more than 7 million fields and 0.5 million tables, along with 400 data products and our full process knowledge. With our acquisitions of Dremio and Reltio, we extend that reach far beyond SAP data and give customers a unified contact layer across their entire data landscape. This structured business ontology means Joule can solve even the most complex tasks with high accuracy, faster results and better... [Technical Difficulty]

Pardon the interruption. One moment while we establish our audio. Thank you, everyone, for your patience. Please continue to stand by while we reestablish the audio. And we are live now again, please go ahead.

Thank you. Okay. Here we go again. Starting with build. Joule Studio gives every developer and domain expert tools to create, extend and deploy agents on top of SAP's platform. And when building customer solutions, customers can select the best model for each use case. SAP offers access to more than 70 models, both frontier and open weight. Unlike some of our peers, SAP is not committing to a preferred model. We give our customers choice, and we will always make sure they have access to the best options available. Moving on to context and reason. Joule tool draws from the deepest business context in the industry through our knowledge graph, anchored in BDC, mapping more than 7 million fields and 0.5 million tables, along with 400 data products and our full process knowledge. With our acquisitions of Dremio and Reltio, we extended reach far beyond SAP data and give customers a unified contact layer across their entire data landscape. The structured business ontology means Joule can solve even the most complex tasks with high accuracy, faster results and better price performance. In Ariba, Joule delivers twice the accuracy of leading frontier models at half the total cost. And we are continuously deepening the ontology itself. Our intent to acquire [ Tech Wolf ] is a good example. It puts us at the heart of every people transformation and extends our [indiscernible] into skills and workforce intelligence. Finally, governance. Joule Work comes with governance and security baked in. Every agent built on the business AI platform is 1 and governed by SAP. We ensure identity and [indiscernible] sales management, compliance with regulations across 60 industries and -- 60 countries and 26 industries and certifications by auditors. Our AI agent hub and [indiscernible] provides 1 central inventory to manage and govern every SAP and non-SAP agent. And 1 key principle is very important. At SAP, the customer's data remains the customers' data. We never upload customer data or process IP into large language models. Not for training, not for model improvement, not for any purpose. And that commitment shapes every partnership with the model providers we make. In a nutshell, we bring together the best models with the deepest business context and govern all of that across our customers' entire agent landscape. That is why SAP provides the business AI for the enterprise. And we will share more announcements across all these areas at [indiscernible] next week. But the story is not just what we ship. It is how we transform into our -- into an autonomous enterprise ourselves. We are reinventing once again SAP's own operating model. First, in engineering. Our 40,000 developers are not just building AI, they're working AI first, guided by clear principles. We are co-innovating with customers, increasing speed and focusing on the data and ontology our AI agents need. The results are clear. Our developers completed approximately 70% more cloud contribution since beginning of the year. Second, in go-to-market, we are using AI everywhere to transform how we engage with customers. This includes running targeted campaigns to attract the right leads and evolving our consulting services into an AI factory to accelerate adoption. The same holds true for AI across our corporate functions. And across SAP, our Joule Work desktop is live for more than 110,000 employees. We are already seeing productivity gains of 20% across our central functions like finance, HR and procurement with even more productivity gains on more complex analytical and planning time. Let me summarize. Q3 was an outstanding quarter. We are successfully executing in a volatile world because our customers see AI as a critical lever for competitiveness and resilience. Our pipeline for Q4 looks strong. At the same time, we remain mindful that Q4 is always our biggest and most important quarter with the greatest movement. Our teams are in full execution mode. The best enterprise AI is not the loudest. It is the AI, you can trust to run your business. SAP is that AI. And we are turning that opportunity into results today. With that, I will hand over to Dominik.

Thank you very much, Christian, and thank you all for joining us this evening. Q3 was another strong quarter as reflected by current cloud backlog expansion and healthy operating profit growth. We continue to see our work bearing fruit as we move closer towards our vision of an autonomous enterprise. SAP remains mission-critical for its customers who want to fundamentally change the way they do work by virtue of AI. As Christian alluded to in his prepared remarks, SAP continues to set the standard for large skin business transformation and now leads the way in redefining how enterprises run in the age of AI. Today, our early adopter care program validates our progress and has been overwhelmingly positively received by customers, a step change versus a year ago. The second half is off to a strong start, and we remain committed to investing in growth by preserving our bottom line in lockstep with the operational framework we've put in place by balancing expense versus revenue growth. We continue to be excited for what's to come. and remain focused on capitalizing on what's clearly a once in a lifetime opportunity for SAP and our customers. Now let me provide more details on our financial highlights. Current cloud backlog surpassed EUR 24 billion, up 26%. SAP continues to be the undisputed large-scale enterprise SaaS and PaaS vendor in terms of growth. While we continue to expect a slight deceleration in constant currencies current cloud backlog growth in 2026, we do foresee continued market share gains. As we outlined at the beginning of the year, the slight deceleration is attributable to several factors, including the continued political macroeconomic uncertainty we see today that could impact our largest booking quarter, which is, of course, the fourth, as well as a rapidly expanding base that continues to grow in absolute terms every year. Cloud revenue grew to approximately EUR 6.7 billion, up 25%, helped by several one-off positive effects, which added approximately 1 percentage point of growth in Q3. This was due to largely a positive revenue recognition effect, which in turn mainly stem from a catch-up of revenues from customers returning from financial distress. Cloud ERP suite revenue increased by 28% in Q3, now accounting for 89% of total cloud revenue. Software licenses revenue decreased by 9%. Finally, total revenue in the third quarter was EUR 10.3 billion, up 12%, a remarkable result when taking a minus 5% year-on-year decline in services into account. Now let's take a brief look at our regional performance. In the third quarter, SAP's cloud revenue performance was particularly strong in APJ and EMEA and solid in the Americas region. Brazil, France, Germany, Italy, India, South Korea and Spain had outstanding performance. While Switzerland, the United States -- United Kingdom and the United States were particularly strong. Now moving down the income statement. Our IFRS cloud gross margin in Q3 was 73.9%, and non-IFRS was 74.3%, down 0.7 percentage points year-on-year. Average operating profit increased by 18% to EUR 2.9 billion. Non-IFRS operating profit was also up by 18%, now surpassing EUR 3 billion. IFRS and non-IFRS bring profit growth benefited from prior year negative impact of over EUR 1 billion each from a change in case law affecting SAP's other tax litigation and from a workforce transformation, which were both recorded in the third quarter of 2025. While the operating profit strength this quarter was a result of the factors I mentioned above, it is important to remember that investments to accelerate our AI transformation and to further enhance our products are in full swing. We're managing the business and associated costs with a disciplined approach, aware that this is a long-term endeavor and we expect to see these ongoing investments continue in the near term. The IFRS effective tax rate was 24.4%, and the non-IFRS effective tax rate was 25.8%. Free cash flow in Q3 was robust at EUR 1.7 billion. Finally, IFRS earnings per share increased by 14% to EUR 1.96 and non-IFRS earnings per share increased by 19% to EUR 1.89. Now on to our outlook. As disclosed in the quarterly statement issued earlier today, we now expect to land towards the upper end of our previously guided range for 2026 cloud and cloud and software revenue while reaffirming our previous guidance for current cloud backlog, operating profit and free cash flow. Finally, we continue to believe that we are well positioned to accelerate total revenue growth in 2027. While we are committed to this call, I would like to take a moment to provide some additional context to help you better understand our path forward. First, as a reminder, the fourth quarter accounts for about 50% of our annual bookings volume. Given the continued political and macroeconomic uncertainties we face today, this could weigh on Q4 bookings and as a result, our 2027 trajectory. Second, as we mentioned in the first quarter of this year and as you can see in the 5% decline, we want to leverage services more intentionally to promote the adoption of our AI solutions and build critical mass. This strategy has served SAP well to date. And as we are seeing positive return on investment, we will continue to double down on delivering even stronger adoption support to our cloud and AI customers which will likely put more downward pressure on billable hours and services revenue also next year. In fact, we believe that in knowledge work industries, business models, which solely rely on selling more hours to customers will be more and more challenged by AI enable lower cost, more scalable and nevertheless, higher-margin alternatives. So further portfolio adaptation in services is a necessity. And lastly, as mentioned earlier, Q3 was another strong quarter as evidenced by growth across top line parameters. As a result, we expect to trend towards the upper end of our cloud and software revenue outlook which will raise the bar against we will measure our 2027 revenue acceleration. Taken together, we are still planning for total revenue growth to accelerate in 2027, but must be cognizant of all the factors that make this trajectory more challenging. To close, SAP was once again able to navigate a choppy environment executing on its strategy and driving commercial success. The second half of 2026 is off to a strong start, and we continue to make progress against the operating plan we set at the beginning of the year. While we have been able to maintain steady momentum year-to-date, we will put our full focus on the decisive fourth quarter bookings required to maintain our growth acceleration against now even more challenging comparables in 2027. Thank you, and we will now be happy to take your questions.

All right, and we now take your questions. [Operator Instructions] Operator, please open the line.

[Operator Instructions] We'll take our first question from the line of Adam Wood with Morgan Stanley.

Congratulations on that very strong third quarter. Maybe if I could just ask on the CCB guide, probably Dominik. I think you still guided a slight detail and change the language around the macro expectations. I think you've said before that slight decel was 2%, which should imply [ 23% ] for the fourth quarter versus we ended Q3 at [ 26% ]. Could you just talk a little bit about why you're not willing to be a bit more optimistic here? Is it just an abundance of caution given the sale of the quarter? Is there anything you're seeing in pipelines or in customer decision-making that would give you that greater caution at this point in the year?

Yes. Thank you for the question. I mean, obviously, we had a very robust performance of CCB growth so far this year. But again, there is half of the bookings of the year to be still executed in Q4. And if I look at the current macro backdrop, it's not so great. I mean you have seen the oil price going up. Luckily, the conflicts in the Middle East not really resulted in massive disruptions of supply chains, but a huge pressure on pricing of many of the commodities that are relevant for the globe. Think about oil price going up from interest rates have gone up, and we've all seen long-term correlation analysis that when interest rates reached like 5% plus and they have reached that in the U.S. now. These interest rates start to develop heat. I'd say it's probably reasonable to not assume that the macro is super conducive in Q4. And yes, I would not kind of like to be understood that we gave a precise guidance on what exactly slightly means, we just referred to was kind of implied in the prior guidance. So it's a little bit of an arrangement. And so let's say, the upper end of that range might be 24% or so. So I don't want to be necessarily saying that there's a point value of [ 23%, ] there's a certain range and possible outcomes. And let's see how the rest of the quarter shakes out.

We'll take our next question from Toby Ogg with JPMorgan.

Just on the backlog again. So 26% constant currency, stable versus Q2 and looks to be slightly accelerating on an organic basis. but your guidance has always been for that slight deceleration, and that's unchanged. So is the backlog currently growing faster than your own internal expectations at this point? And if it is, could you help us with what products specifically are surprising you on the upside? And why that's happening?

Yes. I can take that. And yes, first, as I also already highlighted in my remarks, I mean, what a difference again, a year can make we see in every [indiscernible] first. And just today, I had a large mining company here in [indiscernible]. And the first thing what they said is your agents to accelerate the migration and the transformation with Wise are really outstanding. So Joule for consultants, all our data migration agents, they're up to the roof and we get very, very good feedback. Also, by the way, by all the partnerships we are closing in that space as well, and we have a good partnership with Conduct and many others. And also the SIs are helping us to even disrupt their own business model and accelerate the migration of our SAP customers to the cloud. Second, the business case. I mean, every business case is now with AI, I said it again, the majority of our deals that goes with AI. And even more important, the business data cloud after 2 years now being in the market, Reltio and Dremio acquisition is spot on. I just come from a 2-days engineering workshop. And what I have seen there is outstanding. Now in Q1, we will be able to even also generate non-SAP data products, put it into the knowledge [indiscernible] and we can extend the ontology layer, almost automated also for non-SAP data. Also Reltio, I mean this AI matching with master data records, be it the customer, the material, the supplier, we are on that will come very soon. And that already creates now really a huge excitement and BDC is just a foundation. And customers understand when I have really a messed up data model also for non-SAP data, I have to work on that. Because we also see with AI adoption, obviously, we can adopt AI very fast on the customer side, but it takes longer if you have a messy system. And that has nothing to do with our AI or with the technology. It's just the same like with every ERP implementation, you need still a clean corn. So with that, the BDC is, I would say, this outstanding business, and it's great. And then last but not least, as we said, in Q4 now it's prime time for all of our new AI capabilities, I mean, the excitement inside SAP about Joule Work is fantastic. I mean I love to hear all the feedback from our employees. And no matter if you were building PowerPoint for finance, for HR, for workforce, in sales, creating and of customer briefing deck. This is all done now with 2 work. And again, it's not only about another copilot who creates all of the spreadsheets and PowerPoints automatically. It's also giving you always the data context and the governance, about who is allowed to see actually rich data, and of course, all the identity and the authorizations also sitting in our underlying system of record. So that's actually explains it very well. But probably you heard it. I was also very keen this time to share our line of business wins because now we always said the sweet wins, but now we can go into every customer and says, "You know what? You can build the ontology in pieces with your best-of-breed landscape or you can have [indiscernible] out of the box with an end-to-end ontology layer coming with the suite." And that is also the value proposition for every app in every line of business is also now way easier for us because we can always sell the end-to-end oncology layer, which only SAP can provide. And when you then say, hey, this ontology comes as a service and governed by SAP, this is really huge. And then finally, also with my engineering team, I mean, we are fighting hard, it's always good to stay humble. I mean, there's, of course, also things we have to still improve. I mean one thing what is really outstanding in my eyes is that we are really also making sure that we are not uploading any data or process IP of our customers into these large language models. I mean, which customer wants to see their own process IP, their own data trained in with these large language models and being available to the public. So this is very, very important. And the last but not least, obviously, now in the next months and quarters, we continuously work on the response times of the agents of Joule Work, the accuracy. But all of that together really delivered already a strong quarter, and we see it also reflected in the pipeline. And also what makes me now so confident is also, again, when I see on the product side, the situation now compared to 12 months. Not everything works out of the box. Again, we have to work on certain things like agent response times accuracy, but it's so much better and the customers see the value proposition and that is also now reflected in the numbers.

Our next question comes from Charlie Brennan with Jefferies.

Actually, just to follow up to your previous answer there, Christian. You talked about needing to continue to improve the agents, both in response times and accuracy, can you give us a sense of how far away you are today from where you want to be? Is that a matter of months to close? Or is it a longer body of work than that? . And I guess related to AI, I'm getting partners beginning to fill the push towards AI credit to the business model. Dominik, can you shed some light to whether that influencing some of the metrics in the P&L yet and the dynamics between perhaps what you can book a CCB and what ends up in revenue?

Happy to take the first question, Charlie. Look, probably on accuracy, we will never be done. I mean -- and transformation is always also a mindset change. And -- when product managers come to me and show me in our daily agent review, "Oh, I have an accuracy of 95%. I said, "Oh, no. you raise the bar on your test cases because we want to then solve even more complex problems." I mean that is what SAP is known for. So we will -- we continuously work on accuracy. And again, when I look at the new harness, when I look at the knowledge graph, when I see how we are training those models, I mean, this is, again, a big difference also compared to 12 months ago. But of course, there's always work what we have to do. Response time is also something. We are trading every query down to the data layer, to the ontology layer, we are tracing and how responding, how good are the APIs? Do we need to extend the ontology when agents are going into the loop. And again, this is also a mindset change. I mean this is really so different to -- compared to what we did in software development, but we are making good progress. And I see everyone is so energized to move into that and build the best agents for the industry. So there's always work to do and yes, but we are making great progress there.

So maybe, Charlie, on the question of the credits and whether there's any impact visible, I would say, first of all, it's very important to note that whenever we give credits to a customer, we basically amortize these [indiscernible] accounts spread over the life of the contract. That means that you don't have to worry about some [ eclipse ] or anything happening on that front. And if your question more referred to this more consumption-based approach, which we want to gradually ramp into the mix, that is a very gradual process. So I'd say, in the near term, this kind of logic, how to bring CCB down to cloud revenue is still quite intact. Yes, it will loosen over time as we are going into more and more hybrid models all the way to full consumption based. And as we want to sell outcomes to customers as opposed to tokens. And for outcomes, it needs consumption and adoption, which also means that all our incentives are gravitating towards more adoption. That is not something you should be concerned about in the bridge to '27, this will be still quite traditional, I'd say. And by the way, it was also good to see that maybe our famous transactional business, which has been lingering along for a long time has now bottomed out a little bit and we have actually seen a pretty nice growth year-on-year in that bucket.

We'll move to our next question from Jackson Ader with KeyBanc Capital Markets.

The one that I had was on regions, EMEA and APAC growing again on the in cloud revenue growing again ahead of the Americas. And I'm curious, as AI becomes more of a part of the customer decision making, do you think that so the other regions, right, the Americas could start to catch up to EMEA and APAC in terms of their cloud revenue growth. And as AI might be being adopted faster in 1 region versus others?

I mean, no. I mean when I look at the regional performance, both on AI and data, I mean the regions really performed equally strong. Yes, there's 1 quarter where you might see an outlier. But I would say, when you look at it over the last 4 quarters, there's really a great consistency and also especially in North America, I would even say the customers are really oftentimes the front-runner for AI. We are in many AI projects, especially on industry AI, where we are now codeveloping with customers where we open from a graduation perspective anyway, more easier than when I compare this to all the regulations we find in Europe. So I would say, no, it's actually a U.S. being the front-runner and EMEA and APJ in the meantime, falling really well. So there is really also a great consistency in the performance across all regions.

[indiscernible] the front runner, they are also much more penetrated on clouds already. So I think that is the main...

Our next question comes from Mark Moerdler with Bernstein Research.

Congratulations on the solid quarter and the numbers and details. Dominik, I've been talking about current cloud backlog. Maybe I want to focus on the way you're thinking about guiding to the Q4 and the total revenue, you've given us numbers for FY '26. FY '25 grew at 11% cost currency. First 9 months has grown faster than that. Is it just simply there's just so much that the close obviously helps CCB, but a lot of what closes in Q4 is really not going to necessarily generate revenue in Q4. So what's the factors that are going into the bridge between the 9 months and full year '26 revenue guidance.

You're absolutely right that our visibility for Q4 in cloud revenue terms, which is, of course, the key driver of total revenues is quite high at this stage. And this is why we now say it's now towards the high end of the range and then you can back-solve the implied cloud revenue growth and you will find it's around 23%. So there's no magic. Now you might wonder why is it actually 23% given that we jump off a pretty healthy 25% in Q3, and there are a couple of reasons I want to call out, some I've already mentioned in my introductory remarks, I want to recall that Q3 last year, we had [ EUR 0.2 billion ] of 2 factors that kind of weighed on cloud revenue in that quarter. So it makes the comps easier. Secondly, there is still a little bit of a kind of quarter-on-quarter decel effect by the fact that smart recruiters was actually embarked in our P&L 11th of September last year. So we have a tiny kind of Q3 contribution. So there's still some good growth from M&A, so to speak, in Q3 of '26, which will be -- the comps will be out. So we will be -- we'll lose that kind of edge, but it's, let's say, 0.5 percentage point or so. And then last but not least, don't forget that we have seen the decel in the CCB growth from Q3 last year, it was 27% in Q3 to 25% in Q4. So it's quite natural that we see the remodeling effect. But you're right, I mean, there shouldn't be any massive surprises on cloud revenues anymore. There's always some customer stories where some customers might recover from a certain situation or go in a certain situation, which is adverse to revenues. This noise is always normal. But otherwise, it's a quite clear line of sight for Q4 on the cloud revenue, I'd say. The other items, licenses always notoriously difficult to predict, so don't name me on that one. And -- but it's getting smaller and smaller. And then we mentioned that we had a minus 5% constant currency decline in services. And so it's the EUR 1 billion or about run rate per quarter, and that's another fact to that, yes, will be there probably because we really have good experience now and think that this conversion of billable hours into adoption and support of customers is paying offensively straightforward. You get a meager margin on an hour you sell, but if you can convert that into recurring high-margin AI data business, and we feel there's more value to be had from that.

We'll take our next question from Michael Briest with UBS.