Good morning, everybody. Hopefully, you can hear me. We'll make a start just as the final seats are being taken. Christian Klein, CEO of SAP, thank you very much for joining us this morning. I must start with a safe harbor statement, let's get that out of the way while the final seats are taken. During this discussion, SAP may 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 risk and uncertainties that could cause actual results and outcomes to differ. Additional information regarding these risks and uncertainties may be found in filings with the SEC, including but not limited to the Risk Factors section of the annual report, Form 20-F. With that out of the way, we can start. Yeah, thanks for joining us in Paris.
It's great to have you here again.
Thanks for having me, Ben.
Why don't we start on the current environment? It feels like the news flow is as persistent as ever. A year ago, we were talking about the impact of the U.S. trade tariffs on the global economy. This year, it's more around the Middle East conflict, commodity prices. With that in mind, how would you characterize what you're seeing today, the demand environment, compared to what we saw this time last year with those different sort of macroeconomic factors?
Yeah. Happy to share some insights. The good piece is when you're running 400,000 customers and they are transacting with your systems, you see also a little bit the macro and how the prices develop. Definitely what we see now also because of the conflicts in the Middle East, prices are going up, and I feel in many industries now the prices are factored in also for the remainder of the year. We definitely see that the inflation here and there is going up, which of course some of our customers don't like. The second part, of course, is while we are not seeing so many supply chain disruptions, we are still seeing reshuffling of the supply chains on how to actually avoid the conflict in the Middle East.
Of course, we see a few shortages, nothing yet serious, but when you talk to many C-level executives, they are definitely saying, "Hey, if this continues for two or three months, I can't guarantee that I don't see bigger disruptions in the supply chain." While I feel everything is okay right now, I guess everyone would agree to the statement that we should come to a resolution of this conflict rather sooner than later.
I guess you alluded to in your guidance for the year that the assumption that there was a sort of near-term de-escalation or resolution in the Middle East. Here we are, first week of June. Feels like we're close, but not there yet.
Yeah.
As you look at the year, do you have a sort of line in the sand where you think, "Okay, this may now start to impact customer sales cycles or decision cycles?
No real news compared to what we shared also at earnings. I guess everyone is waking up in the morning, looking at the news. Is there a solution now or not? Sometimes I feel better and my mood is going up. Sometimes an hour later, there is the next news. Not yet come to an agreement. It's pretty shaky. What I can say, obviously, the sentiment in our business, coming out of Sapphire, we had record attendance, and obviously we had a very positive impact on pipeline out of Sapphire. We have this every year, I would say this year was really extraordinary. From that perspective, yes, the business continues to develop well. Again, of course, we are not completely immune against everything, what is happening, especially now in the Middle East.
Makes sense. As you sit here today, obviously, there's a lot of noise that is out of your control in terms of how the year pans out. As you look at the year ahead, what's in your controllables? What are the key variables that you think that you have real influence on that sort of dictate how this year pans out and really into 2027 as well?
I guess what everyone is looking for, and we presented now the first results at Sapphire, that we make business AI real for our customers. That is our focus, our complete focus. There is a product dimension to it, but there's also a go-to-market reality to it. We are forming new partnerships with companies like Anthropic and OpenAI and many others. It feels like a little bit seven years back when we formed the partnership with the hyperscalers, and now we are forming the partnerships with the LLM providers.
Starting with the product, I guess what is very important, what we announced at Sapphire, you can build with any LLM you want on our new AI platform, what is very, very important, and every customer tells us the same, how can these agents better understand the context of my business, better understand the context of my data? Why I build some custom agents, maybe sometimes even without SAP, oh my god, I completely underestimated all the governance, all the regulations these agents also need to understand. It's not only the context layer, it's also the governance layer. You can imagine what's going on in SAP and in our all-in-on AI project. We are building now ontology layers for every industry, for every line of business.
I get daily reports on the accuracy of our agents, how well do they perform, how well do they understand business processes. We are plugging into them into some of the most complex systems in the world. To understand, do they not only perform in a clean system, but do they perform in real life, and do they understand the governance? That's what we can influence, and I guess there is no other company who understands all of that better than us. Plus, obviously now on the commercial side, last week I was together with Dario on the Anthropic team, and we discussed our partnership options. First of all, how to deploy agents. You need to understand business processes. You need to understand system landscapes. What about consulting? Is there a way how we can work together?
I hope there is a way, and there should be a way, because again, a technical agent cannot perform if you are not also implementing the agent in the right way. The second piece, of course, is the commercials. We code software with Claude Code, with GitHub, with others, but I also look at the outcome. I said, "Hey, I cannot pay any price you want. I need to have this in relation to the outcome." The same discussion I have with my customers. Yeah. If Joule now gives you out of our CX a complete recommendation how to price and package this deal, that's great, and customers will pay for that, a premium, but not any price. Yeah. That needs to correlate. That is, of course, with many of the LLM providers, where we are still, of course, negotiating and forming our go-to-market partnerships.
The good piece is we have built into our agent builder a model where we can say we can switch from one model to another. We are not only doing a switch, we can also have a logic in there where we route the prompts to the model who has the best price-outcome ratio.
Yeah. We are not just routing it blindly to one LLM and then good luck. No, we definitely want to also build in certain intelligence into that which correlates to price and, of course, the outcome of the model. That is the things, Ben, what we can focus on, and you can imagine that's also personally where I invest a lot of my time. Yeah.
That makes a lot of sense. We roll back to Sapphire, there was a lot of new innovation that you brought to the market. Two things really stood out for me. The first was that SAP Business AI Platform and then the Autonomous Suite. If we look at firstly at that Business AI Platform, to me it felt like we're spanning and combining developer tools, data, and then finally AI and agents. Perhaps you could just distill for the audience today, what was really new in that versus repackaging of what you already had, and why is this so unique to SAP that perhaps others can't replicate?
Yeah. It was definitely more than a repackaging, and all of our top experts in the company who have not seen their families for the last six months, they definitely know what I'm talking about. Look, I wanted to have a Sapphire where we are not only telling a fancy story but giving proof. There was on the show floor and at that record attendance, we invited all our partners and customers to build agents in the Agent Lab. They definitely realized, "Oh, I mean, that's a new SAP," because we also had our learning curve, and the former agent builder didn't have any of the stuff what I was just talking about. Yeah, you build an agent with Anthropic or any model you want, you still need to stitch together the business data, the business processes.
When you're building agents, it's a great experience. First low-code, no-code or pro code, whatever you want. If there's a business person, happy, build your agent. If there's an IT developer, build your agent. The agent builder tells you want to have this pricing agent. This data sits in the Salesforce system. This data sits in the Business Data Cloud. This sits in the platform. Can I help you to connect that data? There is also a piece of ontology missing where you need to connect data, because otherwise I cannot actually combine the material of this product with the pricing of that product. The second one is, when you are now doing Configure Price Quote, I as an agent tell you, there should be also then a logic on how you do pricing today.
There are certain steps in the business process. Oh, there's end-to-end who can connect you to a certain kind of different process step, what you are having in your company in specific. Suddenly all people said, "Oh, now I'm getting when Christian is talking on stage about context, now I'm understanding why I should probably not build these agents on a third-party platform." You can build it on any LLM you want, but there is something what SAP has, what the others don't have. We actually also showed them the one part and how you can certify an agent. What access does the agent have? You can tell the agent you have read access, you have the right to change, or you have the right to even write into the SAP system.
I mean, who has the right to write into an SAP system? You can also define the autonomous level of the agents, and that is, of course, the new platform. You can imagine there was formerly a platform called BTP, then there was the Business Data Cloud and our AI foundation, and that was engineering work to bring this all together. Plus building now the ontology layer and that's the new way of working. We will also set up our development teams. It's not anymore these teams of 10. It will be team of five, more agile. You prototype an idea with vibe coding with one or two days, then you turn around because the validation brought you some insights which you need to consider. Then you go into the automation side where we can automate a lot on code generation, agent development, testing, et cetera.
The data scientists play a more and more important role because, of course, our agents should be the smartest, and that's why the ontology, the data, the process logic is extremely important. On top, based on the platform, we are now then launching our new agents for finance, for supply chain, for CX. I also don't like to always name them by LOB, so we have a business process because a company doesn't run in HR, finance, supply chain. This one's core. We showed them how on the new platform we can now run businesses with the autonomous suite. We also told the customers, "Hey, I know you still have a lot of legacy with SAP, but we help you to bring costs down. Here's Palantir, here's Accenture. We have our own AI assistants for ERP migration.
We can help you to accelerate the journey because while you can build these agents even now on ECC systems, when you are on the RISE journey, definitely the agents are way smarter when you have your data layer under control and if your process layer is simplified. That was the story. Yeah.
On top of that Business AI Platform, you alluded to it there, but really the Autonomous Suites, that feels like that sort of new layer on top. You announced lots of changes to Joule. Your teams kept calling it Joule 10.0, not Joule 2.0.
This feels to me like the first step from single domain, single task-based agents to the SAP's first step into agent orchestration.
It's actually interesting what's happening inside SAP. When we are now doing our reviews, and customers are coming with new feature requests, we also asking now the customers, "Why don't we build an agent for that? Why can't we build skills for that?" You see that more and more work is now also shifting from the system of record into the agentic AI layer. The good piece is, we're having native access to the system of record. What I just shared with you can decide what the agents are allowed to do in the system of record. More and more is shifting now in the Autonomous Suite to the agentic AI layer. While we coded, in the past, a lot of features, now we are shifting towards coding agents and skills. Feature development will not be come down to zero.
There are still a lot of complex business scenarios where a customer says, "I don't want to have an agent for that. I just want to see that a system of record does this out of the box." You see this is more and more shifting. With that, also our development capacity is more and more shifting. We are doing this, of course, in close cooperation with our customers, who also are very good indicator on how much do we put into the AI layer and how much do we still code features in the system of record.
I'm particularly interested in the economics of this. There's a lot of innovation here, but this audience, your shareholders, want to understand how the economics work. One thing that stood out to me, it was part of your Sapphire presentation. If we take Data Cloud, technology platform, and AI, these consumption-based revenue streams, there was a slide there that says that will increase in your cloud revenue mix from around 10% today towards 30% by the end of the decade. That, by my calculations, is a growth of about 50% CAGR.
Quite exciting numbers. I think the first question to investors' minds is how much of that is on top of, as well as incremental versus you have a huge baseline of subscription there. Is there an element of offsetting or cannibalization, or do you truly think that that is largely incremental?
Yeah. There's, of course, looking at the capital markets, there's a lot of debate about end of software and so on. Obviously, I'm a big believer that there will be still a system of record because somewhere you need to store your data, somewhere you need to store your business logic, et cetera, the regulation. That's why I also said, I guess another important message on that slide was that over 50% of our subscription revenue is already non-user, non-seed related. We price today already a lot outcome based. That was actually a big request from our customers. Don't just price per user, price a bit more outcome based. It's not like that the subscription revenue is almost not outcome related. Now, on top, and that's why I really say on top incremental, why do I believe that?
First, I look at our pipeline. Still there must be a system of record running your business. Now you can say maybe in certain parts of the software world, for some niche players, you could say, "Okay, the LLM can do the job." When I look at our portfolio, definitely I don't see that now for the 95% of parts of our portfolio that you can just say, "Oh, it's so easy now. Let's just use an LLM, and the LLM does the job." No. I'm not seeing that this will go away. I rather see that we will sell a lot of agents. Of course, we will always lead now with AI, but with that comes the system of record. An agent doesn't work without a system of record.
We are now telling our sellers, "When you go in, don't show them the fancy Employee Centra l system of SuccessFactors. You are showing the agent who does the workforce transformation. You are showing how the agents run the payroll." The system of record is, of course, priced in because without a system of record, the agent has zero context. That's why I'm not saying the one revenue stream will substitute the other revenue stream. I rather see this as incremental. Of course, will be there a higher growth, a much higher growth on the agentic AI layer? Definitely. Definitely also the subscription layer, the system of record, they will still see very healthy growth.
When you look at the results we just posted and when you look at what we are indicating for the remainder of the year, I definitely see, and the pipeline development, the two pieces come with each other. Then, the AI migration tools, which all customers are now demanding because they are not willing to pay 10x anymore for every ERP migration. They are now of high demand, and that's completely incremental. We are not a service. We have a services business, but our services concern focus more on adoption.
Yeah
not so much on the traditional services project. That is incremental. With that, and now I hope we can excite our partners to also build agents on top of our new platform, given all the differentiating capabilities I just outlined. When you add this up, I actually see there's definitely not a lack of growth opportunities for SAP.
One aspect that was interesting you talked about at Sapphire, maybe offering some of the AI build consumption, so that initial stage of consumption to get customers up and running, maybe helping customers with that. Could you perhaps unpack that a little bit? What are the limitations around that or parameters you work with to just ultimately drive adoption further down the line?
First, yeah, we offer the development of agents for free, but that is also, for us from a commercial perspective, makes total sense. Anyway, customers want to see value. They want to be convinced that the agents provide outcomes. Then, of course, when we come to the one and the operations, then, of course, we are starting to charge outcome-based. We also said we make the one part for free until the end of the year, next year, we are going to charge. I guess that's now very important that we get our customers, the use cases, the references where they say, "Oh my god, these agents really provide a lot of value." When you are building the community, and it's not only about the customers, you need to build a community also in your ecosystem.
I want to have the Accenture SAP consultants, the millions of SAP consultants running around in the world. Don't code any more features or customizations, code the agents on the SAP platform. To get this going, we also made this offering, which was very well perceived. Now with the Agent Lab, we also have a great landscape where we can train and enable them. Let's not forget, this is also for everyone new on how to build agents, how to deploy agents, how to implement agents. We are also investing into the training, the demo, the enablement part, to really build this community. We have TechEd coming up, and we will do a hackathon, we will do a massive AI agent development together so that they are also then getting used to the new SAP and how we're going to run the customer's business in the future.
When you're in discussion with customers, one of their priorities is visibility, and Dominik, like any CFO, would want visibility on what his costs are.
When you introduce a consumption model, it brings an unpredictability.
Do you find customers are content with that new economic model? Are they asking you for at all a fixed price, almost all-you-can-eat scenario that you then offer to them? I'm just curious how those discussions are going.
Of course, they ask. Look, I'm not even saying we can't do a fixed all-you-can-eat contract in AI. Absolutely impossible. You can, of course, say, "Hey, in the next three years, you can buy a certain AI consumption." When you are then overconsuming it, there must be a threshold, yeah, where you then need to go to the next level. At that kind of models, especially for our large customers who use SAP end-to-end, I can definitely see that, but there must be always a consumption element, even in an all-you-can-eat. You cannot just offer fixed price. Absolutely impossible. Did we do this in the good old on-premise days? Maybe, but absolutely not possible in the AI world. Of course, customers are now still also learning. They say, "Okay, I have your calculator. I see what it costs.
What is the outcome?" There, really, our Signavio and AI process mining capabilities play a big role because I still had a customer this week, they said, "We go all in. We are replacing NetSuite, we are replacing that. On your AI, before we deploy it, still show me before and after. I want to see before and after.
Right.
I going to adopt all of your agents. No doubt about that. It's very important in an outcome-based business model that you show them what is getting better. The good piece is also now dealing with Anthropic and OpenAI, and others on the commercials. The customers are absolutely willing to pay a premium if they see the value. If you can prove that, we definitely see there is a healthy uptake of ACV we now can sell by customer. Again, if we can prove the agentic AI layer, drives a lot of incremental value for the customer.
Yeah. Interesting. If we maybe take a step back in the customer's journey and go back to RISE migrations. It wasn't so long ago it felt like the message from SAP was, "Look, customers, hurry up because this innovation train's running away." This year, it felt maybe a little bit of a change in direction there and being a bit more accommodating. Looking to reduce some of the costs of services, help them maybe give some AI functionality when they're still on ECC, again, providing they commit to that. I think the market's perception of that was either maybe you weren't happy with the pace of those RISE migrations or worried that customers were perhaps pushing back to them.
Yeah.
What was your perspective? Why make that shift this year?
Yeah. Definitely, of course, there was a lot of focus on AI, but we always had RISE & Co. as a key element. Again, RISE & Co. is a transformation office. We tell our customers, "Don't do an ERP migration just for technical reasons. If you want to see business outcomes, you need to transform, and the business needs to sit on the table, and we help you with that." Now with AI, of course, it's not enough anymore to only move from the legacy ECC system to our modular cloud applications to the suite. Now, of course, imagine I would run now into my supervisory board and tell my supervisory board, "You know what? I'm modernizing SAP with some fancy SaaS application." They'd say, "What? Where is AI?" Every CEO gets that question.
What we're now saying with RISE is, while you are modernizing your landscape, and you should absolutely do it, because when you have a broken data model, you will see all kinds of challenges, and we are investing into the data layer in the cloud, also with some recent acquisitions. We also then say, in parallel, you don't need to wait now two years until you have completely modernized your system landscape end to end. We are helping you to adopt AI from the get-go, by putting in some forward deployed engineers and by making our AI agents work. We build some on-premise connectors where now our new AI platform can connect into calling the APIs into an ECC system.
Because I just had this morning a big French customer with me. Also, he said, "Christian, we are doing this and I'm happy that we are doing this. I see that my plans are going live and they are way more productive than before, but I can't miss out on AI. Can we do something for the systems which are not yet in the cloud?" That was the message at Sapphire. We don't want to have you waiting two years until you have finished. We do this right away together, so we do both.
You can bring forward some of that return on investment during the journey rather than at the end of it.
Yeah. The risk, of course, would be if we don't do it, maybe some others will do it. Maybe, hopefully, with less quality, given all the context, what we have from these businesses, but of course, we don't want to miss out this very important move of also building now AI for some of the legacy systems.
I guess ultimately, this strategy here of reducing the time or the cost of migration. Again, this new partnership with Palantir and Accenture looking to deflate the cost and shrink the time, plus bringing some ROI a little earlier in that journey. How does the market judge the success of that? How do you judge the success of that? What do we look at? Should this accelerate the number of RISE migrations? How do we think about that playing into the financials?
I have three KPIs, which I'm looking at. First of all, acceleration of RISE journeys. Will the customers move faster? Second is, of course, adoption of these AI migration tools. I want to see how much customers and partners are spending and using the tools. Of course, ultimate currency is then outcome revenue-wise. It goes from acceleration into adoption. Adoption hopefully will drive revenue. I'm a pretty number-driven guy, so I get my weekly reports, and I challenge the team always to say, "Okay, the enterprise architects, did they really make our customers aware of that they should now use those tools?" I tell all of our SI partners, "Come with us to the new world.
Train your people on building agents on the platform, provide innovation, but don't make less money with ERP migrations. Customers, again, don't want to spend that kind of money anymore with AI.
We had Databricks here yesterday. You have an important partnership with them. It's a key tenet of your offering. It's been 12, 15 months, I think, since the partnership was announced. You've since announced some similar partnerships with Snowflake and with Google Cloud. What can you tell us out there in terms of commercial momentum, how that partnership's evolving?
In my eyes, very positive. You can always ask for more. Ali's asking for more, I'm asking for more. I feel we find a good balance there to have really a win-win-win. The customers, first of all, they have that data strategy, and they love the fact that they can now do zero copy, they can actually harmonize data. We, of course, are very keen on keeping the oversight on the semantical data layer. In BDC, you can now not only create data products, which is a data object that's also available in the BW system, but you now can also join data. You can build intelligent data modules, which then comes into play when the agents need to do smart predictions. They find in BDC now our tabular AI model. They find a semantical model they can work with. Obviously that works really well.
The partner connect to Databricks, to Snowflake, actually enables our customers to share data, zero copy, and build a joint semantical data model. I can tell you, no matter if it's Google, Snowflake, or Databricks, everywhere I go, I was just at the Choose France Summit, and I saw someone from Databricks. They're all loving it. Yeah. Proves on the one hand side that we are sitting on data, which is absolutely key to build great AI. On the other hand, we also are super sensitive of when it comes to our IP. We have certain debates with our lawyers and say, "Hey, it's just not okay." Data sharing is absolutely okay. The customer's data is the customer data. We want to be open. That's absolutely good.
If a few companies believe they can steal our IP by going too deep and actually extracting the process logic, the data logic of SAP, this is absolutely not okay. This is IP we developed over 50 years, and that's what we are selling since over 50 years. There's definitely a borderline we will not cross. Otherwise, we would be completely dumb. Second, of course, a borderline which we also want to protect. Yeah. Because this is what we do. We run businesses. We develop this business logic, and that's why we send out also a few love letters to some of so-called partners where we believe, is this really still a partnership? That's the new world. I really want to underscore, we are open. The new Business AI Platform, the agent orchestrator for free, you can bring in third-party agents.
We don't want to monetize the collaboration of those agents. We have our APIs, we have our MCP servers. We will offer that on the platform, but it's very, very important that the platform is open. The only piece is, of course, the system of record, and there sits the domain know-how, what we developed over 50 years, and that's SAP's IP.
Can we shift gears slightly and look at on the cost side, on the investment plan? Since Dominik arrived as CFO, OpEx has effectively been flat for three years.
With revenues accelerating, you've delivered a lot of operating leverage as a result. You're now actively signaling that cost to revenue growth ratio for next year. That looks like over the next three years, quite a considerable step up in gross investment. There's a few puts and takes that I'd like to unpack. Maybe let's start on the AI efficiencies internally. Sebastian has talked about delivering EUR 2 billion of savings internally from AI. Could you give us an update just on the progress there and specific examples of just how you can extract that much cost savings?
Yeah. We had, last week, a board offsite, and I guess my CFO sometimes values that the CEO is, of course, very focused on the growth of the company and on spending in the right places, but also very focused on being very cost-disciplined in some other places. Obviously, we discussed the more than EUR 2 billion AI efficiencies last week. I told, for example, now on the development side for the R&D budget next year, I said, "Hey, I'm all in for using Claude Code and GitHub, but what is the output productivity-wise? Where do I see the acceleration outcome-wise on our roadmaps? How many more agents do I get faster?" That is now the discussion.
We want to definitely commit to our expense-to-revenue ratio, that's why we are all in for deploying these AI use cases, they need to also then be related to outcome. The outcome has also, of course, a productivity level to it, not only a growth level. We are now going through that. I'm actually very optimistic because the way how Sebastian is driving that, this is not only an IT CO exercise. All the business leaders are asked.
to deploy AI. We have a massive enablement program running that will double down now into half year two, so that we are training all of our people, no matter where they are in the company, to use these AI use cases. We also have a clear measurement on outcome and not just adoption of AI, so usage of AI. That's what we are now factoring in. Interestingly, with many of our customers, it's the same. They're saying, "Wow, my costs here, they are going up on the LLM side like hell.
Yeah.
How can I relate that to more commitments on the business side?" I believe this is not only an exercise what we are doing, this is an exercise what many companies are doing right now.
We take that roughly EUR 4 billion of sort of gross OpEx increase. There's EUR 2 billion of savings. We're talking about an increase of mid-single digit, maybe towards EUR 6 billion that we calculated and you guys alluded to at Sapphire. Feels like a big number. It's about 50% more than what you spent when you sort of invested for the cloud transition era five or six years ago. Where are those key wallet spend, where is that really going? Do you have that sort of penciled in a little bit already, or have you left yourself room that if you feel the need to accelerate certain areas, you can do?
Yeah. I'm thinking back six or seven years ago. Sometimes what happens with companies who are maybe too successful and are dominating a market pretty well, you're getting complacent. You are hiring people which are maybe not always up for the job and not playing Champions League. Barcelona needs to win. They need to have the best of the best. I compare this sometimes to SAP, and they got a lot of these investments now, what we are planning to do. We are not so hardware intensive. We will use the LLMs which are out there in the market. What we need to have, we need to have the best of the best when it comes to ontology, when it comes to training the models, when it comes to building agents and understanding the most complex businesses in the world.
When we now acquired, or intend to acquire, Prior Labs, Frank Hutter, the co-CEO, said to me, "Christian, I will not hire 100 people, but I will hire a few, but they can cost up to EUR 400,000-EUR 500,000." I said, "Do whatever you want, as long as the Tabular AI is the best model in the market," because that's so key for how Joule and the agents will perform. I will definitely now double down, and I told them, what kind of people do we bring into the company, what kind of background, and we will definitely increase the cost per hire. Vice versa, obviously, that's why we are so confident to stick to our commitment. Of course, it's not so much more about the quantity of developers anymore.
You need to invest in certain places, and you need to invest massively now in the next 12 months also on the enablement side. You will not drive AI adoption internally and externally if you are not investing into the enablement of the employees and the partners. That's, of course, another cost bucket. Last but not least, obviously, that the workforce is key. We need some flexibility to adjust our workforce, to make sure the workforce is actually there to lead SAP in the new age of AI and having thriving success both from a top line and a bottom line perspective. Workforce is absolutely key for the workforce transformation.
We're sat here in Paris, in continental Europe, the topic of sovereignty.
European sovereignty. You attended the Choose France event on Monday.
Yeah.
How real is that? How much of a discussion is that with customers? We see the enormous numbers out of the U.S. frontier labs, the hyperscalers. What's your sense of how real European sovereignty is and how that really impacts your business?
Yeah. Look, sometimes I'm not the most famous person with some of our partners when I'm saying, "Hey, we should not now over-invest into sovereignty in a way where I'm not seeing how customers are making decisions." Sovereignty is very important, for sure, but we are having sometimes so theoretical discussions here in Europe about sovereignty that I don't even know what do we mean about it, and everyone is getting a little bit confused and putting money into stuff where I feel it's just not really the sovereignty we need. What I can absolutely say, and President Macron highlighted that and I had to smile because he knows how I'm thinking about that. In Europe, in order to lead in the industries where Europe is leading today, we need to apply AI to make our plants, our factories way more productive. We can't compete on labor costs.
We can't compete in some countries, like my home country, on energy costs. Building now the industry AI models is absolutely key, and there are a lot of startups which we definitely should help to get one of the best of the best startups and companies when it comes to applied AI. With data center, we definitely also will need data centers. Who is the company who will run this data center? Does it need to have an own legal entity like Delos, where Microsoft puts in the hardware, builds the stuff, and we are operating it as a German company end to end? Absolutely, I would say this is a high level of sovereignty, probably the highest you can ever get, because we even decoupled the global network from this data center, and it's an air-gapped data center. There is not a higher level of sovereignty.
That is, of course, something where I sometimes believe it's not about U.S. tech companies or European tech companies, it's about what kind of sovereignty do we need to have. That we then need to invest into data centers and infra is clear, but we should probably focus a bit more our investments on the applied AI layer.
I am not only saying this because I am SAP. I am just not believing that we are going to see in the next three years another Anthropic in Europe. I do not believe. You have to acknowledge certain games are played, and now a new game is getting played. That is not actually bad for Europe because we have so many great companies here in Europe. Let us build together an ontology layer. I just was together this morning with a great manufacturer. I said, "I need you. I need your domain experts helping me now to build applied AI. You know what? If we are building this together, let us also go to market together.
Why not?" I guess that is where it's really decisive for Europe and rather focus a little bit less on, "I know it's cool to open up a new data center and it's fancy and everyone loves it," but we should really also sometimes focus on, "Okay, what does really matter for Europe?" I'm not sure if we are well-served to only now trying to build companies who do data centers or build chips or that. We need to have the best of the best startups when it comes to building the AI modules for the different industries we also need to be strong at in the future.
Fascinating. That's us all out of time. Christian, thank you very much for being with us today. We really appreciate it.
Thanks a lot.