Computers are getting smarter, Silicon Valley's getting weirder, and AI is taking over. At least that's what everyone keeps telling us. Welcome to Machine Gods, a new NPR podcast that will be your field guide to the AI-powered future. We're veteran tech journalists with front-row seats to the industry, and twice a week we'll make sure you get smarter about AI, so you know what's real, what's hype, and what might come next. Listen to Machine Gods from NPR on the NPR app, or wherever you get your podcasts.
Support for the show comes from ServiceNow. AI is moving fast across the enterprise, but without visibility, it's just chaos. Different tools, different models, different teams using AI in completely different ways. ServiceNow turns that chaos into control. With the AI control tower, you see all your AI across the business in one place. What it's doing, what it's done, and what it's about to do. So you stay in control. To put AI to work for people, visit servicenow.com.
Support for the show comes from CrowdStrike. It's not a huge stretch to say that AI is the next major computing platform. Every platform shift changes the way we build software, and it changes security too. CrowdStrike is defining cybersecurity in the AI era with AI detection and response, or AIDR, a solution for seeing, monitoring, and securing AI across the enterprise. Learn more about how leading companies are turning to CrowdStrike to secure AI and secure their business at crowdstrike.com slash decoder.
Hello and welcome to Decoder. I'm Eli Patel, editor-in-chief of The Verge, and Decoder is my show about big ideas and other problems. Today, I'm talking with Roland Bush, the CEO of Siemens. Now, Siemens is one of those absolutely giant, extremely important, and yet fairly opaque companies that we love to dig into here on Decoder. At a very basic, very reductive level, Siemens makes the hardware and software that lets other companies run and automate their stuff. You've seen the Siemens logo everywhere, whether it's under the hood of your car, stamped on the control system of a fancy building, or scattered across factory floors. But since it's not really a consumer-facing company, it's hard to know what ties all those ideas together.
and what some 320,000 Siemens employees across the world are actually working on. How all of those people are organized and how they all work together is wildly complicated. Roland and I spent some real time just talking through the Siemens corporate structure, which for the true decoder heads out there was incredibly fascinating. Roland and I also spent a lot of time talking about automation broadly. And what happens is AI brings automation out of the physical world of factories and into the digital world of the front office, the world of accounting and procurement, the things that help decide what the factory should be doing. Roland's vision is for Siemens to automate the entire factory process, upstream and downstream of actually making things.
And you'll hear him describe that outcome as fairly utopian. Smooth, seamless, optimal operation. Very German. But I wanted to press him on how dystopian that sounds to a lot of us. Because in Roland's vision of pure automation, it seems like there's a lot of people who just don't have jobs anymore. And the ones who do have jobs don't really have a lot of autonomy or fulfillment from them, because they're basically just doing what AI tells them to do. So we talked about that pretty directly. And if that's not a lot of complicated, heavy decoder ideas already, well, Siemens is also a government and defense contractor on both sides of the Atlantic, and a company whose growth historically has been tied directly to free trade and globalization in the post-war era.
There's a lot going on right now that challenges how that world works, especially as tensions keep rising between the US and Europe. So I asked him directly, has Siemens gamed out what it will do if NATO collapses? Because that's not as far-fetched an idea as it used to be. As you can tell, there's a lot in this one, and Roland was game for it all. I think you'll leave with a lot to think about. Certainly more to think about whenever you see all those Siemens logos all over the place. Okay, Siemens CEO, Roland Bush. Here we go.
Roland Bush, you are the president and CEO of Siemens. Welcome to Decoder. Thank you, Nelly. Nice to meet you. It's nice to see you. It's nice to meet you as well. There's a lot to talk about. Siemens is a huge company. It has a long history. You've been in a lot of businesses. You've been out of a lot of businesses. You have worked there since the 90s. The world is...
very complicated right now. And Siemens is a very big, very complicated multinational operating in that world. I'm curious how you are thinking about that. Let me just start at the start. Siemens isn't a consumer company. I think a lot of Dakota listeners have seen the logo, but maybe don't understand the company. How would you describe Siemens today? What is the company? And it's indeed not that easy. We came a long way. It's more than 170 years since the company was founded. And we made, I mean, so many changes in our portfolio in our company. Actually, when people talk about it, I say there's one constant in our history, which is that we reinvented ourselves over and over again. And absolutely, we are now in the midst of another.
reinvention or transformation with one difference. This is the fastest and the most fundamental one we ever had because of technology. And then people ask, what is Siemens about? Because you have now Siemens Healthineers, you have Siemens Energy, you have Siemens. And actually, it's not that easy to describe because Siemens Healthineers has the task in the name. It's about healthcare. Siemens Energy has the task in the name. It's about energy. But Siemens is not that clear. So here is how I explain it. We transform with our technology the everyday for everyone.
Okay, doesn't get you closer. But no, the point is you have to look behind the curtain and then you see what Siemens technology does. When you see a car passing by, eventually all of them are touched by Siemens technology. Is it either cars are designed by our technology or they are manufactured? Every third manufacturing line in that world is run by Siemens controls. If you walk through New York, you cannot walk a block without passing by a building which is automated by Siemens technology. I think we are controlling, I mean...
Something like a little bit less than 50% of electrons are touched by Siemens technologies and our distribution systems, low-voltage systems. And if you talk about health engineers, then, I mean, if you get a scan somewhere in the world, the likelihood that it's a Siemens CTOMR scan is a little bit shy of 50%. And this is what we do. We produce, we have technology which enables others to transform their everyday, and that is what Siemens is about. I listen to that, and I experience Siemens everywhere. I'm the person who pays attention to how is this building automated. I talk to a lot of car CEOs. I hear about Siemens as a supplier to the car industry quite a bit. It sounds like what you were describing basically is you operate things for people, right? Or you build technologies or products that operate other things for people. There's a lot of things in the world to operate. How do you organize the company? How do you think about where there's opportunity and where there's growth and investment? And then how do you think about...
your resources there? Because it seems like we operate things for everyone. That's a pretty wide remit that you could focus down in any number of ways. No, absolutely. And here comes the point, and this is an absolutely valid question because now we are active in so many different industries. I mean, it's industries, manufacturing, process industries, but we also are in buildings and grids and in mobility. So people, trains and signaling systems. So the first basis of Siemens is, and this is where our value sits. It's in our technology platform. Is it our design software? We have one of the largest software companies in the world. If it comes to industrial software, we are the largest. And with our software, you can build the most comprehensive.
physics-based digital twin of whatever product you do. And we are now expanding into molecules. So another one is automation technology, as I talked about. Is it either for discrete process manufacturing? We also go for software-defined automation, which is kind of a disruption. And anything, we are the largest automation company. We're automating grids, we're automating buildings, we're automating signaling systems, we're automating trains, we're automating manufacturing. So the underlying technology is that what the value is. Now, we are bringing this technology to different verticals, so markets. Is it the industrial markets, food and beverage, chemicals, automotive, machine builders?
utilities, mobility operators, and the like. And then at this point, the domain know-how comes into play. So having technology is one thing, but having the domain know-how to deploy it to talk customers' language is another one. And on top comes, quite obvious, the whole thing is now supercharged by AI technologies, which we are rolling in as we speak. We have a long history regarding AI. Actually, one of the first supercomputers doing machine learning algorithms was Synapse I. I mean, this is in the 90s. Siemens had the most powerful one. This was the grand-grandfather of the GPUs today.
And since then, we're working with artificial intelligence technologies. But this is now obviously a new level which we want to bring it to. So the organization, obviously, we're organizing according to businesses. They are reflecting the markets we are acting in. But you have to look at it from the back, the underlying technology, including the data, which is super relevant. I'm very curious to talk to you about AI and automation. I think that's very important. I think Digital Twins, you've been in for a minute. I'm curious about that. It seems like the future of automation is very rich. We're moving from Siemens automating a lot of atoms.
automating the physical world to automating bits. And that's a long conversation I want to come to. I just want to stay focused on the company for one more turn here and kind of ask the decoder questions, because I feel like the structure of Siemens says a lot about the company itself. I was reading your last letter to shareholders. You're talking about how you divested portfolio companies where you weren't the best owner. You're exiting some businesses. You're obviously investing in others. How is Siemens organized today? How is the company structured? We are running according to businesses. One is digital industries.
which is all about the software, the automation piece. One is smart infrastructure. Here you find our building technology, medium voltage, low voltage, but also the grid, grid automation, grid control, grid control software. The third element is Siemens Mobility, where we have our trains, high speed, locomotives, commuter, metro, light rail, but also rail infrastructure, including turnkey projects, which is part of that. And the last one, since we are still consolidating to the Siemens Healthineers, which we hold still some 70%, which is Siemens Healthineers.
own listed company, DAX listed company, we're about to let go. So we announced a spinoff of 30% from the 65-ish kind of, but it's a separate company. So that's how we organized. Of course, we have our corporate organizations like strategy, IT, we run obviously horizontally. We have our M&A department, talked about our portfolio. So that's what a corporation normally has got. And maybe one more special thing is we still have research.
so we still invest uh out of our eight percent in terms of revenue or six and a half billion a portion of that goes into research advanced so we do we work on quantum quantum computer we don't build ones but we run on the software the middleware how to use the applications um and and we have also machine learning experts we're doing research there ai experts so that's roughly how the company runs and then talking about regions um we have in total including house now 320 000 people you have 45 000 sitting united states 30 000 in china 35 000 in india Roughly 85,000 in Germany. Still, it's a German company. There's a lot of manufacturing here. So there's a lot of people. Let me ask you a split between regions. I think a theme on Decoder recently that you were just a part of this trend is a lot of these companies are a lot bigger than people think. You described divisions. You described regions. Those are both potentially first order organizations for companies, right? I talked to lots of CEOs and regions is just the first order organization. Other companies are divisions. You have both. How do those interact? This is a constant discussion because we have very strong regional leads. We have strong businesses.
So this is a matrix, and every company has this matrix. And the first question is, which one is the predominant line? Is it the business or the regions? In our case, it's a clear answer. It's the businesses. So the businesses have the full P&L. Regions are the, let's say, the second derivative behind it. Still a very strong region. In some businesses, let's take, for example, our low-voltage business, switching technology, low-voltage. This business is...
fully run by regions. So you have an independent, you have China, you have United States, you have Europe, and these are, this is their P&L, so roll it up by regions. If you talk about automation, the next level is still, I mean, it's motion control for machine builders, it's factory automation, it's process automation, and then come the regions. So therefore, you still have a different kind of setup depending on the business, how we serve technologies. And then the other, the third dimension, just to make it a little bit more complicated for you, is the verticals. Because factory automation, take factory automation, which is maybe the strongest automation business we have got, they run into any kind of factory automation you can imagine. I mean, the discrete and hybrids, I mean, food and beverage, automotive, machine.
Machine builders is by machine control is another one. But then you have a lot of battery manufacturing and semiconductors. So therefore, this is the third dimension to surf verticals because each of them has a different language, different applications. So we are having a very... let's say, three-dimensional metrics, but there's a clear lead, and this is driven by the business lines. I ask everybody on Decoder how the businesses are structured, and the joke I always make is if you tell me how the company is structured, I can tell you 80% of your problems. But in the case of Siemens, it seems like I'm still trying to figure it all out to even get to where the problems are. When you think about that organization and you describe things like a common platform or shared innovation across these zones or the investments you need to make in AI, a lot of your competitors, they're new. They're essentially functionally organized. There's one person driving the business in the case of some of these startups. Siemens is very old.
It's organized divisionally, and then obviously there's layers of organization between it. How do you think about investing in the core technologies, the core platforms in that structure? Because it seems like all of your division should be doing it, perhaps in redundant or repetitive ways. You're right. When I explain my organization, you can identify the problems or the opportunities, put it that way. And here comes the point. I mean, I talked about investing.
6.5 billion R&D. And obviously, this goes into different businesses. And some of them have a higher share. This is an 8% on average. Some of them are sitting on definitely more than 10, 13, 15. Some others are more on 3 and 4. So capital allocation, is it R&D, is it CAPEX, also spendings? That's done business by business. Each of them has a business case and we allocate capital. In some cases, we obviously want to allocate more capital in higher growth areas. And I talk organic capital allocation if it comes to M&A. This is something that happens on the board. I mean, we come up with proposals and we see where we want to spend more money and where we focus on in our M&A and where not. But then here comes the point. And this is the part of our...
one tech company program. Actually, I started last year together with my fellow board members, maybe one of the most fundamental transformations of our organizations, because you're completely right, we are very much boxed. Below these businesses I talked, like digital industries, we have software and then we have automation. Automation are three boxes, factory automation, process automation, motion control. Below that, we have even segments, so I just didn't give you that complexity as well. So we're very much boxed. And what we want to do is, we are taking layers out, so we bring that into bigger boxes, so actually we are talking for six units. But then we also say we want to create
which is a kind of operating system where we have horizontal ones. So we have a data fabric, we have a technology fabric, we have a sales fabric. So the idea of NDS fabric is a thin layer, but it's a very strong one where we are really scaling horizontally as much as we can. For sales fabric, for example, we want to use the same tools, the same nomenclature for customers. And believe it or not, if I ask today how much revenue we do with BMW, people have to run out and put pile.
numbers together. This stops now because we have one identifier for BMW and with a push over bottom, I know what to do. So they have the same sales methodology. Also the customer journeys, they should all like. Technology fabric is that we don't do things over and over again. When we talk about a digital platform where we sell, we build it only once and sell our portfolio. So this is a change. We are unboxing our organization in two little boxes. And the reason is, number one is technology and AI in particular doesn't respect silos. AI doesn't respect data silos. They don't, doesn't respect any kind of boundaries. The world is squeezing out the small.
you see that this is adding more and more, and the more data you have, the more capabilities you have, the stronger you are. I mean, it's a fact. Look at the big companies. So therefore, we have to play the strengths of Siemens, and this requires a different way of running this company. This is behind our one tech company program, which is really putting in, I mean, horizontal as much as we can, yet respecting different go-to markets and different kind of technologies or applications of technologies, depending on the verticals we are serving. So not losing our strengths, which we have built over so many years, while scaling horizontally. Does it make sense?
It does. I've never been so excited to talk about structure with anyone as I have been with you, because that seems very hard. It seems like what you're describing to me is a multi-hundred-year-old company that has traditionally been very divisional, trying to get to some functional structures so that you can move faster. And that traditionally has come with culture costs. It has come with disruption inside the company. It's come with inertia. How are you dealing with that at a company the size of Siemens?
There's 320,000 people. They can't all be happy with you. Yeah, give a little bit of insights how we did it in the past and how we do it right now. It's our one tech company program. In the past, we had many restructurings and many changes. The point was a managing board. We have now seven people and we used to have 15 or more, whatever. In the past, these guys were going together with their strategies in a room, defining a new structure, laying out a new org chart and then, dear colleagues, this is where we want to go and we reshuffle.
You can imagine how that goes down. This time, we created a North Star, where we want to be. The North Star is basically what I sketched to you, these fabrics, the businesses, but allow them to focus on what they need to do, focus on their customers, on their applications, but yet we want to get horizontals into it. And then we said, this is a North Star, and here are, we call it tracks, the tracks to the North Star. These are the points where we really want to touch, for example, our CRM system for automation, which was completely scattered. We want to do that, and this is a blueprint for the whole company. It can go on and on. And then we engaged people to say,
Let's go. And you work with us now on these tracks and how to change it. That means we give the people an opportunity to contribute, to bring their ideas. But we have a clear idea where we want to go in order to move first. Secondly is obviously you need to communicate a lot. You have to explain what you do because you're describing something where people don't know where it goes. They're not experienced to work that way, but you have to talk it over and over again. Explain why we do it, what's the benefit and what changes for the people. And for some, they said, I don't experience any change. Yes, because we don't touch everything. We touch only things where you really can improve. And some, we just let go because why would you fix something that's not broken?
The next one is we definitely also train our people because transformation is not just shifting boxes. It's a different set of values. Collaboration is a much, much more important element in it. So we are putting a lot of emphasis in helping them also. And change is not only a structure, it's also the processes behind the way how you lead. And the last thing is, obviously, you also want to inject capabilities from the outside. Also on a higher level where you have people who know.
what good looks like if it comes to a super professional sales organization, if it comes to AI technologies and developing models. And this helps a lot. And we do that all over the company, on the lower level, the higher level. But if you're talking about really highest level, if you come with people who really have the gravitas, who bring the experience, where nobody would doubt that if they say, this is how modern software looks like, they have authority by themselves without giving them the stars and stripes. So they just put people together and say, this is how we go. And people listen and follow. So this is the package and you need all of them, all of them.
in order to make this transformation. And it seems to work. I had a huge respect. I mean, I'm a lifetime at Siemens. And I saw in my 30 years of Siemens so many changes in programs. And before pulling the trigger of making this big change, which is the deepest one for the last at least 20 years, you think about twice. Because if this vessel runs in the wrong direction, you have a problem. And we have to deliver at the same time. But I pulled the trigger because I knew that we have to change. The environment technology is changing so fast, so we have to be on the forefront. But I'm quite happy because it runs now for one year.
And good progress. There's another one ahead of us. By the end of the fiscal year, by October, November, we are basically through with all the big moves and the changes. And we are already grooving in the first batch of changes we made. Sales seems to work quite well. We're grooving in. And then within two years, I would say we finally are ready to scale. We're going to take a quick break here. We'll be right back.
Support for the show comes from Vanta. As AI adoption grows, so do your company's security risks and requirements. Most compliance tools promise automation, but you're really stuck doing all the work by hand, chasing screenshots, validating controls, and playing catch up when audit season comes around again. Vanta works differently. Their agentic trust platform is built to scale with you, not slow you down. With over 1,400 automated tests across more than 400 integrations, Vanta collects evidence and monitors your controls year round. Plus, you can access the Vanta agent everywhere you work.
Whether your team is in Vanta every day or living in tools like Claude or Cursor, they work around the clock from full program context so nothing slips through the cracks. The numbers back it up. Vanta delivers 526% ROI over three years and pays for itself in three months. Over 16,000 fast-growing companies like Ramp, Ryder, and Harvey save time and resources while scaling fast with Vanta. Learn more at vanta.com slash decoder, spelled V-A-N-T-A dot com slash decoder.
And you don't ever need to touch a line of code to do it. They also have a mobile app that lets you run your store from anywhere. Their AI e-commerce expert, Sidekick... can help you manage almost any business task with a single prompt, including holiday discounts. With Shopify, you can easily sell everywhere people shop. Online, in person, across social media, and even in AI chats. If you're ready to launch your business before Black Friday, go to shopify.com slash decoder to start your free trial. That's shopify.com slash decoder. Go to shopify.com slash decoder today to hear your first...
Support for the show comes from Granola. Have you ever left a meeting thinking, that was productive, and then 10 minutes later, when it's time to take next steps, you're at a loss. Or worse, you've been heads down the entire meeting furiously typing notes instead of actually listening. And now your brain is fried. Meetings are a mess. Granola fixes that. Granola is an AI-powered notepad that makes your meeting notes useful. It captures what happens in your meetings and turns it into clean, structured notes with decisions and action items pulled out and made easy to find. And the best part? Granola works through your device's audio, which means it integrates seamlessly into the video conferencing tools you already use. No setup, no awkward bots. It's just your normal meeting.
with superpowers. You get to actually listen instead of frantically typing every word, and still walk away knowing exactly what was decided, who's doing what, and what comes next. Once you try it on your first meeting, it's hard to go without. You can try it totally free. Just head to granola.ai slash decoder. That's granola.ai slash decoder to get your time back. Go to granola.ai slash decoder and try it on your next call.
Support for this show comes from Wix. You can make a great-looking website with Wix, and you can do it your way, whether you want AI to jump in or prefer to do things yourself. Get a custom ready-to-use website in minutes with Wix's AI Website Builder, or choose from more than 2,000 templates. Get built-in solutions tailored to your business and enjoy easy, fuss-free domain registration, web hosting included. Wix powers more than 280 million businesses around the world, because with Wix, you can own your individuality, create freely, and scale fearlessly. Ready to create your website? Go to Wix.com. That's Wix.com.
Welcome back. I'm talking with Siemens CEO Roland Bush about what exactly Siemens is. Then let me ask you the other decoder question I ask everybody. This is a big decision. How do you make decisions? What's your framework for making decisions? First thing about decisions is empowerment. Don't pull every decision up to the boardroom. It makes us slow.
It is really not attracting people. People want to really take decisions on a lower level. So basic ideas, take decisions on the lowest possible level. However, empowerment is not anarchy. You have a clear strategy with that set boundaries. This is where you want to go within that frame, within your responsibility. You can act and empower. Empower is a two-way street as well. Empower gives freedom, but it also requires accountability. So if you empower somebody, it has to be accountable for what the people are doing, which is super important. So this is the first thing. So don't decide on things which you can't decide on a lower level.
But then if it comes to, let's say, the bigger rocks, the M&A decisions, this really goes into a very, I mean, we have processes, how we do it. We had a P-proposal, which is a proposal where somebody says, this is a company I want to acquire, this is my business case, outside in, in order to say, okay, no, we believe in that. You give me a trigger, you already know to negotiate, go forward, make a non-binding offer. And then they work on it. We have a lot of processes. Then it comes up, we call it I-proposal, where we finally pull the trigger to say, no, you can invest and you can go. So, and the decisions are, if it comes to strategies, we're engaging as many people as we can, the experts, listening to them in the boardroom. In some cases,
We also ask them not to prepare a super polished PowerPoint. It's not the point. We want to really get the content. And then we have a very open discussion culture in our managing board with our leaders to come to better decisions. Very often I'm also snorkeling around. We're getting advice from others, putting our network if it comes to certain decisions. I would say it's a structured process, but it's a process which...
encourage people to speak up and to bring their opinion in order to come to better decisions. Let me ask one more question to wrap up here. I want to talk about the state of the world, and I really do want to talk about how you see AI and digital twins fitting into automation, because there's a lot there. But if I'm a Siemens engineer working on low-voltage switchgear, right? I'm one of 35,000 people in India, and I'm like, the CEO is at CES talking about fabrics with Jensen Wang on stage, and we're in the middle of a tier transfer. But I just need to get my work done. And this is all just some...
corporate strategy distraction. How do you bring those two things together? Because this is the thing that kills projects at big companies. It kills them dead. The number of times I've heard that story is very high. So how do you bring that together? There's a first one, and this is so super important for communication because, and you got it right, low voltage. I mean, this is as mechanical as it can get. I mean, this is mechanical stuff. There's no software, no KIE, if not in the development, of course, but normally this is a product which is, which is, it's hardware. Not only because we have now a solid state switching, which is disruptive, which brings software into it, but take that aside.
And then talking about being on stage, talking about KI and new models and whatnot and dental twins, the people sometimes feel lost. Bush is always talking about that stuff, but I'm just doing mechanics. So we have to give love to these guys as well because they do a lot in terms of contributing top and bottom line cash flow to our company. They are part of the equation. If you go to any kind of customer, they say, I love your automation, I love your software, but still I have to do some switching here and we are rolling in. I mean, and it's super relevant because if you don't have a switch or if it breaks, then you have a problem. They bring the capabilities to a customer and say, and not, by the way, we have also some low voltage here, but really say, this is important.
it for you to have a very solid operation. It makes sense and they are proud to contribute. So you have to distribute your love not only to the new stuff but also to those which is basically super important and carrying also our P&L forward. And then the other element is, believe it or not, this CES presentation or keynote where you really are on par with, I mean, talk with Jensens there, Sadia there, and you show what we do and people, even the low voltage guys are proud of what we do. I'll give a last one because I have to share that with you. We didn't talk about a completely other area which is mobility, Siemens mobility.
They do trains. This is hard stuff, boogies, frames, mechanical. I mean, these trains are super loaded with automation. These are basically software-defined trains because they tell you whatever. Even before they come to the depot, they tell the depot what they need, which part, what's wrong, and how to replace it. So it's technology at its best. However...
which company can say that we are transforming the whole economy of 110 million people, country, which is Egypt, where we built 2,000 kilometers of railway lines from the north to the south, west to the east, connecting 90 million people and transforming the whole system as high-speed commuter and locomotives. I mean, this makes people proud. And I didn't mention any AI technology, even though it's in their own trains. But this is something where we could say, which company in the world can do that? I'm very curious about all of that. I think there's some amount of, you said software-defined levels of switches or software on the trains, right? Those worlds are colliding.
I want to ask broadly just about the landscape you're operating in to do all of that work, right? These are big opportunities at work. If you want to sell trains, you need to be a global company, right? You can't be a single country train supplier. You have to operate everywhere. I look at Siemens and its size and its history, and I say, okay, this company took advantage of globalization and free trade, right? You're in all these...
countries around the world, you've got tens of thousands of people all over the place. You're building products all over the place. You're taking advantage of the opportunities and the markets you're in, the talent that's in those markets. And then I look up and I read the newspaper and the walls are going up around the world everywhere, every single day. The Trump administration seems intent on putting ever higher walls between us and Europe in particular, which seems very confusing to me. Other countries are nationalists in other ways. How are you thinking about Siemens in that moment, where a company that was able to grow and be such a large provider to so many people because of free trade and globalization now has to contend with ever higher walls and barriers between countries?
Obviously, we believe rather in free trade than in trade barriers, because it brought the world to where we are. And leveraging all technology as fast as possible means bring it to different countries as fast as possible. The good news about it is since, and I mentioned our footprint before, since we are global from the very first beginning, by the way, when Werner von Siemens founded this company 175 years or more ago, even at the beginning, he sent one brother to London and one to Russia because he knew Germany is too small for his technology to scale. So ever since, Siemens is a global company. And now our local-for-local content in the United States or in China is 85%, 87%.
So that means we are so local and we have still goods traveling from different places. The impact on tariffs currently, and we said it last year, it's a public figure, and in 2025, last fiscal year, was a low-mid single-digit bottom-line impact. So, okay, that's good for us. It's maybe not good for others. Our customers are suffering, and with our customers, we are suffering, obviously. We know that for machine builders.
They have reduced volume because their machines are tariffed when they go to the United States. Along with that normally comes Siemens automation. So we see that. But the direct impact is rather low. It's a second impact. And we are increasing our resilience as we speak. As it comes to certain semiconductors, we are trying to double source as much as possible, what we didn't do before in order to scale. We are looking for more localization to invest in the United States. We doubled our capacity for low voltage medium voltage switching. We invested in assembly lines for trains in the United States. We are investing in India, in China anyhow, because this is one of our largest markets. So therefore, good news is we are quite resilient.
Bad news is that for many, many of our customers, it doesn't help and it somehow slows down. When you talk about investing in manufacturing in the United States, I've watched a lot of companies say a lot of things about investing in manufacturing in the United States. I'm from Racine, Wisconsin. I watched Foxconn insist that it was going to build an LCD factory in Racine, Wisconsin, and then simply not do that. I've watched Tim Cook reopen a factory that was already making Mac so Donald Trump in his first term could say that Apple was opening a factory. There is a lot of theater about manufacturing in the United States. And then there is the reality of investing for the long term when presidents come and go.
How are you balancing that? Is it theater? Is it real investments? What is the split? The truth is, it's a little bit of both. Where it's real, I mean, let's talk about the pharmaceutical industry. I mean, this is real investment. I mean, the Swiss ones, the German ones, they are investing in pharmaceutical production in the United States. If it comes to, I mean, some car makers who are not that strong, they do that. But the big wave of remanufacturing in the United States is not happening yet. We don't see that. And the reason is, maybe you mentioned it, number one, is the ability of people.
also trained people. It is that you don't know yet where the whole tariff situation settles at the end of the day. The other reason is that, I mean, why are we in a world which looks like as it looks like? Because, I mean, in particular, American companies, they were basically leveraging low labor cost and low cost in other countries and they made a good living out of it. And you mentioned some of them as well.
So do I believe that this will change? Yes, I believe there will be a wave. You see more manufacturing coming. And I didn't mention semiconductors. Definitely, this is hard fact. Semiconductor built. Maybe even battery factories would come, pharmaceuticals and the like. The only point, what I always advise our customers, if you build a new manufacturing in the United States.
make it as automated and as digital as possible for obvious reasons. Because you cannot get enough labor, let alone trained labor. And technology is there. So if you go there with a greenfield planned, you have all the freedoms to make this whole thing digital before you even send the first excavation machine. Your products, digitalizing, your manufacturing, simulate everything. That's what we do, by the way. Whenever we build a new one, we go all in. And then you build it. It's faster. You don't make mistakes in building. It increases your space productivity and reduces your energy consumption. It increases your output while having more.
variables and more variants of your products. So that will come. But we thought it comes faster, but it seems there's a delay in really ramping up manufacturing in the United States. And again, maybe sector by sector, you'll see different patterns. Do you think that's because people assume that there will be a snapback to normal trade relations in the world? Or do you think it's just slow? The later one. I don't believe this tariffs will be a snapback. Why? Tariffs are just more or less, they are.
They are like taxes. They help closing the budget deficit. And I never saw taxes going back, which help you closing a budget deficit. And I hope that this maybe comes to more normal terms. For example, our machines, the machines which are exported to the United States, they suffer from tariffs, but they have also these tariffs on aluminum and steel on top, which makes them quite expensive. So maybe that goes away. So there might be some adoptions, which I think will come.
But I don't believe that this goes back to where we are coming from. Look, I have a rudimentary understanding of economics, right? I studied this in undergraduate at the University of Chicago 20 years ago. My understanding of all this is this is how you equalize labor rates, right? You say, OK, you can make the products cheaper overseas. We'll just put a tariff on top of it. Now the product is as expensive. You made it in the United States. You might as well make it here. And that is bluntly what the Trump administration is hazardly trying to convey. But what you're saying is we're Siemens. We make automation. We can...
virtually model the entire factory as a digital twin before you build it. In AI, we can automate even more. We can automate people using Excel to program your factory. Just build that. And I look at that and I say, well, that didn't get anybody a job, right? I look at data center investment in the United States and communities around the United States are pushing back on data centers because they're like, this is a lot to extract from the environment and from our land and not enough jobs. I see that same argument being applied to fully automated factories. How do you push back against that? Is a fully automated factory a net add to the economy, do you think? It's a net add with less people per output if you don't automate. Absolutely. That's absolutely clear.
The point is we are living in aging societies. I mean, Germany is aging, Japan, Korea, China is aging. There's a really step curve. So sooner or later, you will see that creating jobs like crazy is maybe not the point because you're missing jobs anyhow, labor anyhow. You might want to deploy the labor you have in jobs which you cannot replace. I mean, the social system, the healthcare system and the like, and use labor really where labor makes a difference. And in manufacturing, you have less and less.
This is changing. You still have people on a shop floor, but you will have less, and you said it right. I mean, an AI factory, fully automated, I mean, this is great. It uses a lot of space, uses a lot of energy, and it creates a limited amount of jobs. I mean, that's what an AI factory is. Who buys the outputs of an AI factory? The tokens, the intelligence. I'm just saying, if I build a fully automated factory to make cars, but no one has a job, who buys the cars? I mean, again, as I said before, what you do is, when you have a fully automated factory, you're driving economy, you grow faster, you can bring...
the production of the United States, which has a value in it. And then people, and once economy is growing, obviously your GDP by capita is increasing and people will be going by cars. But they have a different... But where do they get the money? Again, rudimentary understanding. So number one is, you're replacing blue-collar worker with more trained workers, engineers, maybe also blue-collar worker using AI technology. So a factory doesn't run without. You have a higher output with less people. But you still have, I mean, I talked about it, the whole service sector is super relevant. And they're missing people as well. I mean, Germany, for example, I can share with that. If you would not have...
I mean, 100,000 of immigrants, and most of them are working then in service jobs, including hospitals and the like. If I would take them out of the equation, our whole healthcare system would collapse. So there's a lot of jobs which you can replace. And there's one more thing. When people ask me what to study, and I said, okay, a solid education on mathematics or physics is always good. But if you don't really feel like, if you go for a plumbing job or a handcraft job, this is the last one to be replaced. Yeah, the world needs electricians and plumbers for sure. That much I understand. Someone's got to build and plumb the data centers, and maybe that's the only career in the future. We're going to take a quick break here. We'll be right back.
Support for the show comes from Framer. The future of website building is seamlessly blending all the tools we have available. And you could use a platform that has it all and stays useful even after you launch. That's what Framer is. Framer is the AI website builder that helps creators, teams, and businesses ship production-ready sites faster than ever while getting every detail right. You're able to prompt, inspect, edit, and publish in one place at a whole new pace. On Framer, agents bring speed and scale while people bring taste, judgment, and control. And that all takes place in the same workspace where the real site is designed, managed, reviewed, and published.
Everyone's ideas land on the canvas, stay editable, and can be published when the team is ready. Thousands of businesses from early-stage startups to Fortune 500s are choosing to build their websites in Framer, where changes take minutes instead of days. Learn how you can get more out of your site from a Framer specialist or get started building for free today at framer.com slash decoder for 30% off our Framer Pro annual plan. That's framer.com slash decoder. Rules and restrictions may apply.
Support for the show comes from ServiceNow. AI was supposed to handle the parts of the job you hate. Instead, it just describes them, suggests what to do about them, and then leaves you to do it. That's not help. That's homework. ServiceNow's AI specialists are different. They're not a tool. Think of them as digital teammates who actually do the work from start to finish. Cases get resolved, requests get processed, loops get closed, and most importantly, no extra work for you. Because when you can truly delegate to AI, you can get back to the work only you can do. The work that requires a person with ideas and judgment. And you know, a pulse.
To learn how to put AI to work for people, visit servicenow.com. Support for the show comes from Superhuman Go. AI tools are designed to help you sort through the chaos, but when those tools live in another tab, they never make it into your actual workflow. Superhuman Go from the makers of Grammarly is different than the AI you've tried. Go already has all the context it needs, handling the repetitive stuff so you can focus on the work only you can do. It's right there when you need it. No switching tabs, no losing your place. Try Superhuman Go from the makers of Grammarly. Find out more at superhuman.com.
Support for the show comes from PipeDrive. Sales teams spend up to 50% of their time on admin work rather than selling, building relationships, and closing deals. PipeDrive gives them that time back. It's the intelligent AI-powered sales CRM loved by growing sales teams. PipeDrive just launched meeting intelligence built right into the CRM. Before your meeting, it pulls deal history and past conversations into one brief. During the call, it records and takes notes. After, it auto-drafts the CRM updates ready for you to review and approve. Head to pipedrive.com slash vox to get started and get an exclusive 30 days free instead of the usual 14-day trial.
No credit card or payment needed. Welcome back. I'm talking with Siemens CEO Roland Bush about what exactly Siemens is. So your vision is like a fully automated factory. You're talking about the sort of higher order jobs, like information jobs.
technology jobs, engineering jobs. I would call those software jobs, right? Like in some broad categorization, there's some amount of white collar work that has a laptop involved. That's the next thing you could automate with AI. And you've made some gestures at that, right? We can automate even more of the things I mentioned earlier. It seems like a lot of the frame for how you're thinking is Siemens has traditionally automated atoms. Now you can automate bits, right? And I see so much excitement about AI automating bits. How are you thinking about that? Like moving up from, okay, you've decided how many units to produce. We'll produce them to, we're going to actually.
automate the deciding of how many units to produce? It starts with, and I come to manufacturing in a second, but it starts with the design of your product. I mean, you create a digital twin of your product, but you simulate how it runs on a manufacturing line. And this loop of producing and simulators and producing your products, which you have already in the digital world, this is so powerful because whenever you make a change, a component, maybe you want to be more resilient, design another component, you go all the way back to your design drawings to change it and you know what your manufacturing is impacted. So that's very powerful. But let me go for the manufacturing line itself. The whole idea is that you start building this operating system, which is a layered system. Obviously, you need to get all the...
data which a manufacturing line produces. You have to connect all your machines, the status of your machines, enrich them with environmental data as you want to get the real-time data, even the drawing data of your machines as well. Because once you have that, then you simulate it. And this is what we call the digital twin composer. That means you're sucking in different digital twins of machine, of a line, of a product. You suck it in, and then you have a complete comprehensive digital twin, which ingests real-time data.
And then you can really, I mean, you can go forward, backward in time. You can find out what's the problem. And then here comes the real one when you close the loop of interesting data, but sending data back to the line, which is then the agent, which behaves on behalf of you. And an agent is like a trained supervisor for a line. So when there's a red light blinking, a supervisor goes there, he has a look at it. He says, oh yeah, this is a problem. I know that over and over again, this is what I have to do. And this is what AI agents can do. Finally, you still have to have somebody who is maybe...
removing a blockage maybe updating a software will be automatically but changing a part changing a piece have a switch which is going wrong so you have to have people who know what to do we tell them what to do with glasses i presented also in the in the presentation in the cs so this helps you interacting in your natural words and it helps you and fixing things even if you don't know really all the details this is where agents coming in orchestration agents which is supported by a machine building agent a machine agent a product agent a workflow agent whatever and and that's how the future looks like um which is very powerful it keeps your yield high Your quality process is very high. And you're super agile if there's any change or change in your production because you have a different version you want to produce. It's super smart and it doesn't take a week to reassemble your line, but it really goes automatically. Your agents, when you talk about industrial agents, so there's some line, something's gone wrong, a warning light goes off, and you say an agent will help you figure out what's wrong and potentially fix it by itself. Is that based on an LLM? Are you using one of the models from one of the big companies and it's just an LLM that you've trained to think about? Yeah.
align in that way? But not only. It's based on an LLM, different ones, depending on which one we are working on, different use cases, but it's based on an LLM, but it's not good enough. If it's just an LLM or an LLM-based agent to fix a problem, I mean, the hit rate is near where, I mean, everything that we need. But we are training these LLMs on our data, on proprietary data, product data, machine data, operation data. Once you have that, and even the data of fixes in the past.
Remember, when you walk a plant and you see the whiteboard where a supervisor writes, I have a problem here, this is who we're working on, here I fixed it. All that knowledge goes into this model. So that you just say, okay, this is the pattern of a problem, this is how the fix was. The model knows it because we upload the data. So it's a model which is trained. This is why we talk about an industrial AI model, which is trained on the industrial data. Fundamental, I mean, if it really runs across, but also very specific, if it comes to certain machines, then the hit rate goes up from 60-70% to 90%.
five-ish, eight-ish plus, which is really then what you can do. Are these your models that you're training, or are you augmenting models you're taking from OpenAI and Anthropics? We don't do LLMs, so really large models which are trained on the whole knowledge in the world. This we don't do. I mean, this is what we're using. We're using any kind of. And for specific tasks. So we have some models which are very good in software, co-pilots and agents for software. Now we're working on really genuine newly product designs, not only just having the next code line, but really genuine designs, a completely different world. And in some cases, we're working on co-pilots on the shop floor. We talk about Microsoft, for example.
We have first use cases doing that. So the challenge is, of course, that an industrial AI application doesn't accept hallucination. I mean, you really have to be sure once it's an agent out, it does what you want them to do. This is my fundamental question, and I've asked a lot of people this. I'm very curious of your perspective because the domain is so much different. I am not convinced that LLM technology as it exists today can...
make the leap to do all of the things that people want it to do, right? You see the gaps, even as you're saying, an LLM on its own hallucinates enough to only be effective 60-70% of the time, which is nowhere near good enough for all of the things people want it to do, especially at the labor replacement rates that some of these folks talk about. Do you think it's good enough or do you think it's the actual augmentation that makes it the products that you build with it good? We need the augmentation, absolutely. And do you see LLM technology, the core technology, improving at a rate that might change your assessment of it? Good point. To the question, LLMs will get better and better, but I don't believe that this...
LLMs, if you do not train them really on specific industrial data, and this is where the augmentation comes from, you can train them as long as you want. They will not get to the level which we can use on the shop floor. It will not work. So they need, and I strongly believe that the LLMs need specific, domain-specific, machine-specific data in order to really make a difference. But then, if you do that, then you really can make a step up which is fundamentally higher. I'll give you two examples which are maybe interesting. One is...
when we talk about it was an optical inspection kind of task where we used an LLM and say, okay, show me the problem. Hit rate was okay-ish, but not to a level we needed. Then we start training the model with not so many data, which is anyhow important because if you're manufacturing in a PPM level, guess what, how many mistakes you get a day. But on those which you have, and then you create obviously some synthetic data around those, once you train the model, your hit rate goes up substantially. It's much, much higher than if you use the next best and next best model.
Is there convergence to some extent? But I do believe there's a certain ceiling where you can train as much as you want. If you do not get specific training, then you have a problem. Now, my second example is, and this is now really, I'll give you a little bit of the spirit of how deep you have to go. We have a manufacturer, Italian. They do robots, a grip in the box, a grip in the box for any kind of parts. And obviously, you can train a robot to make this grip in the box. What we did is, we created this hole in the digital world. So we used a digital part.
with a digital arm and a digital software and a digital camera and maybe everything and we trained basically a robot over and over again on the on these on this grip in the box with our technology and then we switched it on and the grip the hit rate was still not satisfying 70 or two level It's amazing because we trained hundreds of hours virtually. Then we used NVIDIA technology with a photorealistic ray tracing of these pieces. Photorealistic ray tracing, different lights. Training the model over again, the hit rate was jumping up substantially. So these little details of having a normal representation of a digital part and a really photorealistic one made the hit rate coming up substantially.
And this is where, also there's a reason why when you train robots now in a virtual world, it doesn't really work. There's a reason we have so many people who are standing there with some handles and training robots to do a job and turn it over and over and over again. Because this is a real training on the real stuff. These little details. make a difference between an industrial application and one which you can not use. So what I'm saying is training models on specific data, on valuable design data, operation data, time series data, brings them to the level which we need in order to deploy them. All that data has to come from lots of different customers, right? You talk about Siemens as having all that data, but that data actually belongs to your customers. Are they willing to let you aggregate so that you can develop the products at the scale that you're talking about? Don't underestimate the amount of data we have got.
I mean, I talk about generations, generations of design data for controls, for trains, for switches, and whatnot, number one. Number two is, we have, I mean, I don't know how many thousands of machines we are operating. We have machine jobs, machining jobs. But if I then go to my AI guys, they say, I mean, okay, you make it now available, everything from Siemens. Then they say, nah, it's still not enough. So, I mean, they need so many data to train models. Now, we have an alliance for machine builders, German machine builders, nine of them. I mean, top, top, top. I mean, these are the names, Trump, Tim Gimori, and the like. They are ready now to share their data.
with us in order to train models, to bring them a model which they can use to make an application which makes their machines running autonomously. So you just say, this is the part, here's the machine, and just get going. They know that their data are not very useful because it's too little. But if you're adding up, if you create these data lines, it works. It requires a certain trust. This is, again, where Siemens comes in because they trust us. We are partners for decades now. We are very mindful about that one. So you're right. You need as many data as possible. You need as many proper data as possible. Would they share the data of their absolute latest machine with us? No.
don't need that. I mean, it can use all the other ones, which is absolutely helpful. Let me wrap up with just a big picture question here. I'm just thinking about Siemens as a company and what it represents and all the places it is around the world and the value of the scale, right? 320,000 people in all the places around the world. And then I just think about the barriers going up. And Siemens is a defense contractor for both the United States government and Europe and a bunch of other countries around the world. Are you planning for an event as catastrophic as the dissolution of NATO? It's great to be like, we can aggregate all the data from all our customers.
but also the world might fall apart. Are you thinking about Siemens as a global company in that context? Yes, we do. If you ask me, do we make a kind of a scenario planning of another war or whatever, some incident in Taiwan, we don't really because I told my people we can now do that over and over again. It comes anyhow different. So therefore, why would you stay agile and be really fast if something happens? So that's one thing. On the other side, we see obviously the trends in the world and we are working more and more on, call it forging of technologies.
so local for local, that you do not always rely on certain technologies from the United States, using China for China and the United States or for Europe. So it's a pity because normally you would like to scale, and we still have core technologies which we can multiply, but obviously getting, for example, we are training our industrial AI applications for China on Chinese LLMs, whereas for the United States, obviously we train them on American hyperscalers or whatever LLMs. So we do that. Have good experience also. And this makes us more resilient. Can I do that for each and everything? Can I fork all my software? I could, but it's just prohibitive, expensive. It doesn't make sense. But for certain areas, we do that and increase our resilience and hope for the best. I realize I'm ending on a down note here, but it just seems like so much of what you're excited about is the opportunity of scale, right? The opportunity to do these things in a cooperative way, in a way that maybe changes the world economy and all of us are just caring for robots in the end. But these are huge global ideas that you have. And I'm just trying to put them in the context of...
Boy, when I hear other people talk about AI, it is in the context of national champions and international competition in a way that feels very old, right? It does not feel like the world of the past 20 years, this feels like a return to a different time. And I'm just wondering how you can keep the optimism of the scale and the globalization when that is happening all around you. Maybe I have this optimism because I'm working in one of the most international companies in the world. I mean, we are a Chinese company as we are a United States company and a European one, and we have so many great people around the world.
And I'll see also how they are collaborating. And I believe that this is a core value of societies, which is super, super relevant also for the future. There might be times where this is not appreciated so much for whatever reason. But in the long run, I believe that a world which is using technologies in order to solve the real problems in the world, I mean, we have to feed now soon 10 billion people. We are cleaning out our climate and whatnot. And we have aging societies where healthcare is a huge problem. So we cannot solve it if we box ourselves too small and so scale it. And here I hope that this, when we have waves of opening and closing and whatnot, that ultimately this pays out if you are acting in a global international network as we do.
And therefore, I'm a more optimist. What should we be looking for next from Siemens? I think the next thing is that we are walking the talk. We talk about that we are building this industrial AI operating system, that we are using AI now for the next level to really not only validate, but also create. that we are leveraging our capabilities of bringing the real world and the digital world together, because the digital world can do so much if you do not have an impact on the real world. And we show that with our customers, like PepsiCo, what I was showing on the CES, like Kion, the supply chain or logistics company, or many, many others. And to see that we can be the entry door for AI technology into the real world together with our partners at scale. Roland, this has been great. I could do another full hour just on a structure scene. I think, as you can tell, you're going to have to come back. Thank you so much for being on Decoder. You're right. Thank you.
I'd like to thank Roland Bush for taking the time to join Decoder, and thank you for listening. I hope you enjoyed it. If you'd like to let us know what you thought about this episode, or really anything else at all, drop us a line. You can email us at decoder at theverge.com. We really do read all the emails. Or you can hit me up directly on Threads or Blue Sky. We're also on YouTube. You can watch full episodes at DecoderPod. You also have a TikTok and an Instagram. They're at DecoderPod as well, and they're a lot of fun. If you like Decoder, please share with your friends and subscribe over to your podcasts. If you really like the show, hit us with that five-star review. Decoder is a production of the Verge and part of the Boxing Podcast Network. The show is produced by Kate Cox and Nick Statt. It's edited by Ursa Wright. Our editorial director is Kevin McShane. The Decoder music is by Breakmaster C
First Citizens Innovation Banking, formerly Silicon Valley Bank, is the bank of the innovation economy. The name may have changed, but for over 40 years, their commitment to bold ideas, relentless ambition, and delivering what's next is as strong as ever. So if you're wondering where the road to innovation starts, it's here, First Citizens Innovation Banking. Head to firstcitizens.com slash innovation to learn more. I'm a voice agent powered by Deepgram. I don't just understand words, I understand context. Like when you're at trivia night and say, I'm on fire, it means you got the last three questions right. But when you're camping and say, I'm on fire, it means a spark just landed on your highly flammable synthetic rain pants. Deepgram understands words and context across STT, TTS, and voice agents. That means better conversations at production scale. Learn more at deepgram.com slash context is everything.
Support for today's episode comes from Delta, official airline of the National Women's Soccer League. Traveling is basically magic. It creates opportunities, it fosters understanding, it expands horizons, it connects people and communities to each other and to their own potential. Those are the values that drive Delta. With up to 5,500 daily Delta and Delta Connection flights to more than 300 destinations on six continents, Delta is committed to helping you explore your world. Whatever gives you the travel bug, Delta is here to help you get there. Visit Delta.com for more info. Delta, wherever the journey takes you.