Move from Collecting Data to Executing on It – Transcript
Continuing the conversation on data, AI and the industrial ecosystem, this episode of the Future Ready Podcast from Siemens brought Rainer Brehm and Siemens and David Humphrey of ARC Advisory Group back to dive into how business are moving from having large data sets they do nothing with, even they are even accessible, to executing on the knowledge help in that data.
Nick Finberg
00:09
Welcome to the Future Ready Podcast from Siemens. I’m Nick Finberg, but I won’t be your moderator for this episode. My colleague Conor Peick will be taking over for this recording from Hannover Fair 2026, where he sat down with Rainer Brehm, the CTO for Siemens Digital Industries and COO for our automation business, as well as David Humphrey, the Director of Research at ARC Advisory Group. In part one, Rainer and David explain the growth curves and perceived adoption of artificial intelligence in the industrial ecosystem, most notably the distinction between what a small fraction of industry leaders are working through versus the wider market. But for this episode, they dive into how businesses can start to scale their deployments.
The reins are all yours, Conor.
Conor Peick
00:51
As we’re thinking about the scaling of industrial AI and SDA as well, you know, in your research are you seeing this importance of the execution layer. Is that reflected in investment decisions or how companies are setting their priorities?
David Humphrey
01:03
Absolutely. Yes. So prior, companies were invested in investing in analytics tools individually in data platforms.
They were hiring data scientists to help contextualize this data so that it can be meaningful to people like you and me that don’t deal with disparate data every day. This is moving into investments in execution platforms, so more at the operational level MES, maintenance, management like I talked about before, definitely edge computing is really big.
And some customers or users are investing in the control platforms. Yes, SDA software-defined virtual controllers are coming, but some are simply upgrading their hardware to more data-friendly processors that do a better job of pre-processing data before they share it with the rest of the system.
Rainer Brehm
01:52
I think this is a very important aspect of you don’t need to rip out everything. And the job of vendors needs to be how we safeguard the investment of our customers and that was for example a very conscious decision what we did on you know when we brought our virtual PLC. the virtual PLC a customer can reuse the entire libraries everything. So basically, it’s our core, S7-1500 PLC core, which we virtualized and put it now on an edge computer, or you can put it on a data center, like Audi is doing it. But the asset of, you know all the libraries, all the programs, the asset of how an operator, as you said, they may be afraid they cannot operate anymore.
They can operate because it feels the same, from their perspective, even if in the background it runs in our data center. So how we make sure that we take our customers at the hand and go step by step in that direction?
Not that so now I have everything is new and rip it out because then they are hesitant, yeah. But you know, do a step, next step and the steps need to build on each other to go into in the future I think is very important. And we see success coming there because their digestible steps, yeah, which customers are willing to go.
Conor Peick
03:03
And is an open ecosystem seems like it would be really important to what you said about it safeguarding your customers’ investment. Seems like openness would be an important aspect of adoption.
Rainer Brehm
03:13
Think now a great thing around SDA is that openness is given and, give you a concrete example though. We take our S7-1500 vPLC, it runs on Industrial Edge, and Industrial Edge is a fully open system. It’s Docker based, so everybody who can now program a Docker container or doing it via agent or Claude code could basically use it and deploy it. And it’s using the same data infrastructure as a virtual PLC. It will communicate with a virtual PLC.
So absolutely open. So, software defined is a beauty thing, you know, you can combine now PLC workload with another workload. On a hardware defined PLC to put different workload into a PLC it was not made for that.
It was made to control a machine or a line. So, it was under this aspect a more dedicated system. I think with SDA the flexibility comes in.
But this flexibility only really is realized if it’s open. Yeah. And that is because else if you do SDA not open, you have limited value. That’s the reason we really having that very open.
It’s not only openness to deploy new applications, as I said, as a Docker container. It’s also open because we said we could connect to all different field levels. So, it’s not you need to rip anything out.
You can say I put the edge computing on there. And you have all different kind of systems, PLCs from all different vendors, and you can connect, get the data, and then add the value on top of it.
David Humphrey
04:41
SDA is by definition open. The days of proprietary systems or platforms for competitive advantage, they’re over. And we we’re learning this from the IT world as we migrate to IT like architectures into an IT mindset.
SDA is a concept; it’s not a product. I don’t go out there and buy a piece of SDA. it’s an idea that that guides my implementation of my architecture and I can choose to use a platform.
There are lots of platforms out there. I can choose in your case, Rainer, the Siemens Industrial Edge. It might be a nice fit with the virtual PLC, but these things can pretty much run anywhere.
Openness is in and it’s here and it’s here to stay.
Rainer Brehm
05:19
I, a little bit disagree with David. Not, disagreeing. I think it matters which platform you are using because still openness is one important feature on the one side.
On the other side, we talk about a deterministic real-time system. So that means we need to build something in the system which is open, but you know, where you guarantee a certain determinism for an application.
It cannot be you have one application and oh and you put another application on there and the other application will not run anymore. So, take the PLC. A PLC is deterministic, so, but it’s open.
So, we need to take care and that we took care on Industrial Edge, for example, that you have a special core only reserved for the PLC because you want to make sure if somebody deploys another app, the PLC still works. Yeah. The the same topic then goes further under the under the aspect of long-term availability.
Because we also know maybe in the IT world you always get updates and which we also need to learn from the OT world that we need to get updates. The IT world gets updated, but how we make it long term available.
If we cannot say after three years, no something new is there, rip it out, do something completely new. So also, we need to make sure that the OT values are still in that platform, and we add the openness topic on top of that.
Conor Peick
06:39
So, thinking about some of these capabilities and we’re discussing the openness of flexibility as well. I’m wondering, Rainer, where are you seeing maybe the combination of industrial AI and software-defined execution? Are we already there with real-world applications or are we still on a journey to connect those dots together?
Rainer Brehm
06:57
Technology wise and product wise we are there on the automation.
Again, we come back to the initial statement probably more than the eighty twenty side where twenty percent of the customers are basically working with it, yeah. And maybe only five percent have scaled it. I would put it even that way.
But yes, product wise we are there. I mean again, coming back to the platform we just discussed, hey, I can put a virtual PLC on Industrial Edge and there is an app called AI Inference Server. And that’s an inference server which you know can run even NVIDIA on NVIDIA cards on the same platform.
And what you definitely can do, you can deploy an application into your PLC and then deploy a trained model on the inference server and they can communicate to each other so it’s there how many customers are really using it that’s again but coming back to the 80/20 rules.
Conor Peick
07:45
So, then David, as we think about the market for SDA and how it’s evolving right now, how do you see the space evolving globally and what kind of growth rates are we seeing? Are companies already restructuring their strategies around adopting SDA, or is this still more vision than reality?
David Humphrey
08:01
It’s a little bit of both. We’re somewhere between vision and reality. Like we talked about before, there’s early adoption, there’s different phases or stages of early adoption. Where we see it is first in discrete manufacturing.
The two obvious industries are electronics because electronics manufacturing lends itself well to early adoption because we change the way we make chips and electronics every few years. There’s always new machines being constructed. And then automotive the Audi case has been talked about.
I love talking about it. At BMW there’s stuff going on and I’m trying to I’m trying to find my way in there. I know some inroads and they’re more secretive than their Bavarian cousins.
But you asked about the growth rates. So, if we take as a baseline growth rate for automation hardware of not the past couple of years because they were very volatile, but typically three or four percent. And if we look at, in industry software, baseline growth rates of let’s say six to eight percent for mature software.
We’re seeing industrial edge platforms grow at rates of fifteen to twenty-five percent. I think I would say we’re projecting that growth. Right now, the growth is actually spiking because we’re starting from practically zero so you go you want to get 100 200 growth in the first couple of years but that’ll level out to like 15 or 25 percent for the virtual PLC again spiky at the very beginning But there’ll be a long phase where the market growth will be twenty to thirty percent during the phase where companies are doing the proof of concepts.
They’re buying individual licenses; they’re not buying en masse. But healthy growth rates, but small markets. And then for a lot of the rest of the stuff, the middleware, the platform technology, things like that, we’re still talking about growth rates in the teens.
Conor Peick
09:47
Do you see differences across regions as well of these adoption rates? Are there certain regions or industries that are leading the transformation?
David Humphrey
09:55
Absolutely. In, in Western Europe, here today at the Hannover Fair, we’re at the epicenter of SDA. Let me just state that all of the companies that have a say in the development are represented here. North America would be a big adopter of the technology because North America has a very seated tradition in software.
When I think of Siemens PLM, where that all came from, it’s all the Americans that are great at software and industrial software in particular. In in Asia to a lesser and a greater degree, to a lesser degree in in selected markets, like I mentioned, electronics, to a greater degree in places like China, because you have a lot of cases of the state driving it. You have a lot of greenfields.
I mean here we, I don’t see any greenfields out there. It’s all brownfields here. But in China you have just places being, factories being built up left and right and you have the opportunity to start with a clean slate, start new, and the state’s saying you got to use digitalization technology as much as it’s available.
Conor Peick
11:09
And are there any key macroeconomic or structural drivers that are that are behind the shift? You know, maybe like labor shortages or companies looking for greater resilience or just productivity pressure trying to, you know, do more with less.
David Humphrey
11:23
It’s a number of factors. You mentioned labor scarcity. I like to talk about skill shifts or are skills wandering from industry to industry that make themselves more attractive. That the high cost of capital is also one reason to invest in infrastructure that is more flexible through software defined architectures.
It lowers my initial capital costs, and it spreads my operating costs out over many years. And then just one general thing that I like to observe is system complexity. I work with machine builders in different, in in different ways and I see the architectures and how they’ve grown to become these spaghetti systems.
I mean we have multiple controllers; we have things that architectures that don’t make sense sometimes, but they make sense because, through necessity. It was necessary to use a different controller to achieve certain servo access performance that wasn’t available like we thought it was initially. And then these systems come together.
Anything software defined, putting everything to a big software pot and mixing it up, that’s a great way to reduce complexity. It puts every, all the applications next to each other. It treats them as workloads, which is one of my favorite new words of the year.
It’s a term we borrow from IT, and we treat everything as a living and breathing asset, whose resources we need to manage and applications that by necessity we might have to move to another platform which we can do seamlessly without interrupting it. But this is the sort of like dynamic world of software management that we’re inheriting from the IT world.
And that helps to reduce and control system complexity.
Conor Peick
12:57
So, then Rainer, does this reflect what you’re seeing in customer demand and real-world projects and sort of the conversations that you’re having? Where are you seeing the strongest pull today?
Rainer Brehm
13:06
I agree with what David said. Maybe to add on, maybe two aspects. I see beside the electronic topic and the automotive which is always kind of front running.
The topic of logistic and intra-logistic currently a field which is very dynamic. And we had here on the fair. KION is one of the key customers that really want to push the limits what’s possible and I think a lot of they can automate more.
I mean that’s really a big return of investor I think here is some is the area and we have another topic here in the booth which I also like, and a new factory concept, because when David talks about greenfield in China, well they build big greenfields, that’s right. What I see and I like here are the topic of pop-up factories. So, it it’s not a kind of a classical big factory.
It’s quite a small one which you might also tear down if it’s not needed anymore. So, there are more kind of smaller size factory. We have some great examples, a company called Desert Control, which puts containers somewhere remotely in the world to produce clay for better agriculture. We have here PepsiCo with Gatorade with a pop-up factory. But also, like two years ago, three years ago we talked about controlled environment farming.
Which is somehow agriculture as a factory, and you can all that make that kind of much more compressed and you can produce very close to the consumption. And we see that kind of small size factory, whether it’s pop-up or it’s whatever in in like vertical farming, they normally adapt into new technologies already. Because like take Desert Control, they produce thousands of containers they deploy in the world, and they want to manage those containers centrally.
So, the workload is there, they update it, they have a new recipe and so on and they want to manage it centrally, and I think their SDA is a perfect platform to do that.
Conor Peick
14:57
It’s interesting that you mentioned the pop-up factory concept. I think it’s such a cool idea that you know about putting a factory right where demand is. And yeah, it makes sense that you mentioned earlier, David, about greenfields versus brownfields.
And I mean a pop-up factory to me seems like greenfield almost on steroids, you know. It’s this little very highly automated modular factory. Yeah, so I suppose it makes sense that they would be so willing to adopt a new technology.
But so, as we think about the move towards software-defined automation as well, David, and we talked a little bit about how it may change the strategic approach of companies. And I’m wondering if you could comment on how, it might shift the balance between CapEx versus OpEx for these companies as they start to maybe plan out new facilities.
David Humphrey
15:40
So, it’s from CapEx to OpEx for some obvious reasons. CapEx means I buy my whole PLC system on day one and pay for it and then I amortize that over the fifteen years of lifetime of the asset.
I still have to buy IOs, I still have to buy operator panels I still have to buy all those drives and the electrical cabinets and everything. So, I have these capital costs up front. But for the control system alone, I’m putting it onto a server now.
There’s a good chance that server might have been purchased already. There’s a good chance that I have a data center that has capacity to add additional control units in containers so I might have no capital cost for that, or I might have the cost of one server. But I can take multiple virtual PLCs and put them all onto one server.
I personally, I haven’t done this yet, but I personally might be happy with three or four. When I think of you know what would happen if somebody kicks the plug out of the wall, but the rule of thumb that’s emerging is like around 10 or 12 PLCs, virtual PLCs on a server. You don’t want to do more than that probably but with hardware developments changing and capabilities growing over time. Who knows?
But my point is that I’m sharing hardware. So, it’s a common platform. I’m sharing it with other PLCs and with other critical applications as well.
So, my upfront costs are probably going to be a bit lower The OpEx costs are going to go up for a couple of reasons. I’m paying licenses. Right now, the rule of thumb is a virtual PLC per year cost about a thousand euros or so roundabout. So, I’m basically amortizing the cost of the control over fifteen years, fifteen thousand or whatever it is, seven years, seven thousand, as a rule of thumb.
And my engineering and maintenance costs are shifting as well. Previously in my operational years, I was doing a lot of intervention maintenance. I was going in with my screwdriver, my voltmeter to find out why and where there’s a short circuit or why an IO module failed.
I’m checking voltages. That’s shifting to a more software-centric maintenance. I’m sitting in a computer somewhere and I’m monitoring my containers.
And I’m getting a lot more information back on the status and the health of these devices because that’s what software defined automation, that’s what virtualization allows. So, I’m going to say that the maintenance costs of the pure automation hardware are going to go down. They’re going to still be there, but my OpEx costs will actually decrease in terms of maintenance.
Nick Finberg
18:14
It really does always seem to be a balance of how businesses prioritize needs and spend budget. But this is a really interesting example with a focus on how maintenance is priced out. There is still more to come from this great conversation in part three, so we hope to have you back to hear from Rainer and David again. But until then be sure to check out our other episodes of the podcast to learn about the different facets of digitalization or how other industries are moving along their journeys Thanks for joining us and see you next time.
