- 18 hours ago
Roland Busch will showcase how Siemens is developing AI, digital twin and automation technology to drive real-world impact in the era of AI – where the real and the digital worlds are combined to transform the everyday. He will explain how AI-powered technologies, domain expertise and a complementary ecosystem of partners are enabling AI to scale in the real world.
Category
🤖
TechTranscript
00:03Please, can you switch the screens on?
00:18This has been technology.
00:22Well.
00:30I want you to think back to the time before electricity.
00:36The world moved at the pace of people, horses and bridged our distances.
00:43Steam powered our engines and ideas moved only as fast as a letter or a human voice.
00:50Then electricity arrived.
00:53A channel purpose technology that turned night into day.
00:59It amplified human capabilities and powered progress that once felt out of reach.
01:07At Siemens, we helped build that world.
01:11Electric grids, telecommunication, the trains that move people and goods, machines that accelerated production.
01:21One and a half century later, another channel purpose technology has arrived.
01:27This time, it is not about energy, it is about intelligence.
01:33It is about artificial intelligence.
01:36And artificial intelligence will be as transformative for this century as electricity was for the last one.
01:45It is revolutionizing how we design, build and operate infrastructure, making entire systems, grid, cities, economies more adaptive and more
01:59efficient.
02:01Siemens helped build the world in the light of electricity and now we do it again in the age of
02:09intelligence.
02:10We are powering the industrial AI revolution.
02:17And it is moving faster than any transformation before.
02:23For Steam, it took 60 years.
02:27Electricity, 30.
02:29Computers, 15.
02:30And AI, 7.
02:32And by the way, half of the time is already gone.
02:37When AI enters a physical system, it stops being a feature.
02:44It becomes a force.
02:45A force with direct, real-world impact.
02:51But AI that runs a production line, manages a power grid or steers logistics network has to be reliable in
03:02the real world.
03:03Reliable and safe.
03:06So, hallucination is not acceptable in industrial AI.
03:13What is the solution?
03:14It is not about going more slowly as we develop AI for the real world.
03:19It is more about having embedded AI technology from the beginning, right from the beginning when you first start using
03:30it.
03:31And that is what sets industrial AI apart from the AI on your phones.
03:38In the next 40 minutes, we will show you the tools that are ready today to help you make progress
03:47with industrial AI.
03:48And how you can use them to create value at every step.
03:54Design, build and operate.
03:57Well, it starts all with an idea.
04:01Ideas becomes designs and from designs you start building.
04:06But long before you build anything, you want to know how does it actually work.
04:13And how.
04:15Siemens software helps you find those answers really very early in the process.
04:23Our tools allow you to build digital twins.
04:28Digital twins, they are replicas of whatever you want to create.
04:33They are physics-based, photo-realistic, in 3D and real-time.
04:41So, no matter whether it is a machine, a factory, an entire logistics network, you run the twin.
04:49You stress test it, you stress test it, you break it, of course you break it virtually.
04:56And then you spend not a single Euro on really building it, not before you know how the real thing
05:05works.
05:06And only when it works in the digital world, then you start building it in the real world.
05:10Our customers, our customers take it one step further.
05:16With our digital twin composer, they connect digital twins of machines, of production lines, of entire buildings.
05:27And they make them play together.
05:31One integrated simulation.
05:35Take Keon.
05:36A leading logistics company.
05:39Their forklifts and automated vehicles keep warehouses moving and supply chains open.
05:47With our digital twin composer, the company can now test virtually how different warehouses lay out effects throughput.
05:56Before moving a single shelf.
06:02Looks fun, looks easy, yes, but all these simulations are hard indeed.
06:09Seriously hard.
06:12Where it becomes almost, almost impossible is aerodynamics.
06:17That means, for example, calculating the air flows around a moving object like a car.
06:22Exciting stuff, at least for me, but then I'm a physicist after all.
06:27So, what you see here is a wind tunnel simulation of a BMW.
06:32And normally these calculations take days.
06:36Now, with accelerated computing from NVIDIA, we drive them by a factor of ten, a hundred, or sometimes even thousand
06:44times faster.
06:47That matters in highly competitive industries like automotive.
06:52Or take passenger experiences.
06:55To avoid uncomfortable vibrations, car makers needed a complete 3D simulation model.
07:02Hundreds of thousands of tiny rectangles or triangles on a mesh.
07:08Build and simulate just one model took weeks.
07:13So, testing another one, you start all over again.
07:16Now, today we are announcing at Weev Attack, now where a software simulation, SimCenter, SimSolid, eliminates this testing and meshing
07:29entirely.
07:29Instead, straight from the design, you go in the calculation, in your physics, and it takes minutes.
07:40Every structural calculation an engineer needs, vibration, thermal fatigue, 30 times faster.
07:50Speed also matters if you plan to disrupt a whole industry as a startup.
07:59And that's why today, we are announcing at Weev Attack, our software collections for startups.
08:08We give up to 95% of discount for startups for design, engineering, and simulation collection.
08:20Faster product design and simulation, much more affordable.
08:27Our technology really makes the difference.
08:30Take the example of Latitude.
08:33A French startup, and they built rocket engines.
08:363D printed for small satellite launches.
08:40And they switched from a patchwork of tools to our software suite.
08:46The result, six months speed on project delivery, 15% higher engineering efficiency.
08:53That's how you turn ideas into design at the speed of AI.
09:01The next step is to engineer the production.
09:05Let's look at the reality as of today.
09:09First, manufacturing engineers create automation code.
09:14For this, they work on our TIA portal.
09:17It's an engineering tool and software engineering platform where they find all what they need.
09:23Automation components, controllers, motors, switches, sensors.
09:27Thousands of them.
09:30In this environment, engineers configure all components.
09:34They document what they do, and they test and debug until their code works perfectly.
09:43And this takes time, a lot of time.
09:46It's a repetitive work on top, but this is about to change.
09:53Less than two months ago, we launched a new revolutionary AI product for industrial automation.
10:01We call it the Eigen Engineering Agent.
10:05And it's not an AI chatbot that only comes up with suggestions.
10:11It is an industrial AI agent that completes engineering tasks end to end.
10:19So it operates inside our TIA portal, and it works autonomously.
10:24You define the outcome, and the engineering agent makes a plan, collects the documents and all the data.
10:33It writes executable code for your controllers, compiles the whole thing, and validates it over and over again until the
10:44code works and it's compiled without an error.
10:47Ready to implement.
10:49What is the feedback from our customers?
10:51What is the feedback from our customers?
10:5150% higher productivity, 2.5 times faster development, 80% higher quality.
11:01And today at VivaTech, we are announcing two new capabilities of our Eigen Engineering Agent.
11:09It understands now, additionally, parts of automation projects, very important ones.
11:15Now, it is grounded in plans from electrical design tools and automatically integrates that information into our TIA portal.
11:27Ready to use for all the engineers.
11:30What used to be days of manual work, now takes minutes.
11:36Secondly, the Eigen Engineering Agent has learned to build exactly along the common norms of your company.
11:45The Agent now works with your engineering framework, that's how they call it.
11:51Hundreds of pages and documents, thousands of rules, and comes up with a correct software structure which is compliant to
11:59your frameworks.
12:00And that fits perfectly.
12:01And again, in minutes.
12:04These capabilities move our agent upstream in the workflow.
12:11That's what industrial AI looks like.
12:14It's not hype.
12:15It's not hype.
12:16Hallucination.
12:17It's real impact.
12:21So, we have designed, we have built, and now the crucial moment starts, we bridge into the real world.
12:31Production in a real factory where thousands of decisions need to happen in parallel with real world consequences.
12:39Industrial AI finds the hidden potential in your machines and processes.
12:47It reads faster than any operator could and learns with every single cycle.
12:54So, let's take a look at this welding line at an Audi plant.
13:02It had been running in the real world for years.
13:06What you see here is robot arms welding car bodies with millimeter precision.
13:12Then comes the quality check.
13:14Multiple cameras capture the process.
13:18They do that, 2,000 welding spots per minute.
13:23And this piece of hardware, it's our industrial PC, is checking 2,000 welding spots a minute.
13:33It has a trained AI module in the cloud which runs on the edge.
13:40It works on our NVIDIA GPA.
13:42It's the crane which we are bringing from the cloud to the shop floor.
13:49Inspecting, evaluating and reporting.
13:53And now imagine this intelligence across an entire factory.
13:59In Erlangen in Germany, we produce the components that make electric motors run efficiently.
14:05And we are turning this factory into the world's first fully AI driven production site.
14:12So far, 100 AI algorithms work on the shop floor.
14:18Digital twins, intelligent automation.
14:21What's the impact?
14:2240% faster time to market, 42% lower energy consumption, 69% productivity increase.
14:31Imagine these results in your production sites.
14:36And we don't stop here.
14:38Erlangen is our blueprint for the factory of the future.
14:41Here, we use physics AI to automate tasks that have simply been out of reach by reasonable cost.
14:51For example, placing a plastic bag of screws into a package.
14:58But these bags, they are flexible, difficult to grip, unpredictable in shape.
15:04So no conventional automation could handle this at a reasonable cost.
15:08And with the highest quality requirements on the shop floor.
15:11Physics AI can.
15:13We taught the system what to do.
15:15If something isn't placed correctly, it detects and corrects itself.
15:21Every factory has hundreds of tasks like this.
15:24Tasks that were never worth automating until now.
15:30Two lightweight robotics arms, minimal programming and maximum flexibility.
15:36And the colleagues have more time for the work.
15:39They really need their expertise.
15:44So the tools are here.
15:46So you might wonder, why do many companies get stuck in pilot after pilot
15:53rather than scaling?
15:55And let's learn from companies that are doing that successfully.
16:01And for this, I would like to invite my fellow board member, Cedric Nike.
16:17Thank you so much, Roland.
16:19Bonjour, Paris.
16:21I will do a quick experiment to wake you up.
16:24I don't know if it's going to work, but it would be great if you could help me.
16:28So it's audience interaction.
16:30I'm going to ask you three questions.
16:33First question.
16:35Please raise your hand if you use AI on a daily basis.
16:38Just raise your hand. It's very simple.
16:41Very good.
16:43Second question.
16:44Are you not only using AI, but are you driving within your organization an AI project?
16:50Please raise your hand.
16:53All right.
16:53And now comes the more difficult one.
16:55Please raise your hand if you are driving a project which is actually super successful and generating return of investment.
17:05Please raise your hand.
17:06These few people, can you stand up?
17:08Stand up, please.
17:09The people, stand up.
17:09Yes.
17:10You have to commit.
17:12All right.
17:12There's fewer people.
17:13Look at them.
17:14They're the heroes.
17:15Give them a huge round of applause because there's very little of them.
17:19Get their contact.
17:22You can sit down.
17:23It's all over for you guys.
17:24So sit down.
17:26But at the end of the day, what I was trying to demonstrate, we all talk about AI.
17:30We all use AI.
17:31We all try to push it forward.
17:32But we fail.
17:33And we fail particularly in a lot of industrial environments.
17:37And the question is, why is that the case?
17:41So, there is something which I call the three traps of adoption of AI in an industrial environment.
17:48I call them the industrial AI syndromes.
17:51So, three ones.
17:53Let me start with the first one.
17:56The first one, I call it the spray and pray syndrome.
18:00So you try to shoot AI at everything.
18:03You define somebody as called, which is chief AI officer, and he does a lot of pilots.
18:09The joke, and the Germans are not very good at jokes, but the jokes in our German company,
18:14we have more pilots than Lufthansa, right?
18:17On a plus de pilot que Air France would be the French sort of aspect of it.
18:21The problem is you get lost in pilot land.
18:24So you do, instead of having one plant you really define, you're doing as much as you can.
18:29First mistake.
18:31The second mistake is what I call the foundation syndrome, right?
18:35Once you have these islands, and it was the same when you went to Industry 4.0,
18:40you had digital islands.
18:42They were all not connected.
18:43You had little pilots, little trials.
18:45None of them were connected in some ways.
18:48I call it the foundation syndrome.
18:51So the hard thing you have to do is you need to build the foundation within your company to be
18:57able to do it.
18:57You need to do it in your CRM system, in your PLM system, in your ERP system.
19:02You cannot.
19:03We talk so much about this huge treasure, which sits in European companies.
19:09If this structure is not structured, you cannot access it.
19:13So the second thing you need to do is you need to build a foundation.
19:16And we're brutal about it at Siemens.
19:18We basically call it the fabrics.
19:20We are forcing everybody to have the same structure capability so we can build our models on it.
19:26So what's the third one that will appear to a lot of the French people here?
19:31I call it the lonesome cowboy syndrome.
19:34What is it?
19:36It's a bit like Lucky Luke.
19:38You're really sort of doing it on your own.
19:41You don't need any help.
19:42You just push the AI projects on your own.
19:45But the reality is you are often falling into the trap of not being able to scale it.
19:51So the three main elements which we have is how do you go about it.
19:56And I'm going to bring somebody to help me discuss it.
19:58And I'm going to bring you two very clear examples.
20:00But the three traps are very, very simple.
20:04The first one is you basically are trying to shoot at everything.
20:10Second one, you don't have a foundation.
20:13And the third one, you do it on your own.
20:14So in order to do this, I want you to help me to give a huge round of applause for
20:17Eman,
20:18the CEO of Capgemini Group, which will help me give you an overview on how we do this.
20:32Thank you, Eman.
20:36Hello.
20:38Well, good to have you here.
20:40And Eman, I shared my three principles, the spray and pray, the missing foundation,
20:45the why is it so hard to bring AI into the industry as a lonesome cowboy.
20:49So give me a quick overview, a quick intro from your side.
20:53Yes.
20:53I mean, so basically we run Capgemini and great partnership with Siemens for many years.
21:02You know Cedric well, he speaks French.
21:04I pretend to.
21:06No, he's very good.
21:07So really, when we look at deploying industrial AI, it's really a lot more complex and a lot more ambitious
21:16than using generative AI tools for productivity.
21:19I mean, we're talking about a different word.
21:22So we're talking about rewiring completely organizations
21:25and designing completely new business models with two key types, two key elements.
21:30One, we talk about new agentic products and services.
21:34And of course, here, we're leveraging accelerated R&D cycle, AI-enhanced software,
21:40and of course, physical products to be able to do that.
21:43And the second thing is the new agentic business processes.
21:47So we are redesigning how works happen.
21:51We are blending employees, humans, with AI agents to get the work done.
21:56But for that, we all know that we need strong foundations.
22:00The first one is on the technology side and the solid data fundamentals.
22:05If you don't have that, it's very difficult to progress.
22:09Then, you're deploying AI agents.
22:12AI agents are human digital workers.
22:14And for that, you need very strong governance, you need very good control mechanisms,
22:19and you need a very strong operating model.
22:22If not, it's like deploying hundreds of employees across the company,
22:30out of control, doing what they want.
22:33And you'd have chaos in your company if you don't have the proper governance,
22:37basically, to be able to manage these agents.
22:40You know, we talk about the control plane, about how we're going to manage all these AI agents.
22:44And beyond chaos, you'll also have a big consumption of tokens.
22:48So the governance and how we manage that is very important.
22:53And finally, you have the human side.
22:56Having AI agents and humans working side by side is not the simplest thing.
23:00Because if humans don't trust the AI agent, it's not going to work.
23:06So basically, you are spending money for nothing.
23:08So that's really what we have to do as basis to be able to do it.
23:13Absolutely. And you talk about trust.
23:14Humans and AI agents trusting each other.
23:16We trust each other.
23:17And we're going to bring about two companies.
23:18Why are we actually working together?
23:20Can you give you an overview of why we're such a good fit also?
23:22I think because it's a great complementarity.
23:24You know, from a Siemens side, you bring all the technology and the tools.
23:28Right? You meet domain and how in many industries.
23:30You know, from process to hybrid to discrete industries, both in automation and software, which is very important.
23:37On our side, we bring all the strategy and transformation capabilities, the end-to-end deployment really in complex environments.
23:44You know, and, you know, a lot of capability in terms of AI ecosystem management and in terms of, of
23:51course, new business model.
23:53And we deploy that across industries, you know, together from aerospace to consumer products to automotive to water.
23:59And, of course, we're building on the trust we have between each other and the trust we have with our
24:04clients.
24:04So one of those clients is Gravity. Can you give an overview of what you do with them?
24:08So we are in Paris, right? You have the Tour Eiffel, 7,300 tons of steel.
24:14So, of course, steel is a concrete component in our world today, bridges, cars, electric cars, many other things.
24:22Small issue with steel, 8% of CO2 emission is coming from steel.
24:27In the world, right? In the world, yes.
24:29So, Gravity came up with the idea of building a green steel plant in Marseille.
24:35Okay? 2.2 billion euro investment.
24:39Small thing.
24:41We replace coal by hydrogen, right?
24:44There's HI for the hydrogen at the end of the name.
24:48And the reality, what we're aiming for, is 2 million tons of steel produced every year,
24:53which means the equivalent of almost 1,2 billion every day, which is impressive.
24:59The only issue with that, there's no pilot plant.
25:03So, you get to go only at 1,2 billion investment.
25:07If you miss it, too bad, you cannot go back.
25:10Hence, we have worked together to build really a digital twin of the factory and of all the operating model
25:16to be able to simulate all the operation of the factory to make sure that we basically get it right,
25:23but also build as well an operation, the agentic AI layer.
25:28On top of that, from an operation perspective, we are able to have a learning environment,
25:33pretty much a learning factory.
25:35It's really the control brain of the factory that learns on how to optimize the hydrogen consumption
25:42on a real-time, do productive maintenance, etc.
25:45So, not only we have a great factory optimized and built,
25:49but also it's what we can call now a learning factory that can basically learn and improve every day.
25:55And that's the advantage, as we've talked about digital twin for the last 22 years.
25:58So, if you build a digital twin, but with AI you supercharge it to come up with the right solution
26:02very quickly,
26:03it makes a huge difference, right?
26:06But let's stay in France and let's look at another player like Sanofi.
26:09So, Sanofi, you know, gravity is about reinventing an industry,
26:14and of course every industry has its own problem and its own challenges.
26:17So, in pharma, the mission sounds simple, you know?
26:19You need to find the best and safest and most effective medicine for patients.
26:25But, of course, getting there is quite complex.
26:31And basically, the challenge, of course, in pharma is basically you have plenty of different factories,
26:43you know, plenty of different sites across the world, and it's a very regulated industry.
26:49And a lot of what you're trying to do is basically how you're going to scale.
26:52And to be able to scale, it's a very challenging point.
26:55Why? Because a lot of this information in a regulated environment,
26:59people manage a lot of the data in a manual way,
27:04and of course it is much more challenging to be able to replicate from one environment to another.
27:10The issue is going to be around scalability. How do you scale?
27:13Okay? And, of course, at this level, where you have 50, 100 production lines across 50 manufacturing skies,
27:21scaling is not simple. So, Sanofi has what's called smart operations.
27:26Okay? And that's what we have developed with them.
27:28So, it's a manufacturing execution system.
27:30And what you do here is really the cornerstone and the foundation of how you make it happen.
27:35It shows what needs to happen.
27:37It records what actually happens. It guides everybody step by step.
27:41It holds electronic batch records, not manual paper.
27:44Instead of searching through documents, of course, the team can see right there, in the right place,
27:50the information that it needs.
27:51And Sanofi, we're working with Sanofi, with Siemens, to be able to roll that across, of course,
27:57all the manufacturing sites across the world.
27:58So, why is that so important? Because everybody here knows Moore's law, right?
28:03Moore's law basically means you have an exponential sort of advantage in terms of cost structure
28:07through the development, which we will share with you later on the microchips.
28:11In pharma, we have something called EROM's law.
28:13So, if you spell Moore's law the other way around, you basically create something which is there,
28:18because every nine months, the cost of developing a new drug doubles.
28:24So, it becomes more and more expensive building drugs, so we need to be able to use AI,
28:28we need to be able to use digital twins, and we need to link because it takes 15 years to
28:32launch a new drug,
28:34plus three years to put it into production, and building this together actually makes a huge, huge difference.
28:40So, the impact is very concrete, right?
28:42We have 70% advantage in batch reductions, 80% less deviations.
28:48So, it basically means that AI brings something to the bottom line, which was my question I had for all
28:53of you.
28:54So, we talked about two industries, we talked about two companies, and what did they do right?
29:02They did not have a thousand AI projects, right?
29:06They built the foundation, the foundation on how to build a digital twin and simulate it,
29:10or the foundation on how to run their factories and their R&D.
29:14And they went with partners, they went with us, they could also go with other partners.
29:17The takeaway for us is that it will become easier if you focus, if you build the foundation, and if
29:24you partner as much as possible.
29:26So, Eamon, so much, thank you for joining us today, giving the overview and to absolutely see.
29:36And I talked about E-ROM's law, which basically meant that we need to bring AI into the pharma world.
29:41I'm going to go back to Moore's law and ask actually Roland to go back on stage with a very
29:45special partner to talk to you about more about this.
29:48Thank you so much.
29:59Well, two industries, two companies, one winning formula.
30:06They did not go for 100 pilot projects.
30:10They are building the foundation and they went with partners.
30:15That's the takeaway of today, that the rest becomes much, much easier.
30:23So, Eamon, thank you very much for joining us and we will continue the work which we did in the
30:31past.
30:32Every AI success story you just heard runs on one thing.
30:38It's compute.
30:39Compute means chips.
30:42And chips need a home, a physical home.
30:46They, it's energy hungry data centers.
30:50And if they're loaded as GPOs, you call them AI factories.
30:54Together with NVIDIA, we've developed the blueprint of the next generation of AI factories.
31:04They are pushing the limits of conventional design and operation tools.
31:09We optimize the energy use and cooling system for the lowest cost token per megawatt.
31:18How?
31:19Of course, with digital twins.
31:23Digital twins of every single subsystem.
31:28And then, remember, our digital twin composer, we're using it now to simulate the entire AI factory
31:35as a complex, complete, very complex system before a single cable is laid.
31:42And by the way, once you have built an AI factory, you will better keep an eye on it because
31:48it's very expensive.
31:50A single power fold can destroy millions worth of equipment in milliseconds.
31:57How do you do it?
31:59Well, back to physics.
32:00AI factories run on direct current, DC.
32:06Solar panels on the roof of a data center, they generate power, and guess what?
32:11It comes with DC.
32:13Batteries to store energy there come with DC.
32:17Every chip runs on DC.
32:20When the entire system speaks the same language, obviously,
32:23you do not want to waste power in conversions from AC, alternating current, to direct current.
32:29That's why the whole AI factories will run on DC, end to end.
32:36But when something goes wrong, direct current is much harder to switch.
32:43And here comes our circuit breaker.
32:46This is a new device.
32:47It interrupts folds in milliseconds, a thousand times faster than conventional mechanical systems.
32:57And think about what you've heard, what you're protecting here.
33:02Tens of thousands of high-performance GPUs worth millions.
33:07And by the way, these switches in their lifetime, they switch a million times.
33:12That's for sure often enough for the lifetime of a whole AI factory.
33:18All of this is only possible with the most critical component of all.
33:24It's the chips.
33:25Making those chips is a very complex engineering challenge.
33:31And there's one company in the entire world which is depending on this manufacturing process
33:37on one company, which is ASML.
33:41They built machines the size of houses.
33:45Those machines, in turn, create the smallest structures humans have ever built.
33:51Please welcome on stage, Christophe Fouquet, the CEO of ASML.
34:12Hi, Christophe. How are you, my friend? Good to have you. Good to have you. We're going to sit here.
34:15So, I just said, at least the biggest, maybe the most complex machines as I know, they are writing the
34:24smallest structures.
34:26We talk nanometers.
34:28You use light which doesn't exist really in natural life with amazing technology.
34:34Do you ever think about how, holy shit, how does this whole thing run?
34:38Well, we have to because, you know, we have hundreds of those machines across the world.
34:47And those machines have to basically print those AI chips, you know, every day, every night.
34:55And they run almost all the time, 95% of the time.
34:59When you think about your car, right, you run your car maybe one, two percent of the time.
35:06Right? If you had to run it 95% of the time, and a lot was depending on it, you
35:11would be a bit more stressed.
35:12So, I think it's a very important thing indeed, but part of the complexity of the machine is, in fact,
35:19to make sure that they can run without too much failures.
35:23And the other point about this running them really with the highest precision and uptime is that, I mean, we
35:30don't talk about the price of your machines.
35:31Some people are saying they're very expensive. I would say it's value for money.
35:35But Moore's Law, also in cost, work only if the machines are really working day and night, right?
35:41That's right. That's right. That's right. So, the economy has to work. So, yes, those machines are expensive.
35:48One more reason to keep them running, because when a machine goes down, this is, of course, a major economical
35:53issue for our customers.
35:55How many components does a machine have, roughly?
35:58More than 20,000, right? So, this is another way to look at the complexity.
36:03You talked about the size, the size of a house. People always look at the size.
36:07I think the number of parts components is very important, because a lot of those parts were designed, basically, for
36:18those machines themselves, right?
36:20Because the technology is unique. A lot of the technology composing the machine is unique.
36:26So, all those parts have to be created, basically, almost specifically for this tool.
36:31So, here comes the beauty, because your engineers are working with Siemens software.
36:37Yeah. You design your machines with our software.
36:40We have TeamCenter with our data backbone, which holds all the information of these 20,000 pieces in the design
36:47phase.
36:47And that's where we're working together.
36:49Yep.
36:50And then you came back and said, Roland, what I need is, I need to know at every point in
36:56time, design phase, manufacturing phase, commissioning phase,
37:02which piece, which part, which component sits in my machine, right?
37:06Correct.
37:08And 20,000 parts is a lot.
37:11And you're right.
37:12We need to have the design information.
37:13We need to manage the entire bill of material to design, to manufacture, to service.
37:21And we need to trust the information, because, you know, if not, we're lost, basically.
37:28And the other important things, we want to innovate in ISML, so we want our engineers to really spend their
37:35time, basically, on innovation.
37:38Therefore, as less time as possible managing those kind of things.
37:43Therefore, the quality of the products, I think Siemens provides, is key.
37:48And that's why we have this great dialogue for many, many years.
37:51And here, I'm sitting next to a very demanding customer, too.
37:54So they always want to have more capabilities in our software to run it faster, more efficient, and we love
38:00that.
38:01What's next? I mean, will these machines getting more complex in the next couple of years?
38:07I think that, you know, they did become more complex over time.
38:12I think we still have ideas for future machines, of course, because the requirements are changing.
38:20Innovation requires, you know, when you come to this kind of limit, you talked about nanometer, right?
38:26Everything you do is complex.
38:29And you touch very, very difficult part of physics.
38:32We talk about plasma.
38:34We talk about things that are very, very difficult to understand.
38:38And this is where AI also becomes important for also the design of our machines.
38:46Because it gives us a chance, using a lot of data we have collected, to basically try to maybe do
38:53things a bit differently.
38:55Things that maybe a human alone cannot do.
38:58So we see AI in this sense as also a next frontier for innovation.
39:05So most probably AI will help us to take one of the next steps when it comes to technology.
39:10And that's what you're working on.
39:11I talked before that we are using now our simulation tools.
39:16We rewrite the code to run on GPUs, which makes them 100,000 times faster.
39:20But this is only one thing.
39:21It still has the same purpose of having simulation, which validates your design.
39:26Now we want to have our software trained on your former designs.
39:32So to create new ones, new ideas.
39:36Which eventually comes up with new ideas and your engineering don't.
39:40But still we should put it in a box.
39:42We call it therefore physics-based simulation because the physics tells you the limits.
39:47So that whenever you build it, it's not a hallucination you build.
39:51It will finally work.
39:52But that's something which needs a lot of training on design data.
39:58Because that's what models cannot do yet.
40:01They don't have domain know-how.
40:03They don't have the contextualization.
40:06And they have to be trained on the data.
40:08And that's the next step.
40:09Yeah.
40:09So there I think we fully agree.
40:11I think I said it before.
40:13Data is key in the value of AI.
40:16But maybe one question for you.
40:18Because I think we already mentioned the word AI a few times.
40:22And well, we know each other pretty well.
40:24And we know that developing this entire AI stack, you mentioned that a few times in your speech, is difficult.
40:31We are two important companies in Europe, Siemens, ASML, with a few more friends, right?
40:37Which we brought together with the tech creators.
40:39Yes.
40:40So what do you think we should do to help Europe to really become competent?
40:44Partner in Crimes, for those who don't know that, Tech Creators is a bunch of companies, amongst them ASML and
40:49Siemens,
40:49who are really promoting at European level to, let's say, to regulate, but in a meaningful way.
40:58We are not against regulation, but we won't have meaningful regulations.
41:02I would say we have a success.
41:03Because on the AI act, the regulation is not regulating machines the same way as they regulate consumer data and
41:13the like.
41:14So that's all consumer AI.
41:17First, the second would be, now we have, next step for us is, I guess, the AI, the data act.
41:22Right.
41:23Because again, I mean, you have to, you should treat personal data differently than machine data.
41:29And so we are, I think, I mean, the short answer to your question would be, we hope that we
41:35can have a Europe which is not over-regulating,
41:39but it's also giving us an environment which gives us tailwind in a world where technology is changing everything so
41:48fast.
41:49So we should not go there to prevent slowing us down, but we should try politics to help us a
41:55little bit.
41:55And I think we agree, I mean, for our companies, I think we are fully aware that if we don't
42:02use AI, if we don't embrace AI, we'll fall behind.
42:05Exactly.
42:06I guess it's the same at the scale of a country or a continent.
42:10Exactly.
42:10And the other one is speed.
42:13I mean, we, I believe we are, the gap is opening, as we speak, we're over-regulating here.
42:19If you look at the most innovative markets, which is the United States and China, you see that there's a
42:23gap going up.
42:24And if you're not accelerating faster, you have a problem.
42:28This is where I said, recently, we have to get into this crisis mode.
42:32Crisis means you understand you have a problem.
42:35Yes.
42:36Number two is you focus what to do and what not.
42:39And number three is speed, speed, speed, because you have to, have to solve it.
42:43That's what we are up to, right?
42:44Together.
42:44So you use AI a lot and you use regulation a lot.
42:48So I guess most probably that's a good message for the company.
42:50Anything on top?
42:51No.
42:53And we have to join forces.
42:55Well, thank you very much.
42:57It's amazing.
42:57Thanks for the partnership.
42:59Thanks for being a demanding customer.
43:01We like that.
43:02We'll continue to be a demanding customer.
43:04Thanks.
43:06Mr. Fouquet, CEO of ASML.
43:19Think about where we started today.
43:23We flip a switch and the room lights up.
43:28None of us in this room thinks much about it, right?
43:32Electricity has become a fact of life.
43:35But it took a lot of effort to get there, to build the system that we take for granted today.
43:46Soon, industrial AI will be a fact of life too.
43:51Factories will adapt themselves.
43:53Machines will ask energy, will ask for maintenance.
43:58And before the break even, an energy will flow cleanly.
44:03Industrial AI will be as transformative for this century as electricity was for the last one.
44:10It's revolutionizing how we design, how we build, how we manufacture and operate infrastructure, and how we create value in
44:21every industry you can think of.
44:24If you find this interesting, well, it is still early enough.
44:29With good ideas, with hard work, and the right partners, you can build every successful business in industrial AI.
44:39Well, with that, I hope that I've been able to demonstrate to you that industrial AI is already there.
44:48There's a lot of tools, but still much, much more to do.
44:51Thank you very much for your invitation, and keep on going and using AI.
Comments