Advanced Technology Insights: Physical AI and humanoid robots

00:00:00: Welcome to What Tech Can

00:00:01: Do, the podcast series from Roland Berger's Advanced Technology Center.

00:00:06: In each episode our experts discuss the technology and industrial shifts

00:00:11: that decisive leaders

00:00:13: can't afford to get wrong.

00:00:15: This series

00:00:16: is for business leaders and investors who need more than analysis.

00:00:20: They need the operational reality behind this strategy.

00:00:24: Welcome to today's episode.

00:00:26: My name is Karina Leiner, I'm a principal in our Frankfurt office and part of Roland Berger's Advanced Technology Center.

00:00:53: Welcome all to this episode and I'm looking forward to your conversation.

00:00:57: In today's episode, we're focusing on human earth robots in.

00:01:00: what separates science fiction from reality?

00:01:03: We want to learn more about where humans robot stand Today.

00:01:07: why?

00:01:07: This is a multi-billion dollar business.

00:01:09: What decisive leaders need to act upon today?

00:01:12: And also what separates the pilot from paying off end?

00:01:17: Is it just an experiment or Are we really in reality?

00:01:21: So starting with Thomas and the Elephant In The Room.

00:01:24: If you look at social media, We see humanoid robots dancing!

00:01:28: We see them doing athletics... ...we see them picking up things.

00:01:32: but are they ready to take over our factories or is this just an expensive

00:01:37: experiment?!

00:01:42: Actually, it's a pretty big elephant that you're talking about.

00:01:47: You also write... It almost sounds like science fiction.

00:01:50: Humanoid robots are robots controlled by AI working around our factory floors and households or airports.

00:01:58: When we first started with humanoid robots as Roland Berger two years ago We published this study called the Outer Topic to be from Science Fiction through Reality.

00:02:10: And when we look at the development of the last three years, We see that human and robots are now pretty capable.

00:02:16: They can walk nearly as fast as humans.

00:02:18: they can operate in general in factory environments... ...they're competing in Olympic-like competitions.

00:02:27: At least from a durability perspective these performing tasks also sometimes work.

00:02:31: I had an easy pleasure to operate the human robot myself with a controller and gloves And you can really feel the technology and see that it's super, super capable.

00:02:43: But on the other side all of human robots are just prototypes for testing.

00:02:50: The hype is real, but the real value at this small to a certain degree.

00:02:54: But you can see there's tremendous opportunity in human world robots... ...but it will take also while until the prototypes we have today even from a hardware perspective are fully capable.... ...until they're really ready for production and they'll be fully industrialized.

00:03:07: Even on the hardware side I think that would take up to twenty-eight or twenty nine until that is at least partly ready!

00:03:14: You recently also published a study where you talked about the gap between hardware and software.

00:03:20: Can AI be a game changer to that or is it rather bottleneck?

00:03:29: First of all, without AI there's no humanoid robot.

00:03:33: so AI really core to humanoid robots.

00:03:35: Humanoid robots are part what we call in the broader picture physical AI.

00:03:41: The reason for that is relatively simple.

00:03:43: When you look at these nice videos, the first ones they were not made with AI but actually made of programming traditional robots in the same way as your program and industrial robots You can really create a nice video or use case But it cannot scale this.

00:04:00: So where companies move now into and see their first deployments It tests if training human or robot extensively on data only one specific task.

00:04:10: This works, but it's not scalable.

00:04:12: And what you really need is a human or robot where you can basically give instructions to human robots that can work flexibly in an effective environment on the household of someone else and therefore you need AI.

00:04:24: Therefore You Really Need Massive Datasets That You Train The Human Or Robot On.

00:04:29: What's the deal in human or robot companies are doing?

00:04:33: The biggest problem that they're facing is, They don't have enough specific data to train their AI models.

00:04:39: So on the digital front with all the let say digital AI models, the company could find data online or just an available form.

00:04:49: but the human and robot companies has a problems as it does not exist at high quality.

00:04:57: so there now moving actually create this data themselves, either in partnerships with suppliers or some of them.

00:05:04: One of the human robots started a pilot where they're giving recording devices to people and they are recording every day tasks like folding shirts in their house.

00:05:13: And that promises all these data.

00:05:16: with massive training on the iPhones and then data centers you can train.

00:05:20: it created general purpose humanoid robot so we can roll out flexibly.

00:05:25: So I see there still is a large gap, but let's maybe be optimistic.

00:05:30: Looking at economics and also the financial side of things in your study that you published... You said even an advanced humanoid robot costs around two US dollar per hour.

00:05:44: with an aging society far higher than two euros or two US dollars, sorry per hour.

00:05:56: That's less then what you would pay for coffee in New York and Munich or even other countries.

00:06:04: so can you give us some more detail about those numbers because it seems strikingly low?

00:06:10: And also this means for labor economics.

00:06:15: If you assume a cost of an advanced one, let's say twenty to thirty thousand dollars.

00:06:21: Even if we take thirty-thousand dollars and assume the lifetime for roughly five years... ...you end up with yearly costs of six thousand dollars and monthly costs of five hundred dollars!

00:06:31: And if you assumed that human or robot is basically working in theory work twenty four hours per day doesn't go on vacation it doesn't strike so That you want to do what could work twenty five hours or twenty-five days in a month.

00:06:45: So basically if you divide five hundred by twenty five, You end up with twenty dollars a day.

00:06:50: and then after the human or robots works for twenty hours today?

00:06:53: You enter actually with one dollar.

00:06:54: Yeah But then you can assume that of course you need spare parts you need a bit of electricity You need other things around like charging maybe spare batteries electric equipment And another elements and therefore two dollars is it's a number that we think Is realistic.

00:07:10: But what you also have to see is that these human or robots might not be as productive of humans, so they may a little slower.

00:07:17: Still if their effect are two-or three-slower there's still tremendous gap and the tremendous value.

00:07:24: The one point I want to point out... Jonas was talking about the operations side but apart from operations this human or robot can work in households for professional service In particular household where we have training data which now becomes available.

00:07:39: And if the human autobots are just doing, let's say one hour of... productive work or saving me just one hour, then the economics are there and we believe a lot of people would want.

00:07:49: You could even for example share that with multiple households?

00:07:52: Or if you have it in your household to do additional things like check the water level of the plants every day make you a nice breakfast or coffee.

00:08:00: so when they're doing more digital AI will also do more physical AI.

00:08:07: therefore we relieve huge numbers once technology.

00:08:11: Yeah, it has been closed.

00:08:13: Looking maybe into China to our colleague Shuo-Ai in Shanghai so as a China understood that problem earlier than the Western world or how does it come?

00:08:25: That China is already producing fifteen thousand units per year around the thirties factor of what North America for example does.

00:08:35: can you elaborate why China than the Western world?

00:08:45: I think we're not solving a problem faster.

00:08:50: In China, we are just adopting different roles of deployment because... We use different philosophies in terms of humanoid robotics.

00:09:03: China is pursuing a deployment first and then pursue for skill and data iterate generation by generation very fast.

00:09:16: While we see the US or Western countries.

00:09:19: their approach is premium first, or AI-first to firstly build the best foundation of big models and then deploy so that... The difference it's that we deployed first to collect data because we know that the data are one key bottleneck in terms of software.

00:09:37: I'll give you some examples, such as one leading China or also the world-leading humanoid OEMs in China.

00:09:48: They are already deployed many robotics over ten per warehouse for a leading logistic and warehouse suppliers in China can simulate more than fifty to one hundred accuracy of the real human activities and movements, such as to pick up a sword or move some standardized parts.

00:10:24: So deployment time is very short in our level.

00:10:33: And besides the logistics and warehouse in retail, in the logistics.

00:10:41: In health pharmaceutical related to China is already deployed quite a lot of human waste.

00:10:51: Another reason it's that mass production speed in China very fast.

00:10:57: For example we know Agibot from a one-thousand unit to five thousand units, just less than one year and from uh five thousand unit to ten thousand unit.

00:11:09: Just three months.

00:11:11: so where we see the figure?

00:11:13: AI for example they deployed some small skilled pilot in BMW factory and Tesla Optimus.

00:11:21: They internally use their own robot but it's totally different way.

00:11:27: So that's why we see, collect the data very fast and use it to solve many long-tailed use cases.

00:11:40: It seems like there is no surprise when it comes to industry but also about different types of decision making.

00:11:52: besides that different approach of addressing the technology barriers and bringing those systems into industrial reality.

00:12:03: We have seen also in other industries, China speed or a China approach and other ones.

00:12:11: So for example, if we look at semiconductor industry We also see the power play between China Taiwan and the US.

00:12:22: Would you compare?

00:12:24: The way how humanoid robots developed today to that same development?

00:12:30: And in the geopolitical dimension behind that do we have two separate worlds competing against another?

00:12:41: I think it is the reality that we are seeing a separating world, but it's not fully decoupled.

00:12:51: The reason is that on one hand, yes.

00:12:54: We're seeing the ICTS regulation in US this year In July we are seeing the FCC new regulations towards Chinese components and towards entire robot.

00:13:07: So these kind of regulations set a very high wall in terms of exporting the Chinese component than their entire robot into

00:13:15: U.S.,

00:13:16: maybe other part of world.

00:13:18: However We have to admit the fact that China contributes nearly over I mean, nearly sixty percent of entire value chain values especially for body parts such as motors and precise axles power modules And many of the, I mean for example the motors and ninety percent of capacity is located in China.

00:13:47: The axles and power modules over thirty or forty per cent of their capacity and raw material are also located in china.

00:13:55: so that's a real business world where regulation cannot fully address.

00:14:03: So now we're seeing quite alot smart ways to collaborate between China and other parties in other countries.

00:14:10: For example, we are setting some equity activities and capitalization related activities on my hand to meet the regulation of equity and a regulation management enterprise but also protect core technology not by breaking the laws.

00:14:41: So back to your question for the semiconductor, maybe we experienced in one of five years to achieve a current decoupled world off.

00:14:54: semi-conductor however may be humanoid robotics were already seeing the trend of the couple.

00:15:02: but what I mean is that Both sides, China side and the US side were seeing quite a lot of positive activities to try to seek flexibility.

00:15:14: And they tried to seek their ways for collaboration after both sides.

00:15:18: Thanks a lot.

00:15:19: that was super insightful.

00:15:22: Now there is a prejudice in industry that we consultants love our slides, but beyond slide and strategy let's go into the ground.

00:15:31: And you're an SSL expert from the operations team.

00:15:33: so if I'm running or manufacturing all logistics operations how do i bring human art robots to life?

00:15:41: What do need know?

00:15:43: what can they for me today?

00:15:51: Honestly, right now we are really in the experimentation and learning phase.

00:15:57: So all of our deployments they're about getting familiar with technology, capturing very important training data to further develop their algorithms.

00:16:21: If we look at the use cases that are getting deployed and in all these examples, BMW figure was just named.

00:16:29: The examples I usually been picking activities so crabbing apart out of a box things like loading components into machine or transporting boxes from A to B in warehouse.

00:16:45: honestly if you look at this case These are cases which can also be automated with traditional robotics.

00:16:54: Some of them then equipped a camera system, more advanced control algorithm.

00:17:00: but overall from an automation perspective this is nothing completely new.

00:17:07: I think where human needs really change the game in medium term if they're able to deal with environments and higher variants one of, let's say the classical use cases where they could add value on top of traditional automation are kidding activities.

00:17:28: So if we prepare an assembly process We need to sort The components that we need for the assembly Put them On a table or card wherever so that the sampler can then pick and conduct the process.

00:17:42: And this is exactly the task Where you have to be able with various different types of components, picking them out from different boxes and so on.

00:17:51: So if the technology gets to a point where you have that variance in your component this is the first time we really unlock new value in our factory.

00:18:02: However... From form factor perspective it's not necessary for us.

00:18:14: Very often it would be sufficient to maybe have just a torso or two arms on top of movable platform, something like that.

00:18:24: Where the human-made form factor really comes into play and unlocks all its advantages is when robots start getting capable of being deployed in multiple tasks and switch over between these tasks without a lot of reconfiguration.

00:18:45: So, if as a factory operator I can come in and deploy my humanoid robotics fleet the morning for loading off machines then move it into the logistics center to basically pack boxes this is where i need flexibility on a humanoid-shaped form factor that lets us say real value of humanoid sketch unlocked.

00:19:11: What is more on the visionary side, and here we are really talking long term.

00:19:16: Is the automation of complex assembly process?

00:19:19: So for example in the machinery industry I need to assemble a laser cutting machine For example.

00:19:27: Here We're Talking Assembly Sequences Of Ten Plus Minutes good configuration about differences from one machine to the other.

00:19:37: So this brings in a complexity and task variety for algorithms that need to be handled, which is right now just overkill what we have.

00:19:48: so overall it's not human.

00:19:52: weeds will come in suddenly takeover and automate all blue color jobs.

00:19:59: They start to fill the gaps, which we see in traditional automation and step by step get deployed as a technology mature.

00:20:08: Looking into our day-to-day business if a client CEO calls you on us I want to pilot a human art robot.

00:20:16: what do i need?

00:20:22: The question is always a little bit What Do You Want To Achieve With The Pilot ?

00:20:28: Do You Just Want To Have A Nice Showcase And we see a lot of these deployments as well, or do you really want to get the maximum out of that humanoid deployment?

00:20:38: Personally I prefer the latter one which basically means it's important not consider the humanoid pilot stand alone.

00:20:48: But more embedded in an overall physical AI strategy than also considers other technologies like autonomous vehicles AI enabled robotics, also machine vision process parameter optimization to extend all the factory automation that you already have.

00:21:08: And in such a setup I see right now three priorities.

00:21:12: first thing is really to map the manufacturing process end-to-end and identify use cases where from complexity perspective they can be addressed The current state of technology, but also with the State of Technology that we expect more in the medium term.

00:21:33: Then it's important these use cases are identified as large enough to actually move a needle from an operational performance perspective.

00:21:43: and when you try to do map manufacturing process understand this case is always the complete set of technologies that are available in mind, because we want to deploy the human needs and their pilots really into applications where they add the most value.

00:22:01: They should not replace other technologies which have already matured or are there but basically fill exactly these gaps as I previously described.

00:22:11: The second important part is to leverage a human pilot physical AI competencies, because if you look at manufacturing processes they are really different company by companies.

00:22:26: They're even different plant-by-plant and no technology provider really have the deep process understanding that is often required to develop and then deploy a solution.

00:22:43: We see that companies who have strong own competence, they are much better in industrialization.

00:22:49: They're much better at scaling up and this important part there able to adjust their process set-up what is needed in order to simplify the deployment of human weights or even make it feasible in those processes.

00:23:06: And then finally third part also use a pylon to start educating, teaching the workforce.

00:23:18: Because if we mention with a larger deployment of human weeds in the future and the factory... The task portfolio that is required from the work force will significantly change.

00:23:32: We move from execution to supervision, exception handling.

00:23:37: We have significant maintenance tasks And this requires quite some change management.

00:23:44: And Humanoid Robotics Pilot is really a good starting point to already begin these activities.

00:23:51: Physically, I generally can provide short-term paybacks but if i look specifically at human reads the purpose of the pilot it's more getting the operational experience and preparing yourself in the company.

00:24:11: I mean, i can absolutely relate to that statement.

00:24:13: And... ...I think this is not something we only see in human-art robots and physical AI but also other technologies.

00:24:21: This whole aspect of training first.

00:24:24: so if you look at AI models or quantum algorithms We always have the same topic about how do we train a model to be able to operate in reality?

00:24:37: That's different than physical AI both for the human interacting with a human or robot, but also for this software and system behind it.

00:24:47: So one thing that I think can be observed everywhere is operators do anything to find training ground in industry.

00:24:58: That's also a trade-off of being seen as risk sharing operational knowledge within Human or Robot OEM.

00:25:10: How do you see that trade off from an industry perspective?

00:25:13: So how do industry leaders approach and what can we expect from this development.

00:25:24: I think it could be a risk, an advantage at the same time.

00:25:27: so you need to really carefully and strategically manage this.

00:25:33: but first of all your right.

00:25:34: as mentioned training data, real world training data in order to train the algorithms.

00:25:40: Yes simulations are getting better and better but there's always a simulation-to-reality gap.

00:25:47: so that is only certain amount of artificial data we can use.

00:25:52: And this also is nice part about the approach China currently doing with deployment.

00:25:59: first because it gives you a fleet.

00:26:05: So, manufacturers have the advantage.

00:26:09: They have their facilities and processes so they have the ability to generate training data And this is something that can be offered for a human-read robotics provider.

00:26:22: The question of what can we get back from this?

00:26:27: I think an important element here in technology access This one part We haven't really talked about when we talk about piloting, but a core reason to already start piloting today is also to secure access to the technology.

00:26:45: Because once one leads will mature and capabilities increase.

00:26:50: supply chains won't be able to keep up so demand will significantly higher than supply.

00:26:57: So if you're all ready Via partnerships, today is the key access to human reads in the future.

00:27:05: This can really be a competitive advantage but of course as you mentioned You need to be careful what you share.

00:27:14: Of course training a human read and logistics process.

00:27:18: this Is most of the time not very critical But if we talk about processes that contain a lot of proprietary company knowledge This really needs to be carefully evaluated, because once a human read learns the process it becomes part of its capability and then can.

00:27:38: And honestly will transfer to competitors as well.

00:27:41: Absolutely!

00:27:44: A trade-off that every company need think about for themselves.

00:27:49: going back to Shuai to Shanghai We've talked about China's speed, and also if we look at this question on how do operators train their models?

00:28:03: And how do they move from pilots to reality.

00:28:07: What do Chinese operators do differently?

00:28:09: And what can Western companies learn from that?

00:28:13: because apparently you are ahead of the curve.

00:28:20: I think the key is the speed of iteration.

00:28:26: The Chinese operators, we usually don't wait for a perfect answer.

00:28:31: But instead they deploy robots that are good enough and go into the production line going to their warehouses.

00:28:40: Instead of waiting for more polished or more perfect product For example BYD They started to insourcing the wrong robotics technology from a twenty-twenty two.

00:28:57: And now this year they have integrated into their seven generation of their robots.

00:29:04: so four years, seven generations very fast and they deployed quite a lot of sampled.

00:29:13: I mean eighty percent readiness off the robot have been deployed by BYD in many production lines of vehicle, in production line of battery.

00:29:26: And they cannot achieve one hundred percent off the target.

00:29:29: however eighty percent is a very good result for BYD.

00:29:35: So this was one and another thing again data.

00:29:39: just as Rihanna mentioned The data coming from real world cannot be matched by the data coming from laboratories, especially for long tail use cases.

00:29:54: Because in real world factory you have un-studied packages and dirty spots of all.

00:30:03: your spot always have some workers, ad hoc needs and working on the road.

00:30:14: So this kind of real and long tail use cases can be very hardily simulated in a lab.

00:30:23: so that's why we mentioned there are fifteen thousand unit deployed is to experiment with the real data.

00:30:37: Another point which is very important that the real pressure in China, we are facing a decreasing labor force.

00:30:47: We projected it would be declined by twenty-five to twenty four percent.

00:31:01: two hundred and forty million people.

00:31:09: So what does that mean?

00:31:11: This kind of this scale is just the same scale for the population in Brazil, so it's quite a big label force in China.

00:31:21: The key question to business owners in China are whether or not they should deploy robots.

00:31:29: If I don't deploy a robot now next day there will be no workers for me to do the job.

00:31:37: So this is a very key reason and that they did real pressures from their business operation, so Western operations can learn from it.

00:31:53: I would say maybe in the perfect version you are waiting for cannot come out forever because the perfection will become out of the real world.

00:32:07: Very dirty data, very daily operation data so that for example the figure AI they use nearly.

00:32:18: I mean eleven months to pilot over ninety thousand components, To move over ninety-thousand of the components.

00:32:28: So this is quite a lot.

00:32:29: however This number ninety thousands it's just one week workload for Just one factory in China within One hundred unit off robotics.

00:32:39: so only After deployment you can see The real problem and then there will iteration and then the real trend of the real world.

00:32:49: So my suggestion is that just to pick up a very narrowed or really unique use case, find production line... Just put around ten robots in it And this robot.

00:33:07: maybe they're not perfect but are good enough.

00:33:09: Then you just operate and month after month you get your data.

00:33:14: That's extremely valuable insights.

00:33:16: Thank you, Shohai and also the dimensions that you were talking about tremendous dimensions I think The topic of an aging society.

00:33:27: That's not only China specific, although it might be severe in your case.

00:33:31: But also in the Western world.

00:33:33: we have generational contracts which don't work anymore and Yorna is elaborated as well on supply chains to develop.

00:33:45: Now Shuhai said that its no matter of if but when.

00:33:52: Thomas looking into you direction Talking about supply chains and the need for supply chains to develop themselves.

00:34:01: The question is, apparently time has already come to act upon it?

00:34:06: Also if we look at your recent study you claim that this can be And one of the numbers that striked in there was even actuators only would be a seventy-nine billion dollar market until twenty thirty five.

00:34:24: So, um... The question is if this is next automotive industry?

00:34:29: Does that not only also pose a chance to existing challenges in the automotive industry, and what can we learn from them?

00:34:38: We see first automotive suppliers moving forward.

00:34:41: So how real is this

00:34:43: opportunity?".

00:34:49: As you said it's tremendous with multi-billion dollars or sometimes doesn't billion dollar... component opportunities, or also assembly opportunity.

00:35:00: So as a supplier you can... It's huge opportunity be it for automotive suppliers industrial suppliers EMS companies semi-conductor company software companies so far lot of companies in particular those that are today more on the system and components side.

00:35:17: this probably is an opportunity of the century they have.

00:35:24: The reason for that is, in the human and robot parts are relatively complex.

00:35:28: You need on one side you don't have a lot of space.

00:35:31: it's a mobile system where energy consumption matters.

00:35:35: It also systems really used intensively over long life time.

00:35:41: so we have high requirements.

00:35:43: We have high scale or on the components side, also to a certain degree global supply chain.

00:35:52: While together with Shuhai we argued that on the OEM site the ecosystems are separate On the supplier's side this tends to be more let say global supply chains.

00:36:02: So there is huge opportunity for companies now To size their opportunities and go into market.

00:36:10: Absolutely If there is such a big opportunity, what do suppliers need to do?

00:36:19: Because you cannot just show up with the catalogue as we did maybe twenty years ago.

00:36:23: and here we are.

00:36:25: Please take me into your supply chain.

00:36:26: so how does winning strategy look

00:36:29: like?".

00:36:33: Yeah that's actually critical point even though on the technology in general business model side human or robot components similar to let say industry and automotive The business model and the success factors to enter are very different.

00:36:49: If you just look at the customers, basically all of these customers either startups or younger industrial more software and AI driven companies.

00:36:57: so we're not talking about an established industry company but a startup that's actually fighting for survival and fighting to be dominant player in this industry.

00:37:08: what matters is speed it's competence and also getting help with the most important issues.

00:37:16: We all argued that data is key, but as a supplier you need to come up with some clever and creative topics like... How can I enable my customer, my OEM to be successful?

00:37:29: Can I provide video data for my own factory.

00:37:33: For example place a conditional order for human world robots.

00:37:36: there's many ways you as the supplier can help your potential customers but it goes much beyond what they say.

00:37:42: is that traditional suppliers do in their business model and requires also on operations side at the Operations Department working together with the Business Development Department because both have go hand-in-hand.

00:37:54: so these The operation side and the business side are for suppliers not too separate topics.

00:37:59: You need to treat this also as one integrated topic, only then we believe you can be successful.

00:38:05: Probably many companies in China a little bit more open to do it on the Western side.

00:38:09: there other first company than move into human or robots but others seem to be a little slower.

00:38:15: I think some of European and American suppliers is at risk.

00:38:20: We talk about that risk to catch the train and then we look again at the far east.

00:38:28: In China, we have at least thirty nine suppliers already on thirteen tech giants automotive companies that actively develop human art robots.

00:38:42: Does a supply chain opportunity Look different in China?

00:38:48: Or were Chinese companies more clever to the complex challenges, as Thomas just called them?

00:39:01: I think there's no good model or better model.

00:39:07: But that opportunity of supply chain is different.

00:39:11: in China We have over fifty clients and developing their own supply chain, and their own ecosystem.

00:39:25: And in the western maybe just five to ten clients picking up the suppliers.

00:39:32: So it's multiple-to-multiple or a few.

00:39:36: so there are different models.

00:39:38: In China we have different investment thesis You bet on the pigs' and the shovels and with the highest value.

00:39:52: So it's totally a different investment logic, And in China we can imagine that one supplier Can at same time supply to provide their products and components.

00:40:04: over fifty OEMs.

00:40:08: They sampling and they iterate their product and don't rely on any single clients To de-risk The portfolio thing.

00:40:18: And the leading OEMs in China, they are already developing their own ecosystem.

00:40:25: One of the leading humanoid robotics OEM's in China.

00:40:30: I talked to them and that they have invested over sixty suppliers In China.

00:40:36: very crazy Over sixty suppliers.

00:40:39: They've been invested Including actuator software even the B-to-P model, the OEN supplier.

00:40:48: And they themselves just developed their own software algorithm big model and the exterior hands such kind of key component.

00:41:00: so this is one another point that you know.

00:41:04: the electric vehicle.

00:41:06: we have a quite immature supply chain off EVs.

00:41:12: five to six out of ten electrical sold in the world are coming from China.

00:41:18: So we have quite a cost-efficient and mature supply chain, battery in motors reducers power modules thermal management.

00:41:31: over sixty percent can be reused or mobilized.

00:41:39: So we don't need to start from scratch.

00:41:43: And another point is that the cost of structure, I take a reducer for example.

00:41:51: so one year ago The reducer bond costs in China Is around maybe three thousand RMB But today it's just half It's around one-thousand to five hundred RMB and the durability increased from maybe five thousand their hours to eight thousand ours.

00:42:14: So in China the supply chain of robotics is not established for robotics, it's established for many industrial sectors such as EV and such as EV tools including the robotics.

00:42:29: so I mean summarize the opportunity.

00:42:37: many clients diversify the client's portfolio, mature and cost efficient full scale of supply chain.

00:42:46: And reuse of electric vehicles ends some in-house capability after EV OEM such as BYD.

00:42:56: so this kind of uniqueness cannot be easily copied by Western countries.

00:43:02: So that is China.

00:43:04: we bet on Whoever makes the best robotics, you have to buy from here.

00:43:12: You have to by components form here.

00:43:13: so that's our industrial target.

00:43:17: So thanks very much for those insights.

00:43:20: it seems there is a better interconnectivity already and much we can learn in Europe And long way to go but still also lot of opportunity for us.

00:43:31: Maybe coming into final question Looking at Europe, we have around twenty two startups.

00:43:38: Eight hundred million of funding.

00:43:40: if we compare the numbers to the ones that we've heard from Schwaie earlier We only have a projected one hundred units for twenty-twenty five.

00:43:49: so we had that.

00:43:51: Has Europe already lost their race or is there still a chance first to succeed?

00:43:59: Jonas maybe starting with you

00:44:02: I don't think we have lost our race or one of the big gaps we still see is on the AI algorithm side.

00:44:14: This really what it's needed in order to unlock additional use cases, which provide additional value at top of traditional automation that already have and as mentioned and discussed access-to data unlocks the further developments on the algorithmic side.

00:44:38: And yes, there China has already a great advantage due to significant larger deployment numbers.

00:44:45: On the other hand if we look at Europe We have historically very strong automation industry.

00:44:52: Um, we are very strong in terms of manufacturing So we have the factories?

00:45:05: if we start to increase number of deployments, increase our ability to capture new data and join together all the different competencies that we have.

00:45:21: We are able especially keep up or even gain advantage on the algorithm development side And this is really an area needed to unlock full potential.

00:45:35: Thanks a lot Thomas.

00:45:36: How would you see it?

00:45:37: Yeah, I agree with Jonas.

00:45:39: So we haven't lost the race and actually these more than twenty startups They're also active in acquiring additional funding.

00:45:46: that's been quite successful recently so And we have actually pretty strong companies in Europe said how to have partnerships of automotive companies an industrial company.

00:45:54: so We have a good base In particular if the you know robot companies the established robotics companies answer automotive suppliers and OEM work together.

00:46:04: We have a strong chance.

00:46:06: And not only do we also lose this race as Europe, because if you imagine the world where we are importing human robots from the United States and China... ...we either have huge risks or let's loose a lot of value.

00:46:22: so we need strong components suppliers and OEMs.

00:46:26: We have the base and I think with the right level of interaction, support or policy frameworks.

00:46:33: Operators that are moving fast in pylons... ...I'm positive.

00:46:38: we can do it as Europe!

00:46:40: Thank you.

00:46:41: Shuai.

00:46:42: over to how would you answer this?

00:46:47: No one is losing a game now because we're just starting for maybe two years.

00:46:56: Especially Europe is also losing again, but we need to accelerate because frankly speaking the Europe enjoyed the biggest base of automotive industry and a very good base for industrial suppliers tier one or two suppliers.

00:47:18: We have a perfect environment for automation good environment for deployment and a very real case of plant in many countries.

00:47:30: The key is to connect the dots, so they plant connected business owners with local robotics OEMs operators more importantly connecting them with ecosystem especially supply chain.

00:47:50: only doing that you can transfer the very good industrial basis into an advantage or a first mover's advantage of new industry.

00:48:04: Especially for the advancement of automation, higher value can be created and technology platform of industrialization is owned by Europe themselves.

00:48:22: So my suggestion is that around the world, we are facing some regulation.

00:48:29: We're facing some wars however and we need to connect all of those by communication... ...by exchange resources within the supply chain And technologies can be exchanged also.

00:48:44: Thank you all very much.

00:48:46: So what I take from this is a lot of optimism and many impulses that our dear listeners can take, and to Shui Thomas-Yunas thank you for your time!

00:48:55: In the next episode we'll go

00:48:57: deep

00:48:57: on another

00:48:58: technology reshaping

00:48:59: how

00:49:00: industries operate.

00:49:01: subscribe

00:49:02: so you don't miss it.