← Everything Electric Podcast
Everything Electric Podcast
The Tiny Tech That's *Really* Powering Electric Cars! | Lars Regers, NXP Semiconductors
00:58:08
In this episode of the Fully Charged Show Podcast, Imogen Bhogal sits down with Lars Reger, CTO of NXP Semiconductors, to explore the critical role that chips play in the electric vehicle revolution. With the average EV containing around 10,000...
Loading summary
Transcript79 lines
- [00:00]
Lars Ragers
Foreign.
- [00:11]
Imogen
Hello and welcome to another episode of the fully charged show podcast, where today we're catching up with Lars Ragers, who is the chief technology officer for a company called nxp. Now, NXP are responsible for designing and manufacturing a huge proportion of the semiconductors used by the automotive industry. Now, for our longtime listeners, if you're thinking this is a little bit of a deviation from your typical topics of electric vehicles and clean energy technologies, let me explain why we're catching up with Lars. Well, first of all, he is just an absolutely extraordinary person. He is so interesting and has so many weird and wonderful analogies to make this topic really, really digestible. But more importantly than that, the semiconductor industry, and particularly when it comes to automotive semiconductors, gives us a really, really good indication of just where we're headed. And the reason is that an electric vehicle has 10 times the amount of semiconductors compared to its internal combustion engine counterpart. And additionally, as we head towards a more connected set of automotive features and more autonomous driving modes, what's happening in the semiconductor industry also gives us that real sense of what we can expect from the automotive industry. And then on top of that, as we move towards these more software defined vehicles, so called, actually the role of the semiconductor companies becomes even more important. Now, NXP's breadth and reach is absolutely enormous. And to put it into perspective, if they cease to exist tomorrow, just poof, disappeared, then many of the cars on the road today just simply wouldn't start. So he is here to give us a glimpse into what we can expect from the world of electric vehicles and the future of automotive. Now, before we get into the conversation which was so enjoyable, I must say I want to define a couple of terms just so that we're all on the same page. Now, first of all, we talk a lot about vehicle architecture and when we say that, we simply mean the blueprint. The blueprint. The blueprint of a vehicle so its chassis, the technology inside it and the layout. We also talk a little bit about tier ones and in that context we're talking about the suppliers who provide equipment and technology directly to the OEMs. And when we say OEMs, original equipment manufacturers, typically in this context, we're talking about the automotive manufacturers. So I hope you're ready for a very interesting conversation. But before we get to that, a very quick advert break.
- [02:37]
Duracell Energy
We've teamed up with Duracell Energy to celebrate their brilliant ecosystem of home energy products and their platinum home owner offer by giving away a Duracell bunny to win. Simply watch to the end and answer a question about fully charged.
- [02:52]
Host
Lars, thank you so much for giving up your Thursday morning today. Can you kick us off by telling us who you are and what you do? Yeah.
- [03:00]
Lars Ragers
Imogen, a big thanks for having me here. It's a real pleasure to talk to you. I'm the CTO of NXP Semiconductors, physicist by education with an MBA from London Business School. And NXP Semiconductors is the former Philips Semiconductor. So a lot of people might recognize still that that name and that NXP Semiconductors got spun out of Philips in 2006 and married, so merged with the former motorola semiconductors in 2016. And with that we are a pretty broad semiconductor supplier. So chip supplier in the industry, basically for everything that is not big storage, big compute, big AI. So not the stuff in the cloud, but everything around you. What nerds like me call the edge. From banking, cards, passports, 60% of the chips that we are selling is going into into cars, industrial robots. So all of that type of electronics for these devices could come from nxp. And that is our value proposition.
- [04:10]
Host
Quite a broad remit there.
- [04:12]
Lars Ragers
Yes. And that is of course, the interesting stuff when you want to build, let's say, little robots around you. And the analysts love us at the moment for the reason, because the forecast is, of course, we are moving from the on demand world. You need an Uber, you press a button on your mobile phone, Lars needs a pizza, press a button, get it delivered. So Hongli is on demand, acting into the world that anticipates and automates. And for that you need to have, and the analysts say, 50 to 70 billion smart connected little robots around you in the next decade. If you believe that NXP is one of the suppliers for that electronics, for these 50 billion smart connected robots, maybe even for a majority of them, yeah. Then of course that has an influence on our share price. Right? I mean, that is how the entire game is going at the moment, and.
- [05:01]
Host
Safe to say, and I hope you don't mind me asking this, but we met some months ago and you described to me that you sort of ended up at NXP a little bit by.
- [05:08]
Imogen
Accident, and I'm sure that.
- [05:11]
Lars Ragers
So let me give you a story. No, I started for Siemens Semiconductor, so a physicist. I wanted to do medicine, medical devices during my studies. Then I started working with electronics, semiconductor electronics, made it to Siemens Semiconductors, which is today Infineon, started working there, then moved to navigation system development in the automotive industry, to Siemens vdo, which is today Continental, and did navigation system Development, then telematics. So I landed as a hybrid in this semiconductor automotive ecosystem. And then NXP called and said, hey, would you like to join us here? We need someone strategist for automotive. And I moved in 2008, and that was precisely the moment where the Lehman crisis hit the market. And NXP just had been spun out a little bit earlier, not really clear on how the portfolio should change and how the setup of the company was. Long story very short, I came here and they told me that they still have cash for one quarter. So I remember me calling my wife and said, hey darling, stop packing bags. I might be soon back, so maybe in six weeks. I need to search a new job again. And since then, a fantastic journey started. I thought it was the mistake of my life, but what really happened was that private equity, who was owning us in those days, really, really in a tough way, looked into each single segment and said, is this fitting to the NXP DNA or is it just because your old mother company wanted to have a chip of a certain kind that they asked you to produce it, even if it does not exactly fit to your assets? And that refocusing of the company into a, yeah, in my language, I would say now into an interface company, into a sensing company, into a smart power company. So all these analog electronics that helped and that was, of course, then suddenly we were profitable again, very profitable soon, and we could acquire Freescale, the former Motorola Semiconductors. And Freescale has all these small, medium, large type of compute brains. Now we suddenly had the sensing part, we had the networking part, we had the actuating part, and we had the thinking part. And how do you build robots? You sense your environment, you connect to the cloud and get your information in. You send this to the little brains, you think of a smart advice, and you send this smart advice to the arms and legs of your little robot. Since think connect act, we were complete. And from the auto side we had the functional safety, so braking systems, you know, fail silent type of architectures. We knew how to do functional safe electronics. And we were building banking cards or mobile phone payment. So we had all of these EGAF chips in passports. We had the security as well. Now how do you build a robot? You sense thing, connect act. And you never would allow a robot to take tasks from you, responsibility from you. I mean, the first moment where Your fridge orders 500 liters of milk for the weekend, this thing is off the grid and you go shopping yourself again. Yeah, first time your smart thermostat turns Your, your, your house to 50 degrees centigrade and your aquarium is boiling. You go change the heating yourself again. And the first time your autonomous vehicle goes diagonally over a crossroads, you start driving yourself again. You don't trust the device. So there is these four ingredients. Sense, think, connect, act. And then you have to trust your device, otherwise this market does not exist. And Inert defines trust as safety and security. So the device always has to work and never turns against you and it cannot be hacked to turn against you. If we are world champion in these six areas, we are the robot company.
- [09:18]
Host
That is absolutely fascinating and I've never heard it sort of described quite so clearly. So Sense, think.
- [09:26]
Lars Ragers
Yes, Sense, think, connect, act, security. Exactly.
- [09:31]
Host
When you cast your mind back to 2008, could you foresee that happening?
- [09:36]
Lars Ragers
No. So I would say, I would love to say it was all strategic wisdom. It was not as so often, but it was also not sitting on our hands and wait, what is happening from the outside world? I mean, there was this dream and the view of what can this company really do with its assets? So no one at that time was saying, we need to be somewhere in the cloud, we need to do big storage, big computer. That was not the point. We are the analog company, we are the interfacing, I called it in these days. We are the interface company. We have everything that is connected to a wiring harness. So we have all of that. We have good sensing capabilities, car to car communication. We started with radar, you know, all of these type of things came on board, followed a little bit that structure, and then Freescale came on board. And it was a complete accident that they had told me at one of the keynotes at an intelligent transport fair. They said, we have just announced this morning that we gonna acquire Freescale. So we have done the signing, but we have not done the deal closing. So, Lars, you are not allowed at all to ever talk about free scale here. Don't make any comment on the company. That is legally not okay. That is impossible. So I was standing there and I had a slide for a domain based car and said, look, this is what NXP is good at. And I had colored blocks on that thing and it was like. And there we are, market leader for radio, for analog can transceiver, you know, all of these type of devices. And then I looked at that slide and just that very moment I said, and for all the gray areas, others are market leaders in most of the gray areas. This is companies like Freescale. And it was dying. Laughter in the entire AUDIENCE but that was the kickoff moment to say, you know what? Honestly, we are rather complete. I'm not saying we have the best silicon everywhere, but we are rather complete. And from there, the story was spinning. Our customers see it in the same way. And only last year, we really coined the word intelligent systems at the edge, or robots. So the story, I mean, the slides are unchanged since almost a decade now. With that value proposition, we can build all of that machinery, and that is, of course, evolving. How much of AI do you need? How do you make sure that your AI, your brain, is functional, safe and secure? I mean, that was all these hula hoops that we did in 2018. I know journalists were asking me, said, last van is NXP bankrupt. I was like, why? Yeah, because it's just a big AI piece in the car and then autonomous driving is done and no one needs your silicon anymore. And I was saying, like, look, a car is as much a brain on wheels as you are a brain on shoes. And then, I mean, it was hilarious. But in principle, there was a reason why in 2018, this big promise, in two years, we are all driving without a steering wheel, did not happen because it was this misperception of just throw AI in and you are good. No, we had to take a step back and had to say, okay, how does a robot really work? But you're talking to a biological robot. And this biological robot here has a very different architecture. I have a real time system in my spine that has reflexes. If I stumble over a pothole, my legs, my spine says, straighten your legs and stabilize yourself. Yeah, I have no clue what I'm stumbling over, but I straighten my legs, right? And I'm fast. This is a inert language, ultra fast and simple real time system. In automotive language, you say, this is an ASIL D type of deterministic system, as it is the highest functional safety level for my heartbeat, my temperature control, my stability control, my stomach control, all of that. And then on top of that, I have my cerebrum. And that is not a real time system only. That device is an AI system. And that device is just at this moment thinking, what is the next funny sentence that I say to you while I'm sitting straight on my chair here? This is the cerebellum and I can do movements with my hands and still talk to you. So this is a multi layered robot architecture. And what we are trying to do at the moment is we try to bring these robot architectures into cars.
- [14:28]
Host
So there's. Oh my gosh, there are so many different things.
- [14:31]
Lars Ragers
Sorry about the long monologue.
- [14:32]
Host
Yeah, no, no, no, I enjoy it. And I knew when we invited you on that my job would be incredibly easy because you are such an incredible speaker. And I can sit back and relax and listen intently, but there are so many different directions that we could take this. This conversation. And before we do, I want to sort of take a step back and just ensure that we've got all of our listeners on the same page and start with some really, really basic terms. So for the avoidance of doubt, I've got two questions. First of all, what actually is a semiconductor? What is a chip? And secondly, why is it that we need different chips for doing different things?
- [15:09]
Lars Ragers
Yeah. So just to make it very, very tangible. In a normal car, so normal combustion engine car, you have thousand chips, about 1,000. In an electric vehicle, you have up to 10,000 chips. And they are all very, very different classes. You have, for example, these little brains that are like your pocket calculator, are calculating things. So this is purely what we nerds call digital electronics. Bits and bytes, shuffling. And you do this with tiny little transistors. And these transistors are switches. Switches that you realize on the surface of a piece of silicon, nothing more than that. You can use them as little amplifiers or as little switches. In digital electronics, you are just switching and you start calculating whether you build calculators like in your computer, then very big systems in the cloud. Yeah, but everything around you, your mobile phone has these little calculating devices. Then you have memory next to that where you can store stuff that you want to memorize. So like your sheet of paper, you write down stuff. And then what you also have is you have interfaces to the analog world. So in other words, you can measure temperature, you can measure a state of charge in a battery cell, in an electric vehicle, you want to measure that you're not overcharged because otherwise your battery cell gets on fire, and you don't want to destroy the cell by depleting it too much. So there is a lot of fine measurement electronics. And if you do this in a smart way, you can condense that all on one little piece of a silicon crystal, and that is this famous semiconductor. So if you look into your old radio, where you have amplifier tubes on and a big, big board with a lot of switches and coils and so on. We try to condense all of that into a piece of silicon, over mold it with plastics, and have a couple of connections to the outside world. And this is these little legs that you see normally coming out of these little black squares.
- [17:23]
Imogen
Yeah.
- [17:24]
Lars Ragers
And this is the chip when I talk about it. Now, this chip is a collection of these electronics. And normally these electronics can run completely software free. But smart guys of course took these microcontrollers and said, hey, this is customizable hardware. This thing cannot only do measure the temperature and over a certain threshold, switch on the heating. No, you can multi purpose use it. And this is what you're doing with your computer, but for multi purpose, using your hardware, you have to program it. This is what software is doing. So hardware and software in these type of systems cannot be detached from each other. There is no software that can be used without any hardware. So you need your computer. Otherwise what are you doing with your software? It has to sit somewhere. And a hardware of that complexity level, without software is also totally useless or would only have one purpose and you cannot change the purpose. So therefore, robots are machines that are customizable, take certain input and you can steer it, it can learn, you can program it, and it takes a certain action. Or in other words, you use software to customize your robot. And if you use software to customize the functions of your car, you have a software defined vehicle. But that has nothing to do with software only. That is one of these myths where the people are stuck and the battle will be won in software. And it's all software. No, sorry, this is wrong. It is a robot architecture that we are building at the moment. And this robot architecture gets configured by programming it. This is what that entire thing is doing. So in other words, I hate the terminology software defined vehicle. I always call it playfully. It's my rolling robot.
- [19:38]
Host
Yeah, because it's not defined by the software, it's brought to life by the software.
- [19:44]
Lars Ragers
Yes. And you cannot detach it from each other. Each part is totally useless without the other. And what you really want to do is in principle, you want to have a intelligent system that transports you and that is trustworthy and efficient. Period. And now, whether this is driving in the uk, so in my point of view, where the cars are driving on the wrong side of the road, or here in Germany. Yeah, in principle, on the software definition, that is exactly one bit that you need to change. Drive on the left lane or drive on the right lane. That is then software customizable, super easy, okay? We talk about the steering wheel that has to move or whatsoever. That is not the part. But your assistant, your driving robot is agnostic. That is really then flipping the software model. And that is what you can of course do with these robot architectures.
- [20:39]
Duracell Energy
Gosh.
- [20:39]
Host
Okay, so I feel like there's a ton of of technical questions that we.
- [20:44]
Imogen
Need to get into in terms of.
- [20:45]
Host
How some of these things are made possible and how we continue that path to the 50 to 70 billion robots that we're expecting in the future. But I think we also have to start by addressing how strange the automotive landscape presently is. We're operating in a period of time where there are changing tariffs, geopolitical tensions. I think there are both trade and culture wars attached to electric vehicles, to semiconductors, to all of it. And what's fascinating about NXP is that you have a very, very global footprint. You work with so many different OEMs, and I wonder whether that makes you quite exposed to some of those things or actually that kind of builds a resilience to them and how you're having to adapt and predict some of these things as nxp.
- [21:30]
Lars Ragers
Well, Imogen, that is a good point. Honestly, there's hardly anyone whom we are not working with. So if NXP would be irradiated from the globe, I guess most of the cars would not start anymore in the one or the other way. And that helps us big time because that drives global innovation. Now, what are we seeing at the moment in the car industry? I'm not talking geopolitics, but what are we seeing in the car industry is we have started 15 years ago, 20 years ago, by changing the drivetrain. Suddenly, people came in with electric drivetrains. Now, until then, the automotive industry had huge barriers of entry. You know, I'm the grandmaster of exhaust gas regulation, and you try to compete with me. How much know how did you have to bring in a combustion engine vehicle? There is thousands of moving parts in an e vehicle, only 300. So from all of that, the biggest disruption was that transformation, the technical transformation, saying, hey, plane level field again. The ones who have capitalized the 140 years since Maybach and Daimler put a combustion engine in horse carriage, their barriers of entry are collapsing and others are coming in. Now, that is of course, particularly nasty for the Germans who had the Daimler Maybach inheritage and were optimizing for 140 years the horse carriage into a better driving machine, an ultimate driving machine. You know all the slogans. Now there is newcomers coming in and they had, of course, difficulties in entering, yeah, driving range is not good, battery stability and so on. But for most of These innovations, for 95% of these innovations, electronics and the attached software were needed. So actually a pretty good game for us. And what I Said earlier, an E vehicle, medium upper class e vehicle has 10 times more chips than a medium class combustion engine car. So for us, square meters of silicon, hey, that's the game to be in. Yeah, we want that. Now we worked, of course, with all these partners. The next thing that came was like, hey, how do we do autonomous driving? I briefly tried to touch that. And we have had to re architect from this crazy belief in everything is AI with wheels into, hey, there is an entire robot architecture. And from there we now see that this entire complicated rolling robot story spills over into other robotic systems. And you see even tier 1s not necessarily here in Europe or in North America, but in Asia, they are doing the next move and say, hey, can we use our automotive know how, our automotive building know how and go into robots or even flying robots names, drones, so air taxis and all of these type of things. And you see that at the moment as the next innovation step for us. Very, very good. If we can lead the innovation. As always, if you are leading, you are the transformer. If you are lagging, you are transformed. And that is not necessarily a pleasant thing. So I don't know who will be the big winners in 10 years from now. What I guess is a lot of the innovation is over in 10 years from now. You know, when everyone can buy from every tier one an autonomous level four system with either a combustion engine or an electric drivetrain. Yeah. Then you are in a mode like 25 years ago where the car industry rather was about quality up and cost down. 25 years ago, I mean, it was like, okay, do I have an Audi, a JLR or BMW navigation system in my car? But by and large, the data sheets and the form factors were all very, very similar. I guess in 10 years we will be in a similar situation again, maybe with totally different players, but again, consolidation is going to happen.
- [25:53]
Host
That's really fascinating.
- [25:56]
Imogen
God.
- [25:57]
Host
And if we are moving in that direction where there's a tremendous amount of innovation that needs to happen to get to that very connected increasing levels of autonomous, increasing levels presumably of electrification, given sort of emissions targets, et cetera, et cetera, that is a lot of chips that need to be made. And obviously lots of people, probably the first time they were really conscious of semiconductors and chips was during the chip crisis, during the pandemic. Would we anticipate that something like that could happen again? If we need, you know, we're talking 10 times more than we presently need.
- [26:32]
Imogen
At this point in time.
- [26:33]
Lars Ragers
So the sheer amount is not the problem, but the supply Chain dynamics. Look what what happened to us in 2020 was the amount of vehicles produced even went slightly down due to the COVID crisis and so on. So it was not like there was not a doubling in car volume that created the crisis. But if each car that is produced suddenly needs 10 times the amount of chips, then each old architecture that goes out into the market that is replaced by a new architecture is nine times the amount of chips of that old architecture on top. Now this transparency in the supply chain can only come from the car OEMs because they only know when and how their new architectures are hitting the market and how many chips of which kind they suddenly need. So there was communication room for improvement to say carefully between 2017 and 2019. Now that gets nasty. If you think of it takes about nine months to make it, sorry, six months to make a make a chip from order to the little black square falls out of the factory, is tested, is shipped and is reaching US Tier 1 or OEM. So six months in that it takes about five years to build a semiconductor factory. And if you come to the conclusion that you have a shortage in 2021, then pedal back to 2016 or 2017 and say who at that point in time would have known that already? Okay, so now today we are communicating much, much more carefully and say are we sure that we don't run into a crisis again? And so on. And everyone int actually has gotten it. But the TR1s are at the moment especially the TR1s are so much under commercial pressure due to the recession that they cannot keep chips on inventory. Normally you would say okay, all easy, keep one year of chips in stock. They are tiny, you can store a lot of them, keep one year in stock, tie up the capital and you are robust against all demand swings. Because when you order again and half year later you have the volumes that you need and so on, you can hedge yourself. Now, if you are super, super squeezed financially and you are between chapter 11 or good inventories. Yeah. Then of course you are reducing your inventory levels. And some of our guys in the supply chain I know are reducing inventory levels to two to four weeks.
- [29:13]
Imogen
Wow.
- [29:15]
Host
So weeks from previously a year.
- [29:18]
Lars Ragers
Pardon? Just a second.
- [29:19]
Host
So instead of keeping a year's worth of supply, keeping just two to four weeks.
- [29:23]
Lars Ragers
Yes, yes indeed. Now. And of course this is not because they are stupid, but because they are so much back to the wall that they say either I go chapter 11 and I have every non finished good in my inventory and then I can be robust against supply swings. Or it's the exact opposite. I'm, I'm reducing all tied up capital and I stay, I stay alive. Yeah. And there we are all a little bit nervous because that requires now a careful communication along the entire supply chain. Who is seeing the first demand, who is on which inventory level, how much do we need to reserve for which customer in case we see that someone is in our view under calling. But also we want to be commercially okay by not tying up the world's supply of chips into our inventories. We also cannot do that. So we are pretty Busy, different than 2019, to talk to everyone, making everyone aware of that risk. And it is always a short term risk. What I mean is a two to three year type of shortage risk where an allocation can hit us again after more than four years. You have the manufacturing sites upgraded with your capacity. That is not an issue.
- [30:45]
Host
So that's so fascinating because I guess previously, let's cast our minds back to 2016 when you say someone might have an indication that those architectures were changing, that they would have a different requirement from a chip perspective. And it's annoying that that communication gap existed. However, if we fast forward today and actually it's people like yourselves, nxp, creating those software defined vehicles, those robot architectures, does that give you the ability now you are the innovator, you are the transformer to say to the OEMs and the car companies, this is what's coming, this is what's possible. And you can be that set the tone.
- [31:21]
Lars Ragers
I suppose so. A lot of things have changed since 2019. So everyone is a virologist these days and everyone is a semiconductor. So that has helped. Now why am I making that joke? Very simply, because in the past, downstream in our supply chain, very often chips were perceived as a commodity. They come in any volumes, they have them on the shelf, they are available, which is true of course, in an oversupply situation. Now, in the meantime, everyone has understood these complications in the supply chain. And due to these escalation talks in this last chip crisis, we also have much better networked connections. So I very often talk to not only the CEOs in the, in the tier ones, but I have sea level contact to the carmakers. And now we are all talking. Say, Lars, what is cooking in your kitchen? What is the newest innovation? What would that architecture be and how can you help us? So you don't have a value chain anymore, but should I say you have a value network? So there is much more discussion along that strict line from us to the tier one. From the tier one to the OEM and every discussion around was forbidden. No, the OEMs are actively reaching out to us now. That accelerates innovation. Because if everyone knows what we are cooking up here in our kitchen that will reach the market in two or three years, then of course everyone is already like adjusting the architectures to that. And we can have a very educated discussion saying okay, if we bring that software defined vehicle architecture, we can reduce the amount of control units, we change the wiring harness, but each control unit has maybe much more mighty chips or more costly chips for more compute performance or more precise chips and so on. So you suddenly start comparing nodes line by line. And that helps of course in that entire innovation acceleration and supply resilience.
- [33:28]
Host
It is always so curious to me that so often these challenges come down to human communication and improving those networks to actually enable innovation rather than it is a technical challenge, say big time.
- [33:40]
Lars Ragers
And the way how, how one communicates determines also how fast an entire industry is with its innovation. What some of The Asian newcomer Tier 1s are doing is they grab the reference designs that we have on the shelf, so test reference designs and they fast absorb them and integrate them into their vehicles. And it is not that much of a. Let's first have a commercial discussion like with the incumbents and I don't need your full system solution, give me the radar chip, but I don't need your software. But these newcomers are sometimes enormously fast and say I take it all, let's later talk about the price, but bring me the innovation first. Now what that means is of course is they are very fast and sitting on the latest and greatest things that we are doing while some others of the incumbents are sitting in their old legacy architectures and say can I have your chip $0.05 cheaper? Now even if you have 100 chips for $0.05 cheaper, it is more expensive to refuel the car. So it cannot be the needle moving thing to be cost conscious in all of that. I understand of course cost in the end matters for a vehicle. But the game changer very often is be fast with your innovation and get going with that stuff. Maybe it's even cheaper to take reference design and Software from the tier 2s and then go to set makers, ODMs. Look for example at Foxconn, they are getting active in the automotive area. We have them at CS last year to jointly show what we are doing. But we are partnering up to say okay, our reference design, can you please industrialize that type of stuff and offer it then to the OEMs in industrial volumes.
- [35:30]
Imogen
Gosh. So it's totally changing the mechanism of who's steering the innovation.
- [35:35]
Lars Ragers
Yes, Imogen, the entire ecosystem for these newcomers is changing. And it's not necessarily said that they are just sloppy and fast and glue something together that is of lower quality because they use our stuff that is on the same functional safety, same qualification levels, same maturity levels. The only point is they are not doubling the R and D effort. They say, can I please have the entire thing? Look, here is already a demo, a placeholder. Can someone industrialize it for me? So in other words, put it into a metal box, put an automotive connector where there is a plastic connector and so on. But that is comparatively small tasks on top of what we have already on the shelf.
- [36:24]
Imogen
So then when you look at the, the global automotive landscape and you can see what's happening in the new organizations, you can see what's happening in the incumbents. Presumably you have to have very, very big, you know, Chinese walls between how you talk to different people. But it must be impossible not to say, guys, what are you doing?
- [36:40]
Lars Ragers
So of course we are not exchanging customer information from left to right, but being asked for. Lars, how would you build a rolling robot? I'm of course giving the best of my know how to anyone who asks very often then it's up to the, to the culture of the of the companies. Small companies are of course much more willing to, to take a higher risk and to be flexible. I give you an example. Incumbents very often go and say, yeah, we need to be multi sourced with your chip. So we take your chip and the chip of one or two competitors. That means if you're not careful, you are developing your software three times. Yeah, because you have to port from A to B to C on the microcontroller. Small companies I'm seeing in the recent months more open to say, what would it mean if we take it all from you? And by the way, can we please anchor a 10 year pricing roadmap? And with 10 years we are in better volumes so maybe we can become more commercially attractive to each other. And the other thing is, can you nxp, guarantee me that you can supply each of your chips from at least two sources and that I get guaranteed supply? You say we can guarantee that to you, but you of course cannot come, let's say in our restaurant and say, hey, I brought my hundred friends, now I want to eat steak, but tell me at least a week in advance that I can can buy stuff. So if this timeline works, if we are working with all of that. I think the industry, the small, the flexible guys are much better off going close with an innovation partner like us and say, okay, we do it all together, we save all the software and re qualification effort. But of course we still keep the liberty that we are not overly dependent from each other. If we get stuck, then we have to source foreign silicon, but not per se and not as a concept. Two or three times redundancy each time.
- [38:43]
Imogen
Yeah, gosh. Okay, I know I need to sort.
- [38:47]
Host
Of move on to a slightly different.
- [38:48]
Imogen
Question because otherwise I get a. I knew this conversation would be easy. Goodness. So just turning our attention then to some of the technical challenges that are around the corner now for anyone who doesn't really know anything about semiconductors, they would have heard things in the media around Moore's Law is slowing down. We're going towards sort of chiplets and different kind of modular types of chips and things. So I wonder if you could share some of the key technical challenges and opportunities that you're, you're currently looking at and how the semiconductor landscape is changing as a consequence.
- [39:21]
Lars Ragers
Okay, so I hope I don't get too nerdy now, but the Moore's Law was crafted for, for these digital chips. So if you want for a mobile phone or for a computer, you have these digital brains, your CPU on a computer that computes all and everything. And the computer has lots of transistors, billions of transistors that are all only good for switching between two states, 1 and 0. So switch on, switch off there. Moore's Law is very, very important because you want to bring on a, on a chip of the size of your thumbnail. You don't want to bring 5 billion transistors, you want to have 20 billion transistors to be almighty, to compute more and so on. For, for most of the stuff in our robots it is not the amount of compute, but what you want to have is you want to have precise analog measurement, battery cell, you want to have energy efficient compute. Your smoke detector has to live on coin cell battery for 10 years. But you don't need a computer CPU up there. You don't understand whether it's burning or whether they smoke or not. Yeah. So the task is very limited. Therefore a lot the vast majority of the automotive chips is not in the high performance compute where Moore's law is hitting you, but it's in the precision and energy efficient compute and analog measurement. Radio, radar, I mean I could go much, much broader now, but this is all not in that mainstream type of discussion. But it's all in totally different areas. How do you bring the best chips for these different applications? And they are sometimes, some people even call it in legacy nodes. So a precise analog transistor for battery cell measurements is 130 nanometer transistor, so huge compared to the 2 nanometer transistors. So 2 nanometers are 20 atom diameters, by the way. 2 nanometer transistors for digital circuitry in a mobile phone. So very, very different purposes, very different recipes. And this is where we are at the moment, building factories here in ESMC in Dresden for efficient compute. This technology is between 12 and 40 nanometers in digital computer or semiconductor language. And we have, for example, here in the Netherlands we have factories for this battery cell management, smart power technologies. They can stand plus minus 100 volts, can shift a couple of amperes of current. But they are of course the ones for the little actuators in your robots or for pumping energy in amplifiers, car radio or in battery cells. So for these energy switching, energy pumping devices you need totally different chips. And it's as much innovation in those areas. But the mainstream and everyone at Wall street only talks about the 2nm.
- [42:15]
Imogen
Yes, I was going to say, because it's all about that sort of increasing computer power.
- [42:19]
Duracell Energy
If you just, if you google the.
- [42:20]
Imogen
Word semiconductor and then click news, those are the kind of things that you hear about. But actually one of the things that you mentioned earlier about increasing autonomy and connectivity is how you take those analog things that are going on in the world, sense them, and then the robot has to react to them. And those should not be energy intensive things. You need all of those different signals that the car is perceiving to behave in an adequate way.
- [42:43]
Lars Ragers
If you wet finger in the air, make a calculation, and it is really 50 billion of these robots and we would build them on today's technology. They could easily consume three times the energy budget of mother Earth.
- [42:55]
Imogen
Wow.
- [42:56]
Lars Ragers
In other words, my dream on today's technology will not work. So I have to be functional, safe and secure for reliability in my robots, otherwise you would never buy them. And I have to be super energy efficient, running at the right level of energy, or send most of these devices into complete energy off mode for most of the time. So what we are building is we are building intelligent architectures with AI on board systems looking like a chocolate bar, a little PCB like this, a couple of our chips on board, and for example, you can glue them to an X ray machine at your radiologist. And this thing is running Large language models. So a lot of AI on these systems. An encapsulated AI model. So not as energy hungry and not as big as ChatGPT in the cloud. What this thing does is it has a very limited task. You get an X ray picture of my chest and then this thing has only to write the doctoral report for this X ray picture. And ideally the doctor even doesn't want to send that X ray picture to the cloud. And you have data security issues there. And everyone can see what Lars has. No, but there is this doctoral report, nicely written up. This is a man in the 50s. Two artifacts on the picture. One could be a measurement artifact. And here he should go to an mri. And then the radiologist reads that so he doesn't have to write the report, but reads maybe 50 of those reports a day. He cannot write 50 reports. Reads it. Okay. Yeah, that fits for Lars. Okay, I send it on. This is the report for Lars. So that entire thing, these limited tasks, but ultra energy efficient, we can do now on systems that maximally consume 8 watts. Gluing in 8 watts system to an X ray machine is doable, making it super efficient. Absolutely doable. And you can wake up these systems. Same system sits at your front door and observes what's going on on the street. With a couple of milliwatts consumption. It sees. Oh, in front of Imogen's house, there is a tree moving in the wind. Don't do anything. It's just this tree in the wind.
- [45:14]
Imogen
Yeah.
- [45:15]
Lars Ragers
There is a raccoon running from left to right. And if head from right to left. Don't do anything. Yeah, just some animals. Okay. Neighbor's dog is coming. Bigger animal, nothing person coming. Aha. Okay. Wake up the next big brother and say. Hey, can you. Can you check is this really a person? Yeah, this is Lars daughter. She has forgotten the key. Okay, Lars, do you want to open the door? Okay. Or hey, person with a black hoodie and a toolbox at 1:30am Shall I directly call the police? And you have these different stages where the full system consumes 8 watts. The watchdog consumes a couple of milliwatts. And these are these intelligent systems at the edge. This is what we can do today already. And this is what you see in very complex systems at the edge. Your car robot with wheels for the outside detections, of course, for the drowsiness monitoring, child present detection, all of these type of things. Even for your gearbox behavior, you have a different driving style than I have. Yeah, but we can learn who is in the car. In which mood is she or he? How do they like to drive? How did Imogen start? At the last traffic light, maybe. Now again, so do I need more or less horsepower? So you know, these systems can anticipate and automate tasks without you even knowing. And these have to be systems that are in the less than 5 watt range for most of the tasks. And this is the big innovation that we have to bring.
- [46:52]
Imogen
And if successful, what do you think? Things like ChatGPT in the cloud and those kind of very intense data processing activities, what could they learn from an energy efficiency perspective?
- [47:06]
Lars Ragers
Well, let me say it this way. Some people asked me some two months ago, was it a surprise that Deep Seq was announced? I said, but the date was a surprise because that was a political date when, when you want to make which statement, which nation is how good for a nerd? It was not a surprise at all because a human brain, a complete human brain consumes on average 20 watts.
- [47:28]
Imogen
Yeah.
- [47:30]
Lars Ragers
So that I need to switch on two nuclear power plants to ask smart questions to chatgpt in the cloud is for sure not energy efficient. So as long as I have not reached for the same quality of output levels between 10 and 20 watts, there will be every year now such a Quantum leap from ChatGPT, Deep Seq, different architecture and so on. I do not know how we get there, but I know what the benchmark is. A 20 watt frame with all its capabilities. You could even do this more extreme look at an ant. An ant as an insect is a fantastic transportation robot. They can even do platooning, right? I mean one ant follows the other end and you see them running around and so they don't fall off the table and so they have their limitations clear. But in principle, such a system runs on a brain power energy consumption of much less than a milliwatt. So in other words, there is even, even 20 watts are a complacent target. And the next 10 to 15 years will drive our innovation. Now, do we need at all to pump all of that data into the cloud or how much can we do at the edge? Therefore also the big AI companies are announcing systems for the edge today. It's not all in the cloud. I mean the value proposition of having these big coffins AI enabled monster units for cloud cloud compute centers. But this use case will be there, but it will not be the only way forward in the industry. The edge is coming big time and this is getting much, much cheaper. It's getting much, much more secure because you don't have to transport the Data back and forth, it might get more functional, safe, because you could have little co pilots next to these AI systems and you don't need to rely on anything outside of your direct proximity.
- [49:37]
Imogen
It seems like in this age of increasing amounts of AI, increasing amounts of data, increasing amounts of connectivity and automated driving systems and electrification, that for different chip companies across the world who previously maybe had quite specific roles of expertise, we're getting this huge convergence and we see it with, you know, for example, exactly as you've described, that there's opportunities at the edge that things like ChatGPT could be exploring in a bit more detail, but also people like Nvidia are exploring things like autonomous driving. Is that something that you've observed as well?
- [50:12]
Lars Ragers
Well, so let's say it this way. Each of these, these industry segments, so the analog world or analog electronics, the on demand world with the mobile phone and the cloud, the world that anticipates and automates, they all have their specific semiconductor needs. So when I say the analog world, it was in the early 2000s, it was gaming consoles, it was PCs, it was home entertainment, these type of things. And there had been a couple of guys who, who made money with that, but most of the semiconductor players did not. Then there were suddenly some people who were developing smartphones, so laptops in mobile phones, and they were data display devices and you needed storage somewhere. So the cloud. So suddenly you had cloud players, Intel, Samsung, AWS and so on. All of these companies got tailwind and you had mobile phone companies, Qualcomm, Broadcom, others who were successful. And now the recipe is changing again and you need to have, in my language, the sense thing, connect, act, safety, security, low power and AI environment. And now the jury is out again. Industrial transformation. Who are the winners and who are the losers of that change? And NXP wants to be, of course, very clearly one of the winners in that. Otherwise I go and retire. So joke aside, I mean that is, that is each time a chance for the players. But you have to change your recipe or if you by accident are sitting already on a recipe that is needed now, you will, of course.
- [51:50]
Imogen
Intriguing. So we have a lot to watch out for for the next five to 10 years of how this is going to unfold.
- [51:56]
Lars Ragers
Oh, absolutely. And a lot don't take anything for granted because what we had in the so Nvidia of course had a fantastic springboard to jump from because they were graphics cards makers. Graphics cards are massive parallel processing. For AI systems. You need massive parallel processing, not only for image handling. Therefore I Mean they had the technology that they could use and and then they made a fantastic success out of it for us. It is exactly the same things like what I just said. I mean, are the ingredients for robot building so intellectually you would get that? Yeah. Okay. We are sitting already on a lot of assets. We could make a success out of it, whether we do it to be seen. But at least we have these ingredients and now we need to have the focus and the will. The capability is there and that is how you will see the industry moving.
- [52:51]
Imogen
So I'm really aware that I have at least 25 more questions that I could ask you, but I'm going to limit it to one because I know.
- [52:58]
Lars Ragers
That we need to bring this very nice podcast marathon. Then let's do this for New Year's Day 24 hour podcast. Oh, we could.
- [53:08]
Imogen
Because honestly, the way that you, your analogies that you have and the enthusiasm in the way that you speak about this topic is just like so infectious.
- [53:15]
Host
So it's very easy to keep the.
- [53:17]
Imogen
Conversation going for a long time. But I'm going to limit it to 1, which is if there was one thing that you could ask for, one wish that you could be granted that would make your role as being the CTO of NXP that much more straightforward, perhaps would result in a few more sort of easier night's sleep. What would it be?
- [53:39]
Lars Ragers
I think the biggest thing is to really getting to platforms that everyone can use and also let's say bringing, bringing ecosystem partners along. Look, NXP is serving about 30,000 customers and for most of the customers, mass market customers, we do not even know what the guys are doing. So building microwave ovens and adding voice control, Alexa like to that type of things, there is a lot of need to scale our knowledge base. So if I would have a magic wand like Dumbledore, I would like to spread just this understanding how the building blocks of robots should be, what the capabilities are and what we jointly need to develop and need to ask for. So the right sensing, connect, act, all of that, what I earlier said, the right software and harmonized platforms and how do we work that in one industry ecosystem? Because I have to say I'd rather push for being 50% owner of a cake than 100% of a cookie. And that is a bit what, what is changing, we cannot do it alone and whether we are the ones who make the most commercial success out of it. I mean again, that is the jury that's out. But in principle, the nations, the subsidy programs, the politicians, customer base Students, analysts and of course everyone in society. If everyone would understand what I'm seeing and would, would, would follow that train of thought. Yeah, then, then my job is, is super easy because I mean, what is the job of a cto? Yeah. Invite people to play with NXP and create out of thought leadership. Create thought followership. Yeah.
- [55:39]
Imogen
Well, I think that is absolutely the most perfect end to note on because we can all agree that upgrading to a cake for from a cookie is, is a world that we want to inhabit.
- [55:48]
Lars Ragers
So I think I, I need it. I, I want to retire in 10, 15 years later and then I need all these assistant systems to help me when I get older.
- [55:57]
Imogen
Well, honestly Lars, thank you so much. We're going to have to continue this conversation another time because there is so much more to cover, but thank you so much.
- [56:06]
Lars Ragers
My pleasure. Looking forward to the next talk.
- [56:07]
Imogen
Well, that is all that we have time for today. Thank you so much to Lars for joining us for this really, really fascinating conversation. And certainly I like the sound of owning 50% of a cake rather than 50% of a cookie. Or perhaps in other words, if we can have that really holistic approach, we can create much more scope to be super, super efficient. Now, before you go, if you could do me the honour of liking subscribing, sharing with a friend, doing all of the above. Honestly, I cannot tell you how valued it is. It really does ensure that we can keep on speaking to fascinating people like Lars and sharing the important and interesting stuff in this clean energy transition. But that's it. Thank you to Louis from our team who will be editing this particular episode and if you have been, thank you for listening.
- [56:53]
Duracell Energy
We're really excited to partner with Duracell Energy to showcase their amazing renewable energy solutions. If you want to reduce your energy bills and join the renewable energy transition, installing Hunter home battery storage and solar panels at home is a great way to start. Duracell Energy's ecosystem of products typically partners with solar panels, but they can be just as effective without it, particularly for electric vehicle owners or anyone looking to take control of their energy. And with Duracell Energy's platinum homeowner offer, viewers can get a custom service that pairs you with top quality products and.
- [57:29]
Imogen
The best installers in your area.
- [57:31]
Duracell Energy
Duracell Energy's batteries, inverters and EV chargers work together on one easy to use app. With features like dynamic tariff integration and grid services, you'll be able to maximize your return. Ready to get started? You can get your quote today. And don't forget, we're also giving away a Duracell Energy Bunny in every episode. Just answer the question about fully charged by following the link in the description. Good luck.