Best of LinkedIn: Next-Gen Vehicle Intelligence CW 23/ 24

00:00:00: Brought to you by Thomas Allgaier and Frennus, this edition highlights key LinkedIn posts on next-gen vehicle intelligence in weeks twenty three and twenty four.

00:00:08: Frenness supports tier one automotive suppliers with early stage market validation for their R&D efforts By combining secondary research direct OEM expert interviews And facilitated customer meetings.

00:00:20: You can find more info the description

00:00:22: Welcome to the deep dive, everyone.

00:00:23: So imagine your vehicle having to actually pull out its own digital wallet to you know negotiate a parking meter with a concrete garage

00:00:32: Right all machine-to-machine

00:00:34: Exactly zero human input and it sounds like I mean pure science fiction right?

00:00:39: Oh completely.

00:00:40: But as we found digging through the absolute sharpest mobility insights on LinkedIn from weeks twenty three And twenty four The AI defined vehicle is well It's already here.

00:00:49: yeah It is.

00:00:51: And it's actively breaking every legacy playbook the industry relies on right now.

00:00:55: So if you're listening to this, You are embedded in the mobility sector.

00:00:58: Maybe your at an OEM maybe a tier one supplier or um...maybe Your actually engineering The software that dictates This whole transition.

00:01:06: Our mission today Is really just To cut through the fluff.

00:01:09: Skip the marketing press releases and look At the brutal fascinating reality of what top Engineering & strategy minds Are actually dealing with

00:01:18: Looking at the discourse online.

00:01:20: The baseline assumption has completely shifted.

00:01:22: Yeah, we aren't just talking about iterative updates to like infotainment screens anymore saying

00:01:27: goodness

00:01:28: right?

00:01:28: I mean We're talking a tectonic reorganization of what a vehicle actually is Who builds it and how it survives in the real world.

00:01:36: So we basically synthesized the conversations from the last two weeks into a few core realities that are defining this next generation of vehicle intelligence.

00:01:44: Which brings us to the first big shift,

00:01:46: right?

00:01:46: Right?

00:01:47: so you look at companies like XPeng and Tesla And they are pushing this incredibly aggressive cloud heavy AI vision.

00:01:57: Oh, massively!

00:01:58: But if I'm say an engineer at a legacy automaker listening right now... ...I might literally be rolling my eyes because

00:02:03: you're stuck on the basics.

00:02:04: Well

00:02:05: exactly Because my department is still just trying to get basic over-the-air updates To.

00:02:10: You know not completely brick The battery management system Right yet?

00:02:14: The thought leaders are already telling us that the goalpost has completely moved like the software defined vehicle the SDV which Has been the holy grail.

00:02:22: for what the last Five years?

00:02:25: Easily five years, yeah.

00:02:26: It's

00:02:26: apparently old news.

00:02:27: now

00:02:27: it is old news and That Is a really painful reality to accept if your organization is still spending billions just To reach that sdv baseline.

00:02:37: I bet

00:02:39: Chauncey made this point incredibly bluntly.

00:02:41: He noted that the sdV is just The foundation like it's essentially Just making the car programmable.

00:02:46: okay The real race, the one that determines survival in this next decade is who reaches AI-defined vehicles AD first.

00:02:54: Wait hold on I have to push back a little bit.

00:02:56: go

00:02:56: for it.

00:02:57: Are we seriously saying the SUV transition?

00:02:59: Is just like groundwork That's already laid but feels Like We're jumping the gun.

00:03:03: because making A multi ton kinetic machine fully programmable and centrally computed i mean that is a monumental engineering feat On its own.

00:03:13: why diminish It To Just a stepping stone?

00:03:16: Because At the end of day, market doesn't reward infrastructure.

00:03:20: It rewards experience.

00:03:23: Devorora actually expanded on this brilliantly.

00:03:25: he pointed out that SDVs laid the necessary groundwork.

00:03:29: so centralized compute OTA architecture but ABE's take raw infrastructure and add predictive adaptive driving layers right top it.

00:03:38: So SDV is just a nervous system.

00:03:40: ADD is actual brain.

00:03:42: Wow, okay.

00:03:42: so think of it like a smartphone revolution then.

00:03:44: The SDV is just building the hardware and the base operating system but the ADV is finally launching the killer apps that make the device feel magic

00:03:52: Exactly!

00:03:53: And Reinhard Seidel actually gave what I consider to be the definitive benchmark for this whole era.

00:03:57: What did he call?

00:03:58: He called AI-defined vehicles Physical AI on Wheels.

00:04:02: Oh wow physical AI on wheels That completely changes the framing doesn't from mechanical product to essentially robotics products Totally...and

00:04:09: crucially.

00:04:10: Cytal added that the user experience, the UX is how customers actually feel that value.

00:04:16: Makes sense!

00:04:16: If you build a brilliant SDV architecture but the car doesn't learn your driving habits or anticipate traffic patterns based on live fleet data... Or

00:04:26: adapt it suspension dynamically using computer vision?

00:04:29: Right exact.

00:04:30: if it doesn't do those things The end consumer does not care about your elegant code.

00:04:35: they just see A car thats less intelligent than the phone sitting in their pocket.

00:04:38: which brings us to the obvious question, who is actually delivering this physical AI on wheels today?

00:04:46: So Benjamin Levith.

00:04:47: looked at the sequential race moving from traditional electrical architecture to SDV, to ADV.

00:04:53: And he noted that Tesla and X-Pen are currently leading the pack by a pretty wide margin.

00:04:58: Oh absolutely they have massive architectural head start.

00:05:00: But there's really interesting geopolitical mechanism underneath this.

00:05:04: we have to acknowledge.

00:05:05: you brought up X-pen.

00:05:06: That ties directly into major strategic move Steven Ma highlighted

00:05:10: Right when these son.

00:05:11: Yeah He noted that Nissan is explicitly making China a global innovation and export hub for their ADV vision.

00:05:19: And we really have to look at the mechanism of why Nissan has doing that.

00:05:22: I mean, it isn't just about cheap manufacturing anymore Not at all.

00:05:25: It's about speed in ecosystem.

00:05:27: Yeah If you want build an AI defined vehicle You need deeply embedded in localized supply chain Where regulatory environment and software talent Honestly allow for ruthless rapid iteration.

00:05:41: Right.

00:05:42: Legacy markets simply have too much baggage and like sequential testing to keep up with the AI iteration cycles happening in regions like China right now.

00:05:51: Yeah, The clock speed is just totally different.

00:05:53: Okay so the vision is clear.

00:05:54: cloud native predictive machines iterating at breakneck speeds

00:05:57: Sounds great on paper.

00:05:58: It sounds limitless when we talk about software models living in the Cloud.

00:06:02: But if I am a hardware engineer i'm literally screaming At my dashboard right now.

00:06:06: Oh for sure.

00:06:07: Because We cannot ignore the physical reality of the vehicle.

00:06:11: You can't just shove a massive data center inside, a metal box park it in The Arizona Sun and expect to run

00:06:17: perfectly.".

00:06:19: And that is the ultimate bottleneck!

00:06:21: It's not lack vision from software teams but the physical constraints for real world pushing back incredibly hard against this vision.

00:06:29: Which brings up quote from H-Manguino which has been stuck on my head all week...it was an ultimate check for this entire industry.

00:06:37: What

00:06:37: did he say?

00:06:38: He wrote Thermal limits, not silicon are throttling next-gen automotive AI.

00:06:44: You cannot OTA cooling.

00:06:46: Well if that's it for a second Cannot OTA Cooling.

00:06:48: It's

00:06:49: so good.

00:06:50: We're just so obsessed with over the air updates fixing every single engineering oversight.

00:06:54: but Thermodynamics.

00:06:56: do not care about your software patch.

00:06:58: Right, if a chip melts down trying to run some predictive computer vision model at eighty miles per hour no line of code is gonna cool it off.

00:07:05: So I really want to dig into the physics of this.

00:07:07: We talk constantly about processing power The newest chips Terra operations per second.

00:07:11: but how?

00:07:12: How exactly does edge compute heat generation actively ruin these utopian ADV plans?

00:07:19: Well,

00:07:19: think about the physical demands.

00:07:21: Yeah To run physical AI on wheels locally On the Edge because let's face it you can't rely on cloud latency for critical driving decisions.

00:07:30: You need immense compute You do?

00:07:32: And these AI models draw massive power.

00:07:35: Massive power generates massive heat, right if you don't engineer liquid cooling or advanced thermal management from literally day one your chip gets too hot.

00:07:44: and what happens when a chip gets to high?

00:07:46: it throttles its performance to save itself.

00:07:48: so Right When Your Vehicle Is Navigating A Complex High Temperature Environment It Needs Maximum Intelligence.

00:07:54: The Thermal Limits Throttle the Processor Yeah...and

00:07:57: your highly touted ai-defined vehicle suddenly loses its brain power in real time.

00:08:02: Wow,

00:08:03: so the hardware architecture has to fundamentally change just to support the heat of the software?

00:08:08: Yep!

00:08:08: But you've argued before that as hard as the physics are...

00:08:12: The human

00:08:13: and organizational limits are actually worse.

00:08:15: Oh they absolutely are because thermal limits are at the end-of-the day an engineering physics problem.

00:08:21: We know how to solve those with enough Time & Money.

00:08:23: Human Limits Are a structural Problem.

00:08:26: Ana Ruta-Dorlo issued this really

00:08:29: stark

00:08:29: warning to the industry that we saw echoing across multiple posts last week.

00:08:33: Yeah, he said you cannot build a software defined vehicle with a hardware-defined organization.

00:08:41: Man, if you are software engineer at tier one listening right now You know exactly what he means.

00:08:46: Your code is directly mirroring your broken organizational chart.

00:08:50: Totally This basically Conway's law in action Which Sushama M applied perfectly to Automotive.

00:08:56: Remind me how she framed it.

00:08:57: So we all know Conways Law, right?

00:08:59: Yeah The idea that systems mirror the communication structures of organizations That build them Right.

00:09:04: Well, Sushama pointed out that legacy SDB architecture perfectly reflects a siloed engineering org.

00:09:10: Like if your powertrain team sits on different floor and fundamentally distrusts the infotainment team Your vehicle software will be permanently silo'd

00:09:18: And you end up with like eighty separate electronic control units doing completely redundant tasks.

00:09:24: Exactly!

00:09:24: Simply because humans managing those domains refuse to collaborate in single centralized compute node.

00:09:31: It's wild But the friction actually goes deeper than just office politics.

00:09:36: Nevzad Eshaniala concluded that The industry is not actually facing a software problem at all

00:09:41: Really?

00:09:42: What is it then?

00:09:43: He argues, It's facing an architecture contract problem.

00:09:46: Oh I love this point.

00:09:47: We know the legacy nightmare of black box tier one procurement.

00:09:51: You buy breaking system and its completely legally locked down.

00:09:54: Exactly!

00:09:55: Its like trying to hire a symphony orchestra but your legal department forces.

00:09:59: you write separate iron-clad contracts with every single violinist and cellist.

00:10:04: Forbidding them from looking at each other's sheet music Or

00:10:06: taking cues form the conductor?

00:10:08: You get noise, not music.

00:10:10: In an AI defined vehicle The suspension needs to talk to breaks based on a camera feed processing a pothole in milliseconds.

00:10:19: If your legacy procurement process legally forbids that deep software integration because of intellectual property silos, the contract is literally destroying the vehicle's architecture.

00:10:29: Wow!

00:10:30: The way you write a legal agreement directly limits intelligence in the car.

00:10:34: That is a staggering blind spot.

00:10:36: I mean we are trying to build an ADV but... The lawyers are still buying parts for a twenty-ten sedan.

00:10:41: Exactly!

00:10:42: Okay, so if physical cooling is hitting the ceiling and rigid old school org charts or legal contracts are legally forbidding integration how does any legacy automaker actually speed up development to survive this transition?

00:10:56: It's tough... Because of China is iterating as fast.

00:10:59: we discussed earlier that clock seriously

00:11:01: ticking.

00:11:01: Well the only way out of that trap virtualization completely abandoning the whole.

00:11:06: do it all yourself model.

00:11:07: Hannah Wolfe laid out the core mandate for entire industry moving forward.

00:11:11: Which

00:11:11: was?

00:11:11: He said build virtual, test virtual iterate fast

00:11:14: And data points backing up that mandate are just wild!

00:11:18: Steve Greenfield pointed out that Nissan is having their vehicle development time down to just thirty months using AI across design and testing.

00:11:26: Having Development Time.

00:11:27: Yeah

00:11:28: we need look at mechanism there.

00:11:29: How does Virtual Testing actually shave literally years off a timeline

00:11:34: What fundamentally changes sequence of engineering?

00:11:36: Yeah.

00:11:36: You know, you aren't building a physical clay model putting it in the physical wind tunnel realizing that side mirror placement causes drag or blocks of LiDAR sensor

00:11:46: and then going back to drawing board for six months?

00:11:48: Exactly!

00:11:49: You simulate air flow thermal dynamics and sensor field view simultaneously on digital twin.

00:11:55: Find failure points before single piece metal is even stamped.

00:11:59: An UM-Ukpam brought up a statistic that really grounds this reality.

00:12:03: He noted, the AWS Bosch Workbench is now validating eighty percent of mobility software

00:12:08: virtually.

00:12:09: Eighty percent?

00:12:10: Think about implications for a software engineer!

00:12:13: You are no longer waiting six months to be built just so you can flash your code onto it and see if breaks work with algorithm... Right.

00:12:20: ...you're testing thousands times overnight in cloud.

00:12:24: But here's catch… You cannot achieve that level of virtual validation in a vacuum.

00:12:30: That's true.

00:12:31: Erkendies stated very clearly, that vehicle testing is simply too complex to insource fully.

00:12:36: now The complexity of simulating edge cases for an AI-defined vehicle Is just beyond the scope Of any single legacy OEM.

00:12:45: Yeah...that makes sense.

00:12:46: Dillek Kabak summarized this new reality perfectly.

00:12:49: actually She noted that competitive advantage in the SDV and AI era no longer comes from scale alone.

00:12:56: It comes from holistic integration, strong ecosystem partnerships.

00:13:00: See this shifts entire power dynamic.

00:13:02: it used to be if you were the biggest OEM You won through sheer manufacturing volume.

00:13:06: Now If your a massive OEM but are slow to partner you lose to smaller player who is deeply integrated into right software Ecosystem.

00:13:15: And we're seeing actual infrastructure for new marketplace model being built today.

00:13:20: Like Tonskip mentioned, SDverse...

00:13:22: Oh yeah I saw that!

00:13:24: Yeah which is a new BDB marketplace specifically designed to source and commercialize STV software faster.

00:13:30: That is fascinating.

00:13:31: It's essentially treating automotive software like enterprise software licensing pretty much.

00:13:37: if you need a driver monitoring algorithm, You don't spend two years and fifty million dollars having your internal team build one from scratch just to say you own it?

00:13:45: Yep!

00:13:46: You go the B-to-B marketplace license approved an algorithm drop into virtual testing environment validated overnight and deploy.

00:13:53: Exactly, it breaks down the architecture contract problem we talked about earlier.

00:13:57: instead of buying a locked black box hardware component you are licensing this specific software capability and integrating into your centralized brain.

00:14:06: And

00:14:06: that is how have year development time to thirty

00:14:08: months?

00:14:09: You got it

00:14:09: Alright.

00:14:10: so let's project this forward.

00:14:11: We've solved thermal throttling.

00:14:13: We bypassed legal silos using B-to-B software marketplaces.

00:14:17: We're testing virtually and these highly capable AI driven cars finally hit the road at scale.

00:14:25: What happens to the massive traditional business infrastructure that surrounds the automotive industry?

00:14:30: Because the car doesn't exist in a vacuum, it exists in a world built for human drivers.

00:14:34: Man this is where we see the ripple effects completely disrupt adjacent industries.

00:14:39: Yeah.

00:14:40: Thaibol Castania brought up a massive paradigm shift in insurance frankly losing their minds over.

00:14:47: Oh, I bet!

00:14:47: He highlighted that autonomous vehicles and heavily AI-defined vehicles make today's actuarial insurance model completely obsolete.

00:14:56: Well because legacy insurance is entirely based on human demographics.

00:15:00: A sixteen year old male with a sports car is statistically at higher risk than of forty five years old female in minivan.

00:15:06: so he pays more

00:15:07: Precisely.

00:15:08: And an AI doesn't have a demographic.

00:15:09: An AI does not get tired It doesn't get distracted by a text message, and it doesn't have gender or an age.

00:15:16: Its risk profile is determined by its software version the health of his lidar sensors And complexity in the operational design domain.

00:15:24: that's navigating.

00:15:26: So Cassania points out underwriting must move to live telemetry.

00:15:31: You aren't ensuring historical demographic.

00:15:34: you are insuring real time performance.

00:15:37: And if the system pushes like a faulty OTA update that degrades sensor fusion by two percent, The premium dynamically adjusts that exact afternoon.

00:15:46: That completely destroys the legacy insurance business model.

00:15:50: and speaking of destroying legacy models We have to talk about the hook from The Top of the Show.

00:15:55: The parking!

00:15:55: Yes,

00:15:56: Harry Campbell brought up a hyper-specific logistical reality of autonomous vehicles.

00:16:01: that sounds funny until you realize the massive infrastructure challenge it presents.

00:16:06: Autonomous vehicles are going to pay for their own parking.

00:16:08: It highlights a major blind spot in how we think about autonomy...we always focus on driving but cars spend what?

00:16:15: Ninety five percent of our lives parked?

00:16:16: Right If you take an empty Moby Taxi to the airport and it drops off, It can't just circle the terminal infinitely.

00:16:22: It has to go wait somewhere right?

00:16:25: Has to pull into a private garage negotiate a rate and pay the meter.

00:16:29: Campbell notes that machine-machine payments are finally making this possible.

00:16:33: Just API to API.

00:16:35: Exactly The vehicle's digital wallet communicates directly with the parking structures.

00:16:39: API No humans involved.

00:16:42: And we need to ground this in the commercial reality of today because it is very easy to dismiss.

00:16:47: This as just future casting, but this is happening right now.

00:16:50: It really is

00:16:51: Allison.

00:16:51: africano nila noted that Waymoe Just bought a fifty five hundred acre proving ground In Arizona.

00:16:56: Oh yeah I heard about That.

00:16:58: and what Is incredible?

00:16:59: But that move is that this The exact same physical facility where Apple's highly secretive car project notoriously died?

00:17:06: No way.

00:17:07: Yeah, Waymo was taking the graveyard of tech's biggest failed automotive ambition and using it as a physical real estate to scale actual commercial autonomy.

00:17:16: That is poetic!

00:17:17: And you know... It isn't just passenger robotexes either.

00:17:21: On heavy-duty commercial side Shibum Shrivastava shared that their gigafusionnet foundation model is powering twenty eight driverless Kodiak trucks hauling commercial freight right now.

00:17:32: Wow!

00:17:32: Twenty-eight massive trucks moving real economic goods on public highways, not in a closed simulation...not promised for next quarter

00:17:40: today.".

00:17:41: So the chasm between R&D and commercial employment has been crossed?

00:17:45: Clearly We are no longer waiting for AI defined vehicle.

00:17:48: we're actively trying to regulate it while its already on road.

00:17:52: As we wrap up this deep dive I want leave you with one final thought to mull over because it fundamentally reframes everything.

00:17:58: Let's hear it.

00:17:59: As we shift toward these AI-defined level two and level three cars, the ones that are physical AI on wheels?

00:18:05: The entire psychological paradigm of what it means to be a driver changes?

00:18:09: Prize Thomas Matthews framed this perfectly.

00:18:11: What did he say?

00:18:12: He said...

00:18:18: Man!

00:18:19: That is a profound shift for anyone who loves cars.

00:18:22: For a century The vehicle has been a mechanical tool that you actively dominate and control.

00:18:29: Now, your no longer operating the machine You're teaming up with an intelligent agent.

00:18:34: The AI is supervising complex environment And you are supervising the AIs high-level decision making.

00:18:40: It's a partnership... Yeah!

00:18:41: ...You Are the Conductor and the AI.

00:18:43: Is That Incredibly Capable?

00:18:45: Symphony Orchestra we talked about earlier.

00:18:47: Exactly If you enjoyed this episode new episodes drop every two weeks.

00:18:52: Also check out our other editions on electrification and battery technology, Future Mobility & Market Evolution, and Commercial Fleet Insights.

00:19:00: Thanks for listening!

00:19:01: Thank you so much for joining us in giving your time today.

00:19:04: Make sure to hit subscribe on the feed So that we don't miss the next one.

00:19:07: Keep innovating And see ya next time.