Best of LinkedIn: Next-Gen Vehicle Intelligence CW 25/ 26

00:00:00: Brought to you by Thomas Allgeier and Frennis.

00:00:02: This edition highlights key LinkedIn posts on NextGen vehicle intelligence in weeks twenty-five, and twenty six.

00:00:08: Frenis 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.

00:00:19: You can find more info In the description.

00:00:21: Imagine spending four billion dollars Just a fix old computer code Like literally four billion while your biggest competitor spends exactly zero.

00:00:31: Wow,

00:00:31: yeah!

00:00:32: Zero.

00:00:34: Welcome to our latest deep dive.

00:00:35: today we are looking at why legacy automakers Are just bleeding cash right now.

00:00:41: and well Why that whole idea of the smartphone on wheels is actually already an outdated concept?

00:00:45: Right

00:00:45: it's completely outdated.

00:00:46: so if you're listening to this We're basically unpacking a massive wave of insights that we've curated directly from the mobility industry's leading professionals on LinkedIn.

00:00:55: This is all from calendar weeks, twenty-five and twenty six.

00:00:57: Yeah!

00:00:58: And if there was one overarching narrative you really need to take away from this deep dive it' s that The Industry Is Undergoing A Violent Strategic Shift.

00:01:06: I mean...we are moving away From the Software Defined Vehicle The SDV And we Are Rapidly Entering The Era Of The AI-Defined Vehicles Or ADG.

00:01:16: But the twist here is that the real battleground isn't actually artificial intelligence itself.

00:01:21: It's the plumbing underneath, it's you know underlying architecture and speed of execution across these massive supply chains.

00:01:28: And this terrifying new challenge how you mathematically validate These highly complex systems before they go on a public road.

00:01:36: Exactly so.

00:01:36: let start with strategic shift because I mean for solid decade now The whole industry has practically been living breathing the term software defined vehicle.

00:01:44: Yeah STV was everything.

00:01:46: Right,

00:01:46: and to give some context on how we got here there was this brilliant breakdown circulating from Aladyn Hamdi.

00:01:52: he laid out the evolution perfectly.

00:01:54: first The industry had a build-a foundation right that was autosar

00:01:57: which basically standardized the messy background code.

00:01:59: Yeah

00:02:00: exactly so different parts of a car could just speak a common language.

00:02:04: And then came the SDV Which gave automakers flexibility.

00:02:08: it fundamentally decoupled the physical hardware from the software, meaning you could update your car's infotainment over-the-air without changing a physical chip.

00:02:18: And that de-coupling was revolutionary at the time but I mean it was still just software following strict prewritten rules?

00:02:24: Yeah What Alladin points out is that the AED, The AI Defined Vehicle brings active intelligence.

00:02:31: yeah and we have to be really precise about what that intelligence actually does.

00:02:34: Totally

00:02:35: Following That Thread I'm Rick Gowenke and Kim Darling-Sarrison added some crucial context.

00:02:41: They noted that the ADV isn't just about using algorithms to control the steering or the brakes, it fundamentally shapes the experience of being inside the machine.

00:02:51: AI is becoming a core architectural layer of car itself.

00:02:54: It's not just flashy voice assistant bolted onto the digital cockpit, you know.

00:02:58: It's

00:02:58: the central nervous system.

00:03:00: I have to push back a little here though because If you've been around the mobility space long enough this sounds incredibly familiar.

00:03:07: Oh

00:03:07: sure

00:03:08: Isn't it just the latest round of industry marketing hype like we spent years being promised?

00:03:12: This magical smartphone on wheels?

00:03:15: and now suddenly The buzzword has shifted to an active personal assistant on wheels.

00:03:19: But i mean look at legacy oems.

00:03:22: are they actually ready to build something that integrated?

00:03:24: well That is the exact tension causing sleepless nights in Detroit, Stuttgart and Tokyo right now.

00:03:31: Adi Akronisto Aguilar addressed this beautifully in his takeaways from the ECV Auto Events Summit.

00:03:37: Oh

00:03:37: yeah I saw that.

00:03:38: Yeah he made it clear that AI is shifting for being just a consumer feature to a defining property of vehicle architecture.

00:03:45: The challenge these traditional automakers isn't whether they have access to good AI models.

00:03:50: Right, they can just buy those.

00:03:51: Exactly The challenge is how to scale that intelligence safely and deterministically across the entire vehicle.

00:03:58: And that reality naturally exposes why current legacy architectures are actively failing.

00:04:03: Which means if you successfully build a genius level AI brain it doesn't matter If car's nervous system is fragmented.

00:04:11: mess right?

00:04:12: The AI literally cant do its job.

00:04:14: Precisely The dream of the AI-defined vehicle crashes incredibly hard into the reality of legacy automotive architecture.

00:04:24: Kat Gan made a phenomenal observation about this contradiction.

00:04:27: What does she say?

00:04:28: Well, she pointed out that Western OEMs have access to exact same frontier AI models as anyone else in the world.

00:04:34: They can license the best tech out there But what happens when they actually deploy them?

00:04:39: They end up caging those advanced models inside, which she calls chatterbox infotainment systems.

00:04:44: Chatterbox!

00:04:45: Wow that's a great word for it.

00:04:46: Right you get really smart screen in the dashboard but has no deep control over car itself

00:04:52: Because underlying hardware won't allow

00:04:53: it.

00:04:54: Exactly legacy cars rely on Dozens, sometimes like nearly a hundred distorted electronic control units or ECUs.

00:05:01: These are basically tiny isolated computers scattered throughout the car

00:05:04: So that you see you controlling?

00:05:05: The power windows has no idea what the ecu controlling this steering wheel is doing.

00:05:10: yeah

00:05:10: They are physically and digitally siloed.

00:05:12: Yeah.

00:05:13: Meanwhile if you look at companies like Tesla and the rising Chinese OEMs they are treating the entire vehicle as a unified robot with a single unified perception layer.

00:05:23: It's like having a genius brain hooked up to body where the left hand can't communicate with right foot?

00:05:28: Yeah, exactly!

00:05:29: And financial cost of that disjointed approach is staggering….

00:05:33: Going back to this four billion dollar statistic from top at show, Dr.

00:05:38: Pradhan shared an example completely blew my mind...

00:05:41: Right in

00:05:42: twenty-twenty-four and major legacy OEM had spend four billion dollars just fix software they originally wrote.

00:05:50: Over that same period, for the specific type of foundational software remediation, Tesla spent zero dollars.

00:05:57: Brunhahn argues this isn't just a technical glitch—this is a profound identity shift!

00:06:02: Legacy brands are still acting like hardware companies who happen to have a software team bolted on as an afterthought….

00:06:08: It's a complete identity crisis and honestly it has driven heavily by the finance department not only the engineering department.

00:06:16: Thomas M had brilliant critique in his exact dynamic

00:06:20: About the hardware decisions.

00:06:21: Yeah, he argued that The real killer of the software-defined vehicle in the West wasn't bad Software.

00:06:26: it was a Hardware decision heavily disguised as A finance Decision.

00:06:31: Historically Western OEMs decided to gate Their hardware by trim levels To protect their profit margins.

00:06:37: wait so you mean By only putting the premium sensors and the high performance computing power In the expensive luxury packages.

00:06:44: exactly If you only put the best cameras and radar on the eighty thousand dollar luxury SUV, And leave them off of a thirty-thousand dollars base model You protect your margin from the luxury car.

00:06:54: Right!

00:06:55: But by doing that completely kill your fleet.

00:06:57: wide data flywheel.

00:06:58: To train these advanced AI models we need massive volume.

00:07:02: We need data for every type road, weather or driver.

00:07:06: By getting hardware Western OEMs actively starved their own AI The data it needed to learn.

00:07:11: Wow Meanwhile, while they were protecting margins companies like BYD and Tesla were shipping their full-sensing stack to almost every single car rolling off the assembly line.

00:07:22: Yes

00:07:23: They effectively turned millions of their everyday drivers into an unpaid data flywheel just constantly feeding raw real world data back to their central AI.

00:07:34: And as Thomas M pointed out You cannot just license a data flywheel that you never took the time to build yourself.

00:07:40: That makes so much sense!

00:07:41: Let me play Devil's Advocate for second though, because re-architecting an entire vehicle lineup to have centralized brain is incredibly expensive.

00:07:48: Oh absolutely.

00:07:49: Anirudha Dorle or Ani jumped into this conversation with really pragmatic counterpoint.

00:07:54: He argued that moving away from those tiny, isolated ECUs to a highly centralized zonal architecture where computation is grouped by the physical zones of the car rather than individual functions.

00:08:05: well.

00:08:06: It only makes sense if economics actually justify

00:08:08: it.

00:08:09: Right!

00:08:09: The economics have work.

00:08:10: You can't just force an ultra-expensive high performance computing chip into basic entry level hatchback Where A three dollar microcontroller does job.

00:08:19: rolling down window perfectly fine.

00:08:23: The variant management and the assembly economics all have to align.

00:08:27: That is a very fair point, And it really highlights the friction between the theoretical ideal of an AI-driven car... ...and the gritty industrial reality of mass manufacturing.

00:08:37: But here's the catch Whether you are building a hundred thousand dollar premium robot taxi Or smart thirty thousand dollar hatchback.

00:08:45: When things go wrong in these new interconnected architectures Diagnosing them as total nightmare

00:08:50: Oh I bet.

00:08:51: Provinter Singh pointed out that we are entering an era of what he calls vehicle triage.

00:08:56: Vehicle triage!

00:08:57: So if you're an engineer trying to debug a problem, You no longer just looking for broken physical wire?

00:09:01: No...you were hunting down a ghost in massive digital network.

00:09:06: When issues spans the cloud The internal connectivity gateways, the vehicle networks and the physical ECUs all at once.

00:09:13: The new superpower for automotive engineers isn't fixing a localized bug.

00:09:17: It's the ability to trace end-to-end data flows across an incredibly complex distributed system.

00:09:24: So if you successfully centralize the vehicle's architecture and give it this single, incredibly complex AI brain... You suddenly create a terrifying new problem!

00:09:34: Yeah—a single point of failure.

00:09:36: Right, if that single brain fails or gets confused the whole car goes down.

00:09:41: So how do you mathematically prove a system like that is actually safe to put on the highway at seventy miles an hour?

00:09:46: That's the existential question right now because old methods of validating safety simply don't scale to artificial intelligence.

00:09:54: Thanethan Sozane introduced a concept into the discussion that he calls, The Validation Paradox.

00:10:01: Yeah for decades the industry's approach to safety was mileage accumulation.

00:10:05: you just drive the car from millions of miles on test tracks and highways but with AI in autonomous systems...the number possible scenarios the vehicle might encounter combinations weird weather unpredictable traffic strange human behavior is functionally infinite.

00:10:20: You literally cannot drive enough miles to prove it safe.

00:10:23: Exactly, so the entire industry is shifting away from mileage accumulation towards scenario-based testing.

00:10:30: They are hunting for critical edge cases that actually matter most.

00:10:34: It's like trying to train a new teenage driver by making them drive on completely empty straight highway for ten years.

00:10:41: Sure they log a lot of miles, but they learn absolutely nothing about how to handle a real emergency.

00:10:47: Right?

00:10:47: They'd panic the first time someone slams at the brakes!

00:10:50: Exactly.

00:10:51: Scenario-based testing is like putting that teenager in simulator where a deer jumps out on an icy road.

00:10:57: every five minutes You are identifying and practicing the exact trick questions which will appear during the exam.

00:11:04: That's much better way to think about it.

00:11:07: you have focus entirely onto those edge cases But the complexity doesn't stop there.

00:11:12: Even when you find those edge cases and train AI to recognize them, The way vehicle's internal network response has to be flawless on a physical level.

00:11:20: Fuddy LeBee made a fascinating comparison between modern cloud computing systems and autonomous automotive systems.

00:11:27: Well, on paper they actually use very similar digital communication patterns things like request-and-response or publish-and subscribe architectures.

00:11:36: But the stakes are completely different

00:11:38: Entirely different.

00:11:39: in a cloud architecture if a server drops a message Or there's a slight network delay you might get a five hundred error.

00:11:45: On a web page where your streaming video buffers for a second You just refresh the page.

00:11:49: no big deal

00:11:50: Right.

00:11:51: In a car, you are dealing with hard real-time deadlines.

00:11:56: If a published command for the automatic emergency braking system drops a packet or arrives fifty milliseconds too late You have life threatening safety hazards.

00:12:04: Wow!

00:12:05: Fifty milliseconds?

00:12:06: Yeah...you what engineers call worst case execution times.

00:12:09: that must be mathematically guaranteed.

00:12:11: You cannot simply coddy and pace an Amazon Web Services architecture into vehicles' breaking systems because the mathematical models of safety don't transfer.

00:12:19: And speaking of the mathematics of safety, we have to talk about how industry actually classifies that risk.

00:12:24: Because it's not as simple asking did the camera break?

00:12:27: Imran Khan posted a really sharp take on the ISO-AISO-RF that stands for Automotive Safety Integrity Level.

00:12:39: He stressed that ASAL is not just a hardware label you print on the box and slap onto to component, it's vehicle-level risk judgment.

00:12:47: This is crucial distinction—a sensor failing isn't actually the hazard!

00:12:51: The hazard is an unintended vehicle motion because of this failure in highly specific driving scenarios.

00:12:58: It's about how the entire centralized system reacts to the absence of that sensor

00:13:03: data.".

00:13:03: And that vehicle level risk doesn't apply at the pristine factory floor?

00:13:07: It extends all the way into the messy reality of aftermarket repair shops.

00:13:11: Oh yeah, The Aftermarket is a whole other beast!

00:13:13: Frank Turlip shared some staggering reality checks from the collision and repair industry that really ground this high level engineering to everyday consumer-reality.

00:13:23: He noted that ADIAS Calibration's Advanced Driver Assistance Systems calibrations appeared on nearly thirty five percent of All Repair Estimates in twenty twenty five.

00:13:32: Wow Over a

00:13:33: third.

00:13:33: Yeah, and that is adding an average of six hundred eighty eight dollars to the repair bill.

00:13:38: It's a massive hidden cost of this intelligence.

00:13:41: But Frank highlights are much deeper critical insight.

00:13:44: The gap between a shop doing a calibration And actually validating it was done correctly Is becoming A massive liability issue

00:13:51: Because its so complex.

00:13:52: now

00:13:53: Exactly Subletting This Calibration Work To Third Parties Is Incredibly Common Right Now because Of The Vast Array Of Specialized Sensors Involved And that makes accountability very fuzzy.

00:14:03: A calibration on an AI-defined vehicle is no longer just a checkbox or invoice, you know?

00:14:08: It's critical safety validating procedure.

00:14:11: if independent shop gets it wrong the car centralized brain operates bad data.

00:14:15: So step back and look at whole picture.

00:14:17: we've got cars transforming into highly centralized AI driven robots.

00:14:21: We have massive internal architecture overhauls causing identity crises at legacy automakers, and we have these mind-bending safety validation challenges.

00:14:31: That's a lot!

00:14:31: It is.

00:14:32: So how was all of this actually playing out competitively on the global stage?

00:14:37: Because whoever solves this first owns the future mobility.

00:14:41: If you look what people like Scott Newton are tracking there is a titanic geographic shift happening.

00:14:47: He recently pointed out that Chinese automakers, specifically brands like BYD and MG are rapidly overtaking traditional German OEMs right in their own core European markets.

00:14:56: It is a fundamental rewiring of the industry landscape.

00:14:59: Yeah And it all comes back to how you manage that complexity.

00:15:01: we just talked about UMOP provided brilliant detailed breakdown.

00:15:05: exactly how BYD is achieving this dominance

00:15:07: The vertical integration Right

00:15:09: Exactly.

00:15:09: Roofless Vertical Integration.

00:15:11: BYD has over three point one million intelligent vehicles currently on road generating two hundred million kilometers of driving data every single day.

00:15:21: But more importantly, they aren't just assembling parts bought from a hundred different suppliers—they control the battery chemistry… They manufacture power semiconductors... They write software….

00:15:30: They train AI algorithms and own the dataloop!

00:15:35: Which solves the exact problem.

00:15:37: Kat Gann pointed out earlier about legacy OEMs having dozens of disjointed ECUs built by different suppliers that refuse to talk to each other.

00:15:45: If you own a whole stack, You can force this system to be unified

00:15:47: Precisely.

00:15:48: and it's not just technology moving fast.

00:15:51: The regulatory landscape is surprisingly keeping pace which acts as massive enabler for global race.

00:15:58: Ian Richards and Raoul LaTour Fortes posted some great insights on UNWP.

00:16:04: The United Nations essentially just agree on the world's first unified global regulatory framework for homologating fully driverless vehicles.

00:16:11: So,

00:16:11: the legal roadblocks are actually being cleared at the international level.

00:16:15: and you see it Paul Comfort noted that in the US, regulators are actively proposing to drop the long-standing safety standards that require manual steering wheels and brake pedals entirely for autonomous vehicles.

00:16:27: The regulatory momentum is definitely there but we do need to temper some of this futuristic excitement with a dose of industrial reality because Speaking at the Infineon Automotive Ecosystem Summit provided a very sobering reality check.

00:16:50: Despite all the hype and billions of dollars surrounding STVs in AI, she noted that ninety percent new vehicles sold in twenty-twenty six will still not be true software defined vehicle.

00:16:59: Wait, ninety percent?

00:17:00: So the vast majority cars hitting dealership lots this year are running on those legacy fragmented architectures

00:17:08: Exactly.

00:17:09: And Joachim Langenwalter's takeaway from the Automobile Electronic Congress AEK-E, really drives home why that gap exists.

00:17:18: He observed that technology itself is no longer the bottleneck in the automotive industry.

00:17:23: everyone has access to high performance computing chips.

00:17:26: Everyone has access frontier AI code.

00:17:29: So what holding them back?

00:17:30: The real bottleneck is execution and ecosystem collaboration.

00:17:34: Scaling these next-generation AI vehicles requires an orchestrated ecosystem of silicon, software tooling and cloud infrastructure working together seamlessly.

00:17:45: The days a single automaker doing everything in isolation or just acting as passive systems integrator are over

00:17:51: Which perfectly brings us to really provocative thought because this evolution from disjointed hardware to centralized AI brain fundamentally changes your relationship with the car itself.

00:18:01: There's a post from John Schneibel that really got me thinking about the ongoing right-to-repair debate.

00:18:05: Oh, well

00:18:06: That's a crucial angle.

00:18:07: The technical architecture always dictates the legal reality.

00:18:10: Right Schneibel pointed out that restrictions on independent repair shops have been around since the nineteen nineties.

00:18:15: This was way before anyone in the industry was actively worried about digital hackers or cybersecurity, it's historically always been about controlling ownership and access to proprietary information.

00:18:25: sure but as we move into this new era where vehicles are highly centralized AI defined robots things get infinitely more complicated.

00:18:35: How will the very definition of car ownership fundamentally change when your local independent mechanic has to interface with an OEM's proprietary cloud connected central intelligence layer?

00:18:44: just a, you know replace physical brake actuator or calibrate bumper sensor?

00:18:50: It is profound question to leave on.

00:18:52: If the vehicle's intelligence and therefore its ability to even function safely is locked behind a centralized AI architecture owned by the manufacturer, do you really own the car sitting in your driveway?

00:19:03: Or are you essentially just licensing a piece of hardware that OEM controls.

00:19:07: Brought to you by Thomas Allgeier and Frennus, this edition highlights key LinkedIn posts on next-gen vehicle intelligence in weeks twenty five and twenty six.

00:19:15: 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.

00:19:27: You can find more info in the description.

00:19:29: If you enjoyed this episode new episodes drop every two weeks.

00:19:33: Also, check out our other editions on electrification and battery technology future mobility in market evolution.

00:19:39: And commercial fleet insights.

00:19:41: Thank you so much for joining us on this deep dive.

00:19:43: don't forget to subscribe So you never miss an update on where the mobility industry is heading.

00:19:47: next.

00:19:48: until Next time keep your eyes on the road and your architecture centralized.