Best of LinkedIn: Next-Gen Vehicle Intelligence CW 17/ 18

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

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 at facilitated customer meetings.

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

00:00:22: Thanks for that.

00:00:23: So let's just get right into it because we have a lot of ground to cover today.

00:00:27: We really do.

00:00:28: Welcome to this deep dive, everyone!

00:00:29: Today we're unpacking well essentially the most critical conversations happening across the mobility industry right now.

00:00:35: Yeah

00:00:36: specifically looking at the insights that Mobility Professionals have been sharing on LinkedIn over weeks seventeen and eighteen

00:00:43: Exactly.

00:00:44: And our mission for you today is simple-we want a cut through all of the marketing jargon & noise...and bring you reality where next generation vehicle intelligence actually heading

00:00:55: based on the people who are literally building it on the ground.

00:00:57: So to start off, we need to look past what the industry is currently obsessing over like right now.

00:01:04: everybody's still super fixated on the SDV Right?

00:01:07: The software defined vehicle?

00:01:09: Oh totally I mean It has been a holy grail for A decade Yeah

00:01:13: But the smartest voices were tracking.

00:01:15: they're already looking completely beyond that.

00:01:17: They talking about the leap into the AI Defined Vehicle

00:01:21: Which makes sense because the Software Defined vehicle It was never actually the final destination.

00:01:26: Yo,

00:01:27: it wasn't!

00:01:35: Well,

00:01:36: Albert Poulot actually had a brilliant analogy for this.

00:01:38: that really grounds it.

00:01:40: He used

00:01:40: cabin climate control to explain the evolution.

00:01:43: Oh right!

00:01:43: I read that post.

00:01:44: It's great way of visualizing.

00:01:45: Yeah.

00:01:46: So he says you know if look at hardware defined vehicle You're just turning physical dial.

00:01:50: You want it colder.

00:01:51: Turned the dial Opens mechanical vent

00:01:54: Completely manual

00:01:55: Right And then the software-defined vehicle comes along, now you're adjusting a digital screen.

00:02:01: The underlying code tells the compressor and fans what to do... ...and the OEM can send an OTA update later to make it more

00:02:08: efficient.".

00:02:08: But then you get into AI defined vehicles….

00:02:11: …and that whole paradigm just

00:02:13: fractures!

00:02:14: Yeah how does he put it?

00:02:16: Well..the climate control basically becomes autonomous.

00:02:19: It's not waiting for your input.

00:02:21: The car actively learns your historical preferences.

00:02:24: Oh, wow.

00:02:25: It reads contextual data like it checks the outside weather... ...it looks at trajectory of the sun hitting wind shield.

00:02:30: Yeah!

00:02:30: It tracks the Sun

00:02:32: Exactly and even monitors your vitals using in-cabin sensors.

00:02:36: So adjusts environment proactively before you realize that you're getting hot.

00:02:40: That

00:02:40: is wild.

00:02:40: Basically goes from being this dumb tool to an active partner.

00:02:47: But here's catch To make car a active partner You need unbelievable amount of localized computing power.

00:02:53: Right, because you can't run that kind of real-time complex sensory analysis on like legacy microcontrollers.

00:03:01: No!

00:03:01: You can't.

00:03:02: and That brings us to the infrastructure reality that Manikav Asagam Sundaram highlighted.

00:03:08: He was saying we really have to stop thinking of cars as just quote connected devices.

00:03:12: So what are they then?

00:03:13: They're high performance compute nodes on wheels Like literal HPCs.

00:03:18: HPC nodes okay yeah

00:03:19: Because to run an agentic AI backbone, and by that I mean an AI who doesn't just answer questions but actively executes complex tasks across the car.

00:03:29: You need massive processing power!

00:03:31: How massive are we talking?

00:03:33: Well, he specifically referenced Turing chips delivering three thousand TOPS

00:03:36: Okay.

00:03:37: Three thousand TOPs.

00:03:38: and just to clarify for anyone listening who isn't deep into chip architecture, TOPS stands for Terra Operations Per Second Right.

00:03:45: So three-thousand TOPS means the car's brain is processing three thousand trillion operations every single second

00:03:51: which sounds insane, but you need that horsepower to instantly ingest data from the cameras.

00:03:56: The radar or the lidar fuse it all together and make life-or-death driving decisions...

00:04:02: And without waiting for a cloud server to ping back?

00:04:04: Right?!

00:04:05: Exactly!

00:04:05: The heavy lifting has to happen right there at the edge physically inside of car.

00:04:09: But okay let me pushback on this little bit.

00:04:12: If the car goes from being like A smartphone To a proactive super computer How on earth do we validate that?

00:04:20: Well, how do you validate a machine that keeps changing its own behavior after it leaves the factory floor?

00:04:26: Like How Do You Know It's Safe?

00:04:28: Ah.

00:04:29: Yeah well That is arguably The Biggest Unsolved Problem in Automotive Engineering Right Now.

00:04:33: I

00:04:33: would imagine

00:04:34: Satya Pradaprahaan actually sparked a huge discussion on this exact issue.

00:04:38: He pointed out that ISO-to-six to sixty two,

00:04:41: which is the functional safety standard right?

00:04:43: Right

00:04:43: he noted That The entire Standard Is Fundamentally incompatible With AI Defined Vehicles Because ISO-two six Two Six Two Was Written To Validate Fixed Deterministic Behavior.

00:04:54: Oh I see like if X Happens Y As Guaranteed To Result

00:04:57: Exactly.

00:04:58: But Ai Doesn't Do That.

00:05:00: It learns and adapts in real time.

00:05:02: So if it encounters some weird edge case on a snowy road, And like...learns-a bad habit you can't just hope that correct itself?

00:05:08: No!

00:05:09: You can't wait for an over the air patch from a server before something does dangerous Which is why Dr.

00:05:15: Marina Zebian proposed really compelling solution.

00:05:17: Okay what's her fix?

00:05:19: She argues industry needs hybrid architecture Basically four strict guardrails operating directly on the edge.

00:05:26: Let's break those down because that sounds like the ultimate safety net for you, The Consumer.

00:05:30: Yeah so first Safety Guardrails.

00:05:33: These prevent physical harm and act as a hard override if AI tries something dangerous.

00:05:38: Second Security Guardrail to lockdown cyberattacks.

00:05:43: Third are Ethical Guardrailes.

00:05:45: So ensuring privacy Like making sure the cabin cameras aren't broadcast in your face To the internet

00:05:49: Definitely need Those.

00:05:50: And fourth resource guardrails.

00:05:53: These manage the physical limits of a car, monitoring the compute load power draw thermal output so that AI doesn't literally melt on board chips.

00:06:01: Wow!

00:06:02: Literally melting... Okay this is all massive theoretical and architectural shift but setting up theoretical guard rails as one thing doing it profitably another which naturally brings us to China Specifically AutoChina.

00:06:14: twenty-twenty six because that's where this is actually hitting the pavement at scale.

00:06:18: Yeah, auto China is the ultimate reality check.

00:06:21: Vivian G shared this massive takeaway from being on the ground there.

00:06:25: What

00:06:25: did she see?

00:06:26: She said The Chinese market has officially shifted From Being EV first to intelligence.

00:06:32: mandatory.

00:06:33: Intelligence mandatory wow yeah

00:06:35: having an EV powertrain and it's not a differentiator anymore It's just the baseline.

00:06:39: Intelligence is the core purchase driver now.

00:06:42: Which tracks perfectly with what Augustine Friedel was posting about, he noted that there's this intense brutal price war in China right now.

00:06:49: but paradoxically tech expectations just keep rising.

00:06:52: Right.

00:06:52: consumers want The Bleeding Edge But they refuse to pay a premium.

00:06:55: Exactly and Philip Stiebling had this revelation from the VW brand night in Beijing.

00:07:00: That sums it up perfectly.

00:07:01: He realized that traditional legacy benchmarks you know platform discipline complex mechatronics perfect panel gaps

00:07:08: The things Western automakers love.

00:07:10: Right, those are being completely overshadowed by digital speed and local tech integration.

00:07:15: Chinese consumers just care about the software experience

00:07:18: And that's forcing western OEMs to adapt aggressively.

00:07:21: Yeah, NBHEU foot and Peter Bosch were analyzing Volkswagen new in China for China pivot.

00:07:27: Oh

00:07:27: yeah They basically stopped trying to force their European software platforms into china didn't they?

00:07:31: Completely!

00:07:33: AURA T-SIX on the new China Electronic Architecture, the CEA.

00:07:37: Right!

00:07:38: And that fully integrates advanced ADS OTA updates and onboard AI agents.

00:07:43: but here's the kicker.

00:07:45: they didn't build it alone in a closed system

00:07:48: Right.

00:07:48: They partnered locally, didn't they?

00:07:49: Yeah a massive shift for VW!

00:07:51: They partnered with X-Pang and Horizon Robotics... ...they realized that if you want to move fast You have to leverage local experts

00:07:57: Which really begs the question And this is something you as an industry professional Have to be wondering.

00:08:02: If VW can do this in China by partnering What is structurally holding Western legacy automakers back from moving at china speed At home.

00:08:09: Well Heechim Samudi offered A blunt but very accurate assessment.

00:08:13: Oh What did he call it?

00:08:15: He

00:08:15: said European OEMs are stuck in integration.

00:08:18: hell.

00:08:18: Integration Hell, let's explain what that actually looks like inside the car because It is a nightmare.

00:08:24: right picture of traditional European cars.

00:08:26: compute structure.

00:08:27: you're looking at fifty maybe up to a hundred highly fragmented ECU's electronic control units.

00:08:33: Yeah, like one tiny computer for the wipers another for windows and other breaks.

00:08:38: Exactly!

00:08:38: And they're supplied by dozens of different.

00:08:40: Legacy Tier Is running different code not sharing data seamlessly.

00:08:44: You simply cannot build proactive agentic AI on a fractured nervous system.

00:08:49: Sindeep Karey had an amazing analogy for this.

00:08:52: He said legacy tier-I suppliers are facing an HDD to NAND flash moment.

00:08:57: Oh, that's good.

00:08:57: Unpacked that!

00:08:58: Yeah

00:08:58: so think about old computers with spinning hard disk drives HDDs.

00:09:02: they were mechanical decentralized and just off slow.

00:09:06: right

00:09:07: then we shifted to NAND flash solid state drives instant data centralized new moving parts going from a hundred ECUs to a few powerful zonal controllers is the exact same leap.

00:09:17: it redistributes power

00:09:19: completely.

00:09:20: But the automakers themselves are struggling with this too.

00:09:22: Nadia Jamal was at the Automotive Masterminds event in Berlin, and her major takeaway was structure before software.

00:09:29: Meeting what?

00:09:30: Exactly!

00:09:30: That.

00:09:30: The Problem In The West isn't a lack of ideas or engineering talent.

00:09:34: it's organizational gridlock silos legacy culture slow decision-making

00:09:40: And data totally backs that up.

00:09:42: Compeel cited an Alex Partners survey showing that forty-one percent of Chinese OEMs build their SDV architecture entirely in house.

00:09:50: Wow,

00:09:50: forty one percent!

00:09:52: And the West?

00:09:53: Just twenty-five to twenty seven percent.

00:09:55: Jeez!

00:09:55: Yeah, Chinese OEMs know that controlling the software stack is ultimate competitive advantage

00:10:00: and That lack of control in The west directly impacts the consumer especially In the digital cockpit.

00:10:05: like phallux walter argued that the automotive launch and forget model Is completely dead.

00:10:09: Oh

00:10:10: for sure

00:10:10: customers don't care about an oem's internal silo problems.

00:10:13: They compare their cars interface To Their smartphone which updates all the time.

00:10:17: Right, and if your car feels obsolete six months after you buy it.

00:10:20: You're done with that brand which is why we are seeing massive signals from players who get Tim Tuerdahl and Patrick Brady posted about GM rolling out Google Gemini to four million vehicles.

00:10:32: Four

00:10:32: million?

00:10:33: Okay, wait I need to push back here.

00:10:35: is Rolling Out a chatbot two four million dashboards actually useful?

00:10:39: or is it just you know A marketing gimmick because my current car's voice assistant can't even navigate me To the grocery store properly.

00:10:45: It feels like a parlor trick.

00:10:47: i get this skepticism.

00:10:48: but This Is different rigid keyword-based voice recognition, its deep natural language integration.

00:10:55: Okay

00:10:55: how so?

00:10:56: Brady pointed out that Gemini is tied directly into the car's diagnostic data.

00:11:01: So you can ask it like How do I prepare this exact car for an automatic car wash?

00:11:05: Oh Yeah and The AI instantly reads the owner's manual checks the cars state And gives you tailored step by step instructions.

00:11:12: okay That is actually super practical But there's always a. but what happens when you drive in to a tunnel and lose your cloud connection?

00:11:18: right?

00:11:19: The dead zone problem.

00:11:20: Yeah, does your supercomputer just turn into a brick because it can't reach Google's servers?

00:11:24: That is the exact vulnerability Felix Friedman from NVIDIA brought up.

00:11:28: He says that future relies entirely on local inference.

00:11:32: Meaning intelligence lives physically inside the car

00:11:36: Exactly Using small edge-friendly models running on local hardware like Nvidia AI box.

00:11:42: So if you work seamlessly without internet You get zero latency Which

00:11:45: is critical for driving

00:11:47: Totally and it ensures total data privacy because your queries never leave the vehicle.

00:11:52: That changes everything, and you know widening the lens here this on-device intelligence isn't just making the cabin nicer?

00:11:58: It's literally reshaping commercial fleets in infrastructure.

00:12:01: Oh autonomy is graduating fast!

00:12:03: Yeah Alexandra Shaw broke down California's new AV framework which opens the door for heavy duty autonomous trucks A city robotaxi story to massive freight and supply chain logistics.

00:12:17: But fleet operators are so ready for it, Sarah Gallagher recapped RJ Skerring's talk at AC Textbow.

00:12:23: The Rivian CEO said commercial fleets are now viewing their vehicles as rolling data platforms

00:12:28: Not just mechanical assets anymore.

00:12:30: No they're using embedded AI for predictive maintenance Instead of waiting for a van to break down.

00:12:35: the AI detects a microscopic vibration anomaly And predicts the failure before it happens.

00:12:41: It optimizes routing, manages battery health.

00:12:44: It alters the fundamental economics of a fleet.

00:12:47: But the infrastructure has to keep up with that right?

00:12:48: You can't just have smart trucks on dumb roads.

00:12:51: Exactly

00:12:52: Heather Wills shared some amazing work On a v-two x vehicle To everything and three d lead our tolling pilot.

00:12:58: in North Carolina The road itself communicates With vehicles to manage traffic And

00:13:03: it's not just roads.

00:13:05: Pedro Manzano noted That Abu Dhabi is testing AI powered autonomous maritime patrol boats

00:13:11: Boats, okay that's awesome.

00:13:12: Right cities are realizing they have to become multimodal orchestrators of autonomy.

00:13:17: It's mind-bending to think about a city operating as one giant AI network and it all traces back To that first leap from mechanical to learning software?

00:13:25: It does.

00:13:26: but you know we should probably end with A sobering reality check That Matt Damacino brought

00:13:31: up right residual values.

00:13:33: Yeah if cars Are now defined by their tech cycles instead Of how long Their engines last Residual Values They're Gonna plummet.

00:13:40: It's true, best-selling EVs from just a couple years ago are bleeding value because their silicon brains already functionally obsolete.

00:13:48: Exactly!

00:13:48: So we want to leave you with the provocative thought to mull over.

00:13:53: if car ages exactly like smartphone how will that completely upend leasing economics in used car market?

00:14:01: Are we entering an era where you don't even own a car.

00:14:04: You just subscribe to its compute power until your forced upgrade?

00:14:07: It's

00:14:08: huge question!

00:14:09: If you enjoyed this episode, new episodes drop every two weeks.

00:14:12: Also check out our other additions on electrification and battery technology Future mobility in market evolution And commercial fleet insights.

00:14:20: Thanks so much for listening everyone.

00:14:21: Yep

00:14:22: thank you.

00:14:22: Don't forget to subscribe.