Software-Defined Systems: The Architecture Decisions Shaping The Future Of How We Move, Live And Work - Six Five On The Road at CES 2026

What’s shaping the next generation of vehicles isn’t always visible from the driver’s seat.

From CES 2026, Olivier Blanchard sits down with Mark Ng, Director of Automotive Systems at Texas Instruments, to explore how software-defined architectures are changing vehicle design. They dissect the systems that now decide how vehicles perform, scale, and stay safe, from semiconductors to real-time intelligence, and the constraints OEMs and Tier 1s face as software takes the lead.

As vehicle lifecycles stretch and updates become continuous, the takeaway is straightforward. Platforms that can adapt to growing data, new AI workloads, and rising system complexity will define what scales next.

Key Takeaways:

🔷 Software-defined vehicles depend on strong architectural foundations: The most impactful automotive innovations now happen below the surface, where system architecture determines how safely, efficiently, and reliably vehicles can scale over time.

🔷 Edge intelligence is becoming essential: Real-time decision-making increasingly requires compute to move closer to where data is generated, reducing latency while enabling faster, more responsive systems.

🔷 Performance and efficiency must advance together: Automotive AI demands higher capability without sacrificing power efficiency, reliability, or functional safety, forcing tighter engineering tradeoffs across the system.

🔷 Architectural choices today shape longevity tomorrow: Decisions around zonal architectures, centralized compute, and sensor fusion directly influence how vehicles adapt to software updates, new workloads, and rising complexity over long lifecycles.

Sponsored by Texas Instruments.

Watch the full video at sixfivemedia.com, and be sure to subscribe to our YouTube channel so you never miss an episode.

Listen to the Audio Here

Disclaimer: Six Five On The Road is for information and entertainment purposes only. Over the course of this session, we may talk about companies that are publicly traded, and we may reference that fact and their equity share price. Please do not take anything discussed as investment advice. Six Five Media and its hosts are not investment advisors, and viewers should not treat the content as such.

‍

Oliver Blanchard: 

Hi everyone. We're Six Five On The Road and we're here at CES 2026. Today, we've partnered with Texas Instruments on this segment to talk about how software defined architectures and real time intelligence are reshaping system design. And because we have a lot of questions and topics to discuss, I'm going to jump right in.

So Marc, thanks a lot for being with us. So the first question I'm going to read out and then we'll get into a conversation. Okay, so we're at CES 2026. Um, impression so far. What are you excited about? What's your main kind of takeaway so far from the show?

Mark Ng: 

I think the resounding theme of CES is, of course, AI. You cannot go to any booth here at CES without seeing AI or intelligent or physical or software defined. No matter what technology. I think what I'm really excited about is that there are so many different people there talking about AI. But what I want to see is basically five years from now, who are those companies that are going to be able to deliver on the promise of an AI car, or AI boat, or AI golf divot fixing machine?

There are very specific applications which I know and I can feel will make it in automotive and robotics. And I think I look forward to seeing the semiconductors, which are foundationally required for any type of technology with AI to be rolled out. That's what I look forward to.

Oliver Blanchard: 

Many of the most impactful innovations tend to be invisible to most end users. And yet they're critical to safety, efficiency, scalability even. And when you look across how people move, live, how they work, what role do you see semiconductors playing as the foundation that makes all these advances for UX and technology possible?

Mark Ng:

 I think one of the best examples, you know, we're very, very accustomed to the word AI here at CES. It's everywhere. We tend to think about those big processors and everything, and we think that's the invisible parts of it. But if you think about data centers and even cars or robots, there's a lot of unsung heroes that are out there. There's the communication networks that happen in order for the processors to be able to talk to other parts at very high speeds, the ability to sense real time information, the ability to compute properly.

So there's even the requirements to power those semiconductors. So when you look at these boards, there are a significant amount of devices, not just that one central chip that we tend to think of. There's just a lot of devices that are required to make that work.

Oliver Blanchard:

 So I'm thinking, okay. Texas Instruments has been around forever. I remember seeing your brand on devices that I had as a kid. So you've been a player in the edge for a very long time before the edge was called the edge, right? And so now here we are, this long household name of a brand that has this storied, very long history of excellence and innovation and products. And we're in this new era of self-driving cars and robots and AI. From your perspective, what separates companies that are architecting for that future today? As data and AI demands continue to grow.

‍

Mark Ng: 

There are several keywords we see this year. Intelligent, physical AI, software defined. The companies that will make this transition happen properly are those that are focused on the whole picture. I can make the world's most advanced processor, but if I'm not focused on efficiency or communication with other devices, it does not work.

It's the companies focused on sensing the physical world, communication between chips, and compute. That ability to sense, communicate, and compute creates the ecosystem for intelligent systems.

‍

Oliver Blanchard:

‍That brings us to a challenge. As AI workloads move into vehicles and robots, engineers are being asked to deliver significantly more performance without compromising power efficiency, reliability, or safety. Umm Are there trade-offs? Right to balance those. And how do those choices impact long-term scalability? And actually, what do you think about the right balance is for that?

‍

Mark Ng: 

‍I think that engineering is all about trade-offs.  The easiest things we can think of is in vehicles and robots and all that stuff. Um, we’ve, we’ve all heard 48 volt in market and very famous person in the industry as Ted Talks about 48 volt  in vehicles and not that technology which actually transferred over form, you know, communications  equipment and server, you know, data data centers over to, to, you know, vehicles and our robots. Does it really make sense to apply 48 volt everywhere? 

And the answer is it probably doesn’t at this moment.Right. Because I don't I have a lot of legacy parts, um, that are built on, you know, maybe 12 volt architecture. 

‍

Oliver Blanchard:

‍Okay. 

Mark Ng: Um, you don't want to upgrade that all at once because it may not make sense from a cost perspective, even from a technology perspective.

An interesting soundbite. Maybe if I tried to do a 48 volt windshield wiper. You know those for reference, those are quite simple, but there are use cases, right? Maybe there's a you have some and this is real, right? You might have a carmaker that says I need a 48 volt windshield wiper because, um, you know, my cars are designed to drive at 200mph, and I need to keep that, that windshield wiper down.

So there are situations like that where I have very unique requirements for very specific features. But then in our opinion, it's really the the ability to, um, you know, not say that we're going to make a full transition over to a new technology, and it's the ability to give the engineers the choices to, to, to transition.

So with regards to this 48 volt ecosystem, whether it's robotics or whether it's, um, you know, cars, we are enabling our silicon ecosystem to be able to handle that. Right. And we will basically maintain both 48 volt as well as 12 volt compatibility. So it's basically two separate families. Now. The overall the way that you design for the future is basically as I mentioned, you have to be able to communicate.

So workloads are going up. Data centers. You have basically, you know, cars are essentially, you know, data centers on wheels, except you try not to have liquid cooling in a car. Right? Right. So the the overall network or the overall communication network that you design has to be able to be future-proof.

So whether you're looking at Ethernet protocols or PCI protocols or, you know, temp based ones protocols can protocols, Lin protocols, those protocols will exist for a long time. But those who actually do really well are the ones that are actually able to embrace legacy and also actually be able to kind of merge everything and put it into modernity.

‍

Oliver Blanchard:

So actually so I have a follow up to that, since we're talking about future proofing all these technologies. Right. With with longer vehicle life cycles, um, in part aided by OTA or continuous software updates right over the year updates, um, becoming the norm. Um, what architectural bets should OEMs and tier ones be making today to support that type of longevity to.

To invest in the right technologies and the right partnerships, uh, to be able to achieve this. And I guess I can ask again, uh, if we get into into the weeds a little bit, but what assumptions should the industry, um, sort of put their bets on? Right. 

‍

Mark Ng:

I go back to that word physical, which actually I wasn't too keen on, but I see it more and more. I think the key assumption is that the physical aspect is going to be present always, whether you're in data centers, robots, uh, cars. Because for me, where I work on cars, what's physical to me is things like radar or things like lidar or things like batteries or, you know, communication, like audio, uh, speakers, all that stuff in the physical world is going to continually change and it's going to overall that's going to enhance our overall life.

The way we move, the way we live, the way we work. Yeah. Um, and so I think it's that ability to, um, capture the physical world and to actually put that into our overall electrical infrastructure. Right. So we're going to have to take that physical world and sense that better. Yeah. And we're going to have to invest in technologies that are going to bring us to that next level.

And I think then it's those that the the investments that would be made would be to choose the best in class physical devices and then to choose the best in class communication networks. And then, you know, processing is the central component of everything. Um, there are,

you know, SoCs or processors that are out there that are specifically designed to handle AI. Right. We have these neural processing units. Different vendors are attacking in different ways. Some are using GPUs to do it. Some. We're basically using our real time cores. But future proofing requires the fact that you acknowledge you're going to have more physical data, and you're going to need to maybe process multiple streams in different fashions and basically accommodate for AI.

Um, and so yeah, AI is here regardless of what we, uh, whether we want it or not. 

‍

Oliver Blanchard:

Uh, yes, it depends who wants it and where. Um, so, okay, I'm going to circle back to the the sensor fusion with, uh, radar, lidar and cameras because you just mentioned that. But first, um, I'm thinking okay. So systems are becoming more defined by software, right?

Um, so we have, you know, growing data flows, real time decision making. Um, and so a lot of the conversations that I've been having at CES, specifically and for the last few years, not just this year, have been about zonal architecture and centralized compute. And I'm seeing some of that move beyond the automotive space and go into robotics as well.

So you feel free to to talk about either one or both. But how are zonal architectures and, and centralized compute reshaping vehicle design. Um, and what can that evolution tell us about how intelligent systems are already being built and where that's going in the future?

Mark Ng: 

From a vehicle perspective, there's no doubt that we're moving into this zonal architecture.

And you know what I mentioned the the the intelligent, the software defined. That's what it's all about, right? So from a, um, of a very high level, as I mentioned, it's a car is essentially moving into a data center on wheels kind of thing. And so we have a central compute unit. Uh, typically you might have one big processor on there sometimes has two.

That's basically where I'm kind of having the central intelligence of the vehicle. Now, um, with the zone architecture, I am trying to consolidate functions into various zone boards that surround the central compute. So I might have 3 or 4 zone boards that surround it. But the overall concept is that I am consolidating all of the, you know, what I would call legacy edge nodes.

You know, like maybe a window or a door or a seat. If you're looking at a robot, you might be looking at a hand or a leg, right, a different joint or appendage. And I'm basically trying to consolidate that into one bigger board. And the advantage of doing so is that I'm reducing wire complexity and weight, and I'm reducing weight.

I'm reducing the overall cost of the system by actually reducing components. Because think about I don't know. We can think about my high school car, which maybe had well, I didn't have a lot of features in my high school car. But simply imagine that if you have a car without zonal architecture, you're going to have a different ECUs or electronic control units.

And so maybe you might have a window module and you'll have like a, like a roof module and a seat module. You're going to have like a fan on a blower. And so you can imagine there's hundreds of what we call MCAS. And basically they're all redundant at a certain point because I'm buying it from a and I'm buying it from B, and I'm just sniffing it in there.

So what if I tried to consolidate that all into one main board? Because, you know, silicon is getting more advanced day by day. We're not in the 1980s anymore. Yeah. And we have the ability to consolidate. So that's what this zone architecture is all about. Now the advantages of it I mentioned are really in terms of reduction of weight.

And, um, you know, it's really in terms of the intelligence aspect and also the software defined aspect, because I'm actually able to consolidate and centralize the way that software is developed and also rolled out because I have a central compute unit, maybe that's connected to the cloud, and then that can actually roll out updates across the entire system or half of the system.

I can do it real time. I can do like, you know, over-the-air updates and so on. So that's that's really what Zone Architecture has, has brought us. And it really is no different from a lot of what we've done in industry. It's it's just a natural evolution. Right. It's just we don't want to put many things together and see what happens.

We kind of want to have a rhyme and a reason for why we're doing things.

‍

Oliver Blanchard:‍

And also, I'm guessing it also simplifies the supply chain and just allows you to scale a little bit faster?

‍

Mark Ng:

It does. Yeah. Yes. 

‍

Oliver Blanchard:

All right. So talking about zonal architecture and going back to as I promised to sensor fusion across radar lidar cameras.

Well, that sounds great, but I know it's difficult. Real world deployment is hard for having having these systems work together properly. So how are companies working to solve the challenges that come with implementing really complex designs like those, especially when you're trying to connect them together?

‍

Mark Ng:

So I think that all has to do. That's that. The third angle, the most the third most important angle, which is compute I think from TI. We are trying to address that from the processors that we're doing. So like our TDA family, which we just announced is the TDA five that has you know, I won't get into the technical details, but that has like, you know, Taro teraFLOPS we're talking about, you know, trillions of flops of of operations per second.

And these are devices that are designed to handle multiple loads of radar data that's coming in on one one IO stream. And there's basically a camera data coming on, coming through on another IO. And then you've got like, you know, communication that's happening elsewhere. So the devices that we're making are basically designed to handle independently.

Yeah. Um, multiple streams of data and process them independently because we're not actually using the application core to do it. We're actually accelerating via via hardware. And so that's it's not unique to T.I. Um, you know you have to do hardware acceleration. Just some companies maybe do it by like GPU acceleration.

Right. Um, and when we choose to do it by real time acceleration through our C7 cause, um, essentially our, you know, if you look at T.I. or what we've built our name on is our DSPs, our C7 real time cores are designed to do things very quickly at very low power. So what I like to say is that it is an extension to what I what I said earlier, cars essentially are data centers on wheels, but what we are trying to differentiate on is really to develop.

We're trying to develop our design data centers on wheels that don't require liquid cooling, because we're not trying to throw GPUs at it. Correct. Yeah. So that's kind of how we try to handle. Sensor fusion is basically to handle it at the compute level to have processors that can, um, you know, hardware accelerate to be able to do it at extremely low power.

But we also are trying to innovate in terms of how we, you know, when we sense, right, the ability we can sense very well every single generation. We're actually sensing better, more resolution faster, but that requires more bandwidth. Yeah. So so bandwidth is only going to go up. So that's where the the communicate aspect we are thinking about how to you know, we talk a lot about edge processing.

If you take a step back when we have sensor fusion, you may not necessarily want to process at the edge because you might want to just get raw camera data from a robot coming in. You want to get like the radar data coming in and maybe some lidar to confirm if you're driving a car and if you're able to get a very powerful processor that's already hardware accelerated, to accept all those streams with data streams that are raw and not pre-processed, I'm actually better able to, to to to sense the physical world.

So yeah, it's still to us comes down to that central theme of how do we sense, how do we communicate and how do we compute?

‍

Oliver Blanchard:

Cool. No, that makes sense. And actually, I really like that you, uh, you you gave us the the nugget that I was looking for that what differentiates Texas Instruments and your approach from a lot of the other ones that we've seen so far in the industry and why that matters.

So thanks a lot, Mark, I appreciate it. This is all the time we have for, uh, for our interview. Hopefully we'll talk again soon. Uh, I'm going to dig into your, uh, your announcements, uh, a little bit deeper and maybe do a little bit of coverage. Thanks for tuning in to Six Five on the road at CES 2026. Uh, don't forget to hit subscribe.

Obviously, catch us on the socials and check out all of our coverage at SixFive.com and until then, see you next time.

‍