What’s the progress of AI and SDV in 2026?

0:00 Introduction to COVESA Meeting

Today in The Garage, we’re recording live at the [2026] COVESA all-member meeting in Porto, Portugal.

One of the things that we do every year at the COVESA all member meeting is Sonatus has worked with Omdia, an important research firm in the industry, to survey the industry and understand the trends in software-defined vehicles. This is our second year sponsoring the annual SDV survey. And last year, working with Omdia, we significantly expanded the survey set to a much larger pool. We found incredibly valuable conclusions last year about how people across the industry were driving ahead towards software-defined vehicles, what they felt was important and value-creating, what are some positive signs, and what are some warning signs about how technologies are being deployed. In this week’s episode, we talked with Maitê Bezerra, who’s principal analyst for software-defined vehicles from Omdia.

We go through the important conclusions of the SDV survey and look at it from a lens of different regions, different subgroups, and what we can take from it to learn about the evolution of software-defined vehicles. It’s a very exciting update to this year’s survey compared to last year. We look forward to sharing with you. Let’s go!

1:29 Guest Introduction: Maitê Bezerra

Welcome to The Garage. We’re so excited to have Maitê Bezerra here with us today. Maitê, welcome to The Garage. You have the distinct honor to be the first three-time guest of The Garage. Because you’ve been here many times talking about this SDV survey a year ago and so on. But before we jump into that, I want to give you a chance for our guest to introduce yourself and tell us about your company and your role.

Thank you, John. First of all, thank you for having me for the third time. So I’m a principal senior analyst, in software-defined vehicles, and I lead the digital mobility coverage for Omdia. So I have been looking at software-defined vehicle as a transition for the automotive industry for quite a while now, and pretty happy to be working in the survey with you.

2:13 Transition from Wards Intelligence to Omdia

So, you know, I know you were with Wards Intelligence for a long time, and I think maybe you want to explain for people the kind of changes that have taken place over the past couple years.

Of course. That’s a very good point. So we were all Informa companies. Right? But I was with Wards Intelligence before, but now we transitioned the name to Omdia because we merged with the Informa Research Group.

But we are essentially the same company now, just a different name.

Okay. Fantastic. So the SDV survey, you’ve been running this for quite some years. This is our second year sponsoring the survey with you. Last year, one of the things we did was expand the sample size so we were covering more geographies and more people. So can you begin by telling us about the survey methodology and how many responses we got as part of the survey?

Of course. Well, it’s interesting you say this because we are at COVESA, and I was in COVESA in 2024, and presented the survey. The sample size was a hundred. So I’m very pleased with the sponsorship because now we have a five hundred and fifty nine sample size, which we had last year and we repeated this year.

And with our sample, we always try to get, an equal distribution of automakers and suppliers. Right? So we have 50% automakers and then 50% divided between tier one suppliers and tier two suppliers. We also try to get representative geographical distribution.

So we have seven countries in North America, in Europe, and in Asia as well.

That’s phenomenal. And we recognize it was so important to be across the world. And we also saw that the data was so valuable, we really wanted to get a good sample size. Now we both have our iPads here because there’s so much data, and we have to go into the saying that what we’re gonna cover is a tiny tiny fraction of this total survey content, but we tried to pull out some things that we thought were really relevant. First, what are some of the key conclusions we took away?

4:03 Key Conclusions from the SDV Survey

So I would say the first conclusion is about the state of the industry. So if you think about software-defined vehicle as a technology that has zero to five levels, our respondents really think that we are in the zero to two levels today. Right? So level two is the max that we have achieved today. Right.

And we’re not talking about self-driving, SAE self driving levels. This is SDV maturity levels, which we’ll introduce on the next slide. But sorry, please tell us.

No. I’m glad you make this difference in between the two. And our respondents think that we’re gonna reach SDV level three by 2030.

So, really, we are the inception of this transition.

The second thing we noticed was we kind of called the great rebalancing. Data, of course, continues to be critically important to the industry. There’s no change in that. But in last year’s survey result, we saw a big focus on data monetization, and we’re seeing, and we’ll get into this in more detail, a shift towards more creating value, creating new capabilities instead of just monetizing data directly. A very interesting conclusion.

Yeah. I really like that one. But another thing that was really interesting is that when you think about readiness for software-defined vehicles, we see that a lot of the enabling technologies are going to be deployed in the near future, but we are still lacking when it comes to the organizational readiness. So there’s a kind of delay of four to seven years in between technology and organizational readiness.

5:25 Understanding SDV Levels

So technology is there, but are the companies ready to take it in? It’s really interesting. Precisely. So let’s begin with this these SDV levels, and maybe you can introduce the SDV levels and talk about what were some of the conclusions we saw.

So this survey, it’s based on this framework that we’re trying to standardize the SDV levels. Right? There are lots of important people in the industry that came up with levels, and what we’re trying to do is get these people together and come up with the common areas of these levels. So trying to to explain them very simply. Level zero is a car that is not connected and the software remains static unless you take the car to the dealership. SDV level one is when you have embedded connectivity and you have OTA firmware updates.

SDV level two, you will start introducing new functionalities over the air, but only in the infotainment. SDV level three, you add new functionalities to nearly all vehicle domains and areas.

SDV level four, full abstraction of hardware and software. So regardless of your hardware version, you have access to the latest software, like it happens with the iPhone today. And SDV level five, it’s the agent orchestrator. So it’s when you have agents that become sort of the brain of the car, and they are working autonomously, inside the vehicle, but they also connect the vehicle to the external world through cloud or connecting to consumer devices or the ITS.

Now, John, we put these levels for the respondents, and we asked them, where do you think most of the vehicles on the road today are? So what did we find?

7:01 Current State of Vehicles on the Road

So as of today, as of 2026, the the predominant level was level two with about thirty percent of the vehicles at level two, which is providing some upgradability, some limited upgradabilty. But looking forward to twenty thirty, the expectation in 2030 is around 30% will be at level three, which is what we call a real beginning level software defined vehicle with obviously some amount of vehicles at higher levels and and a number below that. So it’s showing a shift from level two to level three over the next four years as the center spot.

Which makes sense. And it’s it’s quite interesting that the industry is aware that we we are only at the beginning of the journey.

The next thing we looked at was the timeline of critical ingredient technologies for SDVs. And this chart, which is very complicated, you can see on the screen and we’ll walk you through this, is looking at what was the timeline that respondents felt each of these respective technologies was going to be deployed.

And what you see at the top, and the first four technologies are ones that are over 20% where respondents feel these technologies were already deployed. And then there’s a number of other important ingredient technologies that were below 20%, starting with things like service oriented architecture, time sensitive networking, containerization, and hypervisors.

But then what was interesting is looking forward to the future, what the respondents said. What was the outlook?

So this is what I call a deployment cliff. Right? So this is when we look at the data, most of the respondents believe that the enabling technologies of SDV will be deployed in between 2026 and 2027.

So it’s really really in that between that the enabling technologies of SDV are going to be there. Right? So I I thought that this this is a it’s really it’s interesting, but also it poses, you know, some areas for thought. So if everyone is deploying at the same time, we may face some bottlenecks, we we may have some problems with interfacing of all of technologies. So it’s important to prepare for it.

So it’s interesting because the survey respondents were from OEMs, from tier ones, tier twos. There’s a lot of different people coming together. So if you look at the the bars on the chart, the dark blue bar on the left is those respondents that thought these technologies were already deployed already. And then the next bar, which is the sort of the lighter blue with the red outline.

So if you look at that combined, what you’re seeing is more than fifty percent of all of these technologies are expected to be deployed, and in some cases, more than fifty percent by the end of 2026/2027. Now we’ll see on the next coming slides why there may be some reason for pessimism, why it’s quite this aggressive. But what it does suggest, as you said, the cliff, which I think is interesting, is to think that a lot of these technologies are right at the precipice of being deployed. And if not, maybe not exactly 2026/2027, certainly going through the process of being deployed.

9:52 Reality Check on Technology Deployment

And there is some optimism about these ingredients, which are crucial to making SDVs.

Exactly. The recognition that these are the enabling technologies. Right? Which is an excellent evolution from where we were a few years ago.

Right.

Right? Where people were asking, what are the key technologies for SDV?

Right.

So if we look at the next slide, we call this the reality check. And this was looking at the survey results of what technologies were listed as already implemented. So this is looking backwards in what people said in last year’s survey and what people said in this year’s survey in which technologies were already implemented in the past. Now we should point out that this survey, large sample size, has a margin of error around four percent.

So you have to take with a pinch of salt when you see things less than four percent. So what we saw is three technologies are showing increase in what the respondents said were already deployed this year versus last year, which is containerized applications, hypervisors, and time-sensitive networking. Now of these, only containerized workloads were which was showing ten percent is above the margin of error. Hypervisors and TSN at five percent was just a hair above the margin of error, but nonetheless showing a positive trend for these technologies. But then below, there’s some cautionary tales on some other technologies.

Yeah. So when you go to technologies such as COVESA VSS or microservices, we see that we have a lower percentage of people saying that these technologies have been deployed. Now what this indicates and as you can see, it is within the margin of error. So what really indicates is that we didn’t really see much progress, Right?

These technologies are very likely more complex to be deployed than these people thought. So it’s not necessarily fully deployed. But, yes, as you say, let’s take a step back. Looking at this, what essentially it’s saying is we haven’t moved significantly ahead, but still deploying this technology.

So obviously, you take with a pinch of salt that that that there’s some noise in the system, but regardless, some reasons for optimism, but still some technology still in front of us. Then we looked at the near term future.

12:04 Near-Term Future of Technology Implementation

So this is a similar analysis, but not what’s already implemented. But what would be implemented in 2026/2027 thought of as last year’s survey versus what will be implemented in 2026 in this year’s survey. Very interesting results here. What do you take away from this?

So it’s interesting because we’re talking about the deployment cliff. So most of technology is being deployed within 2026/2027. But when we compare this survey with last year, the the percentage of people in this bracket is actually smaller. So the timelines have increased a little bit. Right? So overall, when you look at this, when you look at these technologies, hypervisors and etcetera, microservices, it’s actually they are going to be deployed, but I would say that this 2026/2027 timeline i actually 2027/2029, right? So if we if we actually take this survey, take a step back, we can actually say that these enabling technologies will be deployed by the end of this decade.

13:01 Diversity of AI Use Cases

It’s important when you’re doing a survey to ask questions in multiple ways, because a lot of times you can you can tease out trends where maybe there’s optimism in one question and there’s pragmatism in other questions. And what this says is that these technologies are at this precipice. They are at this deployment cliff, which is good for the industry. But we shouldn’t be, you know, counting our chickens that every one of these things is going to be happening in the next year, but we need to watch them closely. Interesting. One of the things that that is particularly notable is hypervisors and virtual machines, which we talked about before is something that is is growing.

Actually, the optimism of that deployment actually has really fallen off. That’s a notable one that’s fallen off in this year’s survey compared to last year.

Yeah. And I think this is, well, we can take, of course, their external factors for this sort of, you know, I wouldn’t say lack of optimism, but it’s actually realization. Right? Yeah.

We had the tariffs, we have the geopolitical situation. But I think when it comes to hypervisors, it’s we still have this discussion on where should the workloads run. Should we mix the criticality with infotainment? So I think this is still like, how do we separate these these different workloads?

This is still a question.

I agree. I think it’s for me one of the most interesting things to watch because a number of different, you know, silicon providers and tier ones, tier twos, they all have a slightly different take on that. And there’s and you’ve we’ve also done another question, which we’re not going to summarize here today, looking at vehicle architectures and what you’re seeing is a huge diversity of thought. You would think that at this point, when you think about phones and think about other devices where there’s a sort of a fairly clear architecture and so on, in vehicles, you’re seeing tons of heterogeneity still. And I think that that’s bearing out in these hypervisors because it’s affecting not the hardware, but it’s also affecting the software, how it goes.

You’re spot on. You know, I usually call it people say hybrid architecture. I call it a hybrid zonal architecture. Right? Not just zonal because everyone has a different way of interpreting what zonal is.

That’s true. That’s true. Yeah. Absolutely. Then moving on to data. And we’re here at COVESA, and, you know, despite some a little bit of pessimism of the deployment of VSS, Nonetheless, we’re all here because we believe the importance of data.

15:17 Shifts in Data Monetization Strategies

We believe in the importance of standardization of vehicle data, and we are seeing a proliferation, but maybe a little slower than we thought. But we also asked survey respondents, what would what data collection use cases provide the most value to OEMs? And one of the interesting trends, we mentioned this at the beginning, was that last year’s response, the number one response by a mile was vehicle data monetization, which if I were to summarize it in a way, it’s like directly charging for a feature. Pay for this feature, and we will get direct revenue from that.

That was fifty one percent of the respondents by far the the strongest response by a lot last year was the top answer. This year, that’s changed. It’s dropped a lot. And also it’s dropped regionally.

If you look at the the responses around 51% last year said it was number one. Now it’s down to 44%.

But there’s still a number of other applications of data that are still very strong. And I think what we took away from that was that data is still important, but instead of directly charging for data, we’re using the data to create other value. What were some of the other things we saw in the survey?

Precisely. And I think not only just charging directly, but also getting this raw data and just selling the data. Right? I think the OEMs are realizing that, first of all, the data may not be as valuable as they thought initially.

And also that there were some problems with privacy, with consent. So it’s and I think there’s also this realization that this data may give you a better value if they are used back. So if you use this data to improve your products, enhance what you are offering. Right?

So if you see ongoing improvement in ADAS, data driven product development, it’s something that has grown. Right? Right. So it’s the data going to other areas within the company.

We also saw regional split. Very interesting when we split for the responses from people who said that data monetization was the top answer.

We looked at that, and China is significantly down compared to last year where China said 59% of people said data monetization was valuable. Now it’s down to 34%. And I took away —we chatted about this—I think that our takeaway from this is China is not saying they don’t care about data, but rather I think they’re using the data to improve their products or using data to to improve the driver and the customer experience as opposed to directly charging for data capabilities themselves. I think that’s what I took from that.

Me too. And a strong trend in the US as well. Yeah. Now one thing about Europe that we have been discussing is interesting because Europe still thinks that data monetization itself is the highest application for OEMs. And it could be that Europe has a more developed ecosystem for data monetization. There has been lots of initiatives around this and data sharing from OEMs.

Strong privacy protections in Europe…

Precisely. Precisely. Or it could also be that when it comes to deployment of SDV technologies in production vehicles, we see that other regions are a bit ahead than Europe. Europe’s now catching up. So perhaps Europe will still face this realization that there is more value in actually using the data internally.

Yeah. We’ll see and you will see some data later in the survey about how some European perspectives on some of the data are different from some of the other regions, which is not to say it’s wrong, but it was a different perspective. Different. Yeah. We also looked at how the data strategy varied by region more generally and some really interesting trends here. What were some of things you you observed?

18:55 Regional Variations in Data Strategy

So it’s when you look at data monetization, vehicle monetization, we spoke about it. Right? So going down in some regions. But one one thing that we see that is strong everywhere is using vehicle data to enhance ADAS and autonomous driving.

And we have seen increased optimism regarding autonomous driving, especially now with end-to-end systems that are capable of handling situations that the traditional rule-based systems couldn’t. We also see ADAS becoming table stakes in China with BYD launching ADAS in their in their vehicle line.

Many many regions have stronger requirements. Europe has stronger requirements for automatic emergency braking. Lane keeping is becoming more and more standard.

And I think that it shows a recognition that we can’t think of ADAS and autonomous driving as a solved problem. We think of it as a continuously tuned problem as we learn more, as we gather more data, and we improve things.

Precisely. Precisely. And again, the data driven product development, I like this because it’s actually the OEMs thinking, I can actually make a sustainable business case if I use the data, I understand my customers, I understand how my products are performing, and I make them better. Right?

Another interesting trend we noticed was personalization where vehicle personalization, which many people and and we had Roger Lanctot in the podcast as well. We were talking about the importance of personalization. But regionally, the survey respondents were were split on this. China felt that far and away personalization was a key value creating aspect. And you’re seeing a lot of Chinese OEMs perhaps coming with consumer electronics roots in some cases, coming with a very customer first, customer facing approach.

Versus if you look at the US, Europe, even Japan, the importance of personalization was much lower on the survey. Very interesting trend.

I really like that. It’s a a really different way of thinking of the car. Right? Yeah. And what delivers value.

And I know that we know in China, for example, there’s a big focus on integrating with people’s digital experience more broadly, with smart home and so on like that. That’s probably less proliferated in other regions, I would say, partially because of less standardization, different, you know, regionality in different states and localities. The other thing we noticed is if you look at other uses of data strategy, there’s a number of categories we we split into some small pieces. We looked at diagnostics to reduce or avoid recalls, diagnostics for better service experience and customer loyalty, as well as data driven product and and improvement, which you mentioned.

If we think about those which we split into small slices as a mega slice, they would be far and away the top usage, and we’ll see that in some other data in a minute. So using data to enhance products, using data to improve quality, using data to enhance loyalty, clearly a value creator from the survey respondents across regions.

Yes. And it has been validated in here. I think it resonated really well. We we got a lot of feedback from from people, from automakers and suppliers that this is actually the trend.

Yeah. We should say that here at COVESA, just earlier this morning, we presented this to the COVESA all member meeting participants, and we were had a number of people, as you’re saying, discussed with us afterwards that they really appreciated the conclusions, actually, many different conclusions I had people resonating with after our presentation.

22:15 The Role of AI in Automotive

Which is great.

Now shifting over to AI.

Of course, AI is crucial for autonomous driving and ADAS. Of course, we know that. So we actually said to our respondents, putting that aside, because clearly that would be the top choice, what other applications beside that are use cases for AI, and what were some of the conclusions?

Well, it was really interesting that we have smart diagnosis and predictive maintenance as the top application for AI. Right. So I like it because it’s a vehicle-centric application. Right? It’s delivering value to your customer.

Another thing that was interesting was the vehicle AI assistant. So last year, we had vehicle AI as vehicle assistant, and he ranked around fifth position. Right?

Now this time, we added the agentic part and it’s right on top. Yeah. And I think that the point you’re making here is that when you add the agentic element, the voice assistant is not just a voice assistant anymore. It’s actually an autonomous system taking actions for you in the background, which can improve the vehicle quality and the vehicle response.

Now it’s important to say that we we spoke in the SDV levels that a full agentic system level five is certainly probably in front of us. But the question, the exact survey question was, what are the most promising use cases of AI in vehicles besides ADAS and autonomous driving? So this may be looking to the future. But as you say, it’s clear that agentic AI is clearly a value-creating thing that people are looking at. And there’s a lot of companies, including Sonatus, that’s spending a lot of time focusing in that area.

Other things we talked about, I think if you look at other use cases of AI, what we found is besides those two standouts, which is vehicle diagnostics and AI systems, there’s a wide range of things that were almost tied pretty much, which is feature tuning and continuous improvements, cybersecurity, virtual sensors, route planning, driver safety monitoring and scoring, all of which were almost tied for use cases. So what we’re seeing is a real diversity of uses for AI. And collectively, I think it provides a huge opportunity going forward.

And only features that your phone cannot deliver.

It’s true. Right? Because I think a lot of some people feel and some automakers are saying, well, I’ll just use my phone as the way to interact with vehicles—with drivers. But the reality is a lot of these capabilities require knowledge of the vehicle systems, not necessarily safety critical because that’s another thing. But bringing AI into understanding and improving the vehicles even outside of the safety loop is still hugely value creating if you’re tightly coupled.

Yes. And I for me, this is a direct path to monetization.

Right. Yeah.

Fantastic. And then speaking of predictive maintenance, one of the things that we found is it was a really consistent priority across respondents. What were some of the people we saw respond to this?

So it was it was really interesting because sometimes when you look at your regions, you see a lot of variation, right, or different companies. But, you know, it is top among all of them, and it’s very strong in China. So it’s really interesting seeing China that it’s really investing in AI and putting an emphasis on using AI for the predictive maintenance of the vehicle. So they’re really understanding that the vehicle quality is what they should be looking at at the moment.

Right. China was very strong on that. North America as well, 43% of respondents thought AI was a top priority. Tier two suppliers as well, which is really interesting because you generally think that if we’re going to be doing predictive maintenance, the tier two suppliers who are making subcomponents, really, they’re going to have to be engaged in this if we expect it to succeed. They see value in this, and it means they’re probably investing in it. There was one standout, though.

There was an outlier, which is Germany. It’s quite interesting that they put very little emphasis in in predictive maintenance for their AI investment, which could indicate that they think they can do it without AI. True. It’s true. They have different priorities.

Yeah. I mean, we we were joking that, you know, the German quality, attention to quality is famous and historic. And so it may be that they’re they want to build in quality in such a way that they don’t need to build in AI, or they want to use more conventional approaches to to solving their problem. So but it it is interesting trend. And more generally, we looked at monetization and we we said what kinds of features and services would create the most customer loyalty or after-sale revenue. So it could be loyalty or revenue. What are some of the things you saw there?

26:44 Predictive Maintenance as a Key Focus

Well, what is on top again? Predictive maintenance.

Right.

Which is really, again, looking at the quality of the vehicle and how the vehicle is delivering value to how the drivability actually is is taking place. Right? But, we also see automated driving, ranking really high, and enhanced personalization. Although it’s interesting that there is a lot of regional differences in here.

Right? But I’m gonna talk a little bit about that, but what I think is important as a takeaway is that the OEMs can’t really take one strategy. Right? One strategy fits all.

You need to have a software defined platform that is flexible enough for you to adapt to different regions. Right?

Because most OEMs in in most cases, not all, but in most cases, companies are selling worldwide.

And thinking and we’ve been talking about this for years. Thinking that you can have a one size fits all strategy across regions is not a good approach. There’s regional regulatory differences in some cases. There’s privacy differences. There’s regional preferences differences. And so one of the key values SDVs may bring is the ability to tap into diverse markets around the world with a common hardware platform in a reusable way.

Precisely. And this leading to a sustainable business model. So then you can deliver high in-vehicle entertainment in North America, which is something they value. You can get a give an emphasis in automated driving in Japan or in China where they value it a lot and predictive maintenance in North America or vehicle ride customization, which is highly valued in Japan.

It’s the second year in a row that I have to point out that that automated driving in Europe is was low last year and continues to be low and has dropped. I think the Europeans just like to drive their cars, and they want to be in control of what to do. But in China, it’s it’s strong and growing, and in Japan, strong and growing. Very interesting.

Yeah. But who knows? With the new new systems, the end to end systems, I think maybe we will see some opinions changing in Europe at some point.

The last one which is looking at this, we talked about predictive maintenance. It’s something that Sonatus has been focusing a lot on. We see a lot of interest from our customers in this area, and we’ve been working with our AI Technician product very extensively. It was so interesting to see this data because what what it showed was predictive maintenance came out on top in multiple different directions.

It came out as the top AI use case outside of autonomous driving. 34% said it was the top use of AI. It’s also the top revenue and loyalty driver. 44% of respondents said it was the top driver.

And it was also consistently seen as valuable from tier ones, tier twos, and automakers, suppliers. All of three all three of them said that was a key key provider, a key value creating thing. Very interesting.

Very interesting. I think this is definitely going to be the focus for the industry moving forwards. And, again, I’m happy with this. This is what the industry should be focusing again in making vehicles better.

Making them better, improving the loyalty experience. We were chatting earlier about how customer loyalty is incredibly valuable for for OEMs. Customer acquisition is hard. So if you can retain a customer, it’s much better. So if we can improve the service experience, if we can improve the quality, you’re going to get the driver to come back and buy another vehicle from you in the future.

And you touched a really important point, which is differentiation. Because moving to software-defined vehicles, many OEMs get worried about how they’re going to differentiate and and keep their unique selling point. Now if you focus in the quality and keeping your vehicle for a long time in the roads and vehicles that can receive new functionalities for a long time, I think you you were definitely going there.

30:28 Accessing the Survey Results

Fantastic. Well, Maitê, this is a huge survey.

Let’s talk to people about how they can get access to it. On Sonatus’ website, very soon you’ll see a link, we’ll provide this in the show notes. We’ll have the survey, some key conclusions of the survey you can get from us. And if people want to dig in even deeper than that, they can reach out to you, and we’ll have your contact info on the the show notes as well. They can reach out to you at Omdia to learn more about this survey.

Yes. And learn more about the other questions. It’s a very large survey, and I’m very excited about that.

Thank you as always for joining us every year. It’s a pleasure. We did this at COVESA last year and exciting to see it again. And thanks for for joining us on this important initiative to understand the industry.

Thank you.

If you like what you’re seeing from this episode, please like and subscribe to see more like it, and we hope to see you in another episode again very soon.