Ajit Chander Swaminathan of Frost and Sullivan

0:00 Introduction to the Guest

Today in The Garage, our guest is a Ajit Chander, who’s associate partner and mobility practice leader for Frost and Sullivan. Today’s conversation, we talk about a recent analysis report from Frost and Sullivan that was partially sponsored by Sonatus, where we looked at cloud and edge AI and the impact and benefits of those incredible innovations in the mobility space in automotive. It’s a very detailed report where we looked at a number of different, application and use cases that provide tangible benefits. And in today’s podcast, we talk through that report and hear from Ajit who has incredible background in automotive and mobility. Let’s go.

0:50 Ajit’s Background and Experience

Today in The Garage, our guest is Ajit Chander from Frost and Sullivan. Ajit, welcome to The Garage.

Thanks for having me.

We’re so excited to have you join us today. Start by telling us a little bit about yourself.

All right. Well, I am an engineer. I was born and raised in India.

And I actually joined Frost and Sullivan in India before I moved to North America. It was primarily to Toronto. Been with the company a little over seventeen years.

Seventeen years, that’s a long time.

That’s a longer time I care to admit, yeah.

And I worked across multiple industries on the IT services sector. I worked in the manufacturing sector before that.

So I bring that cross industry experience.

Primarily all these years with Frost and Sullivan, I’ve been in the automotive and transportation industry working across the different players in the ecosystem from OEMs to suppliers to the finance of institutions to the oil and gas companies to the energy players.

So a pretty broad …

Very diverse background.

That’s great. So you got to start with a fun fact about you.

Well, so my daughter, so I have two daughters. My younger daughter, she picked up playing volleyball, a sport that I’ve never played. I played other sports. I’ve been decent at some of the sports, but never played volleyball.

But, you know, when when when your kids, they get get started and start involving the games, know, as a dad, you start to get involved in it. So I decided to train her. So I started learning the sport and then we hired a personal gym and started training her on a weekly basis.

And we started seeing some results in the games, so did her friends. So they wanted to also you know once they realized that you know the parent parents started asking me what’s going on, I mean what’s the change? So we had to tell them what we are doing so they started bringing their children then. In no time, I started managing a teen girls team, the snacks, the drama, and the group chats. Also survived a lot of eye rolls, but still had them run all the drills.

Well, that’s fantastic that you’re getting so involved in your in your family and your and your daughter. My fun fact, I always like to have a fun fact back to our guest. My daughter was in some many years ago when she was in fourth grade, a science competition. I’m on a board of a science center in in San Jose, and they joined the science competition for kids, you know, fourth through twelfth grade.

And so I was a coach for a team of fourth grade girls doing the science competition. So it was so exciting to teach them about how to use tools safely and how to build things because they didn’t know much about that. And so it was good fun. So there’s my my fun fact back to you.

Nice. Okay. Okay. Nice.

3:36 Overview of Frost and Sullivan

So tell us about Frost and Sullivan. Frost and Sullivan is a is a famous consultancy and analyst firm. Tell us about your your role and and what are the scope of Frost and Sullivan they do?

So I am a associate partner and I lead the mobility practice at Frost and Sullivan. I’ve been with the company seventeen years, like I said. I decide the business strategy for mid to long term.

At Frost and Sullivan, we look at disruptive forces and how it reshapes the industry.

We are looking at technology that impacts our clients. We are actually in the right in the middle of the forefront of technology integration. We’re bringing together OEMs and suppliers and the tech integration partners. And as part of this, we do publish a lot of research studies. We engage on strategic projects.

We also publish a lot of industry wide papers looking at emerging opportunities.

That’s what we do at Frost and Sullivan.

4:36 Research report on In-vehicle and Cloud AI

Well, and that’s quite a good transition that you were the lead author on a report we were involved with you recently on cloud and edge AI. Looking at how AI has potential to really transform vehicles and add value.

It was so great to work with you and so thank you for your partnership on that. And it was also an incredible work product. I’d love it if you start with an introduction of what were some of the key conclusions and takeaways you saw in doing that work?

Sure. So let me start with what are the impending challenges in the industry to achieving the SDV objective? So there are the fragmented ML Ops framework. The data required to train these models comes from different ECUs that run on different models, more different models use different frameworks like PyTorch, TensorFlow and the validation, testing, deployment and management of these different workflows are handled differently in terms of tool chains and teams. So this causes significant silos, and there is the disparate hardware and software.

That’s such an important conclusion because, you know, as we’re working to deploy AI into vehicles, one of the things we want to empower OEMs is to use the diverse hardware they have. We shouldn’t be thinking that, well, we can’t do AI until there’s central compute, or we can’t do AI until there’s a giant optimized AI accelerator.

The reality is, and I think this was a conclusion borne out in your report, is there’s so many use cases across the vehicle that can be much more modest performance requirements and can get deployed on today’s hardware.

And that’s one of the things that Sonatus is trying to do and it was exciting to see that I think your report bore that out.

Yeah. You’re right. Mean, so to continue, the data is coming from different ECUs and ECUs run on their own RTOS and operating system and middle west stack.

And that comes with data access and lifetime OTA management challenges.

Right.

Now look at what are the key enablers of SDV, right? The decoupling of application software from hardware with middleware stack and operating system, standardizing interfaces, creating modular designs, continuous updatability and upgradability with integration.

Right.

And then creating SDV ready hardware.

Yeah, exactly. Absolutely. And your report said this was quite a sizable market. I mean, what’s the outlook on the potential for AI in vehicles and AI in automotive?

7:17 Market Outlook for AI in Automotive

Alright. I mean, we estimate the SDV spend by 2030 to be upwards of $500 billion. And then we and we analyze the different use cases.

There’s a current impact in 2025 of about $48 billion that grows up to $238 billion by 2030. This includes includes several interesting emerging use cases like prognostics, usage based insurance.

So this is the AI part.

So the AI portion of SDV spend going from 2025 to two hundred and what in 2030?

From $48 billion to $238 billion.

So this is real money. This is not a niche application and it’s also you were starting to say…I’d love to understand some of the use cases you found were the most maybe compelling or value generating.

8:05 OEM Strategies and Investments in AI

There are there are quite a few. I mean, I would like to just cover one more point before that. I mean, we looked at what are the challenges in the journey to SDV objectives. We looked at what are the, you know, SDV enablers.

So we also now look at what are the OEMs doing. Right. I mean, as you looked at General Motors announcement of central compute architecture. And then that increases their AI compute capabilities by 35 times and their OTA capabilities by ten times. Now OEMs are trying to convert these assets to monetizable opportunities.

Frost and Sullivan research shows that ten other OEMs or ten plus more OEMs are moving towards a central compute architecture with middleware stacks with, you know, ethernet backbones and OTA for their DevOps transition

To secure a five-plus year window for continuous upgrades. When look at what Tesla did in just in 2025, They sent, know, to eight OTA updates, made approximately, I think, about forty two feature updates just in 2025.

Significant. Yeah.

I think there is an opportunity to deploy AI models existing hardware in the current generation silicon. So no advanced NPUs or high performance SoCs are required.

And then this has immediate impact in terms of the use cases that we have analyzed as part of the white paper.

That’s great. Your report was so impressive. You’ve done and we’ll put a link to the report in the show notes so people can can look it up and find out and find out about it. You put a lot of numbers behind the impacts and the the potential benefit of some of these use cases. Now here at the at the show at CES, just outside, we can just outside where we’re we’re sitting here, we can see our AI Director, Sonatus AI Director examples where we’re showing many different use cases. I’m wondering which are some of the use cases that stood out to you as particularly impactful, particularly valuable?

10:00 Emerging Use Cases in Automotive AI

I think there are several emerging use cases that we analyzed as part of this white paper. I mean, a lot of them on the ADAS and cockpit use cases are well documented.

They’re well known, yeah.

Yeah, so, but we looked at several other emerging use cases that has an immediate impact. We looked at prognostics, usage-based insurance, sensor virtualization, edge cybersecurity, in-cabin sensing.

And we thought a lot of these have a significant revenue opportunity, monetization opportunity for OEMs. And things that actually stood out for us in terms of higher impact are sensor virtualization. And I can give you a couple of use cases that we looked at.

And in fact, you have some real world partnerships on those with COMPREDICT.

We’ve been working with COMPREDICT, it’s a fantastic partner that shows how something that would have normally be done with an IMU, know, a motion sensor that detects angle of vehicle and accelerometers and so on. They can remove the IMU, save twenty dollars in vehicle BOM costs. And twenty dollars may not seem like a lot, but it’s a colossal important cost savings as OEMs are trying to squeeze cost out of the vehicle. And that’s just one use case. That’s correct. There’s many other use cases for things virtual sensors just in your example.

Yeah.

And then the EU is mandating it from 2027 That’s right.

For headlight leveling sensors to go on all the vehicles, and it’s going to add BOM costs for OEMs. And of course, there is the indirect TPMS that can infer tire pressure from antilock braking, inertial measurement units, speed sensors, visual input.

And it can detect leaks, deflation, and efficient tire management solution overall.

11:59 AI Not Always Driver-Visible

It’s true. What struck me, and we’ve been visiting with many, many customers throughout the show here this week, and I’ve been explaining and talking through these different technologies. The thing that struck me from the report and from these many meetings is it’s not only that we’re adding some new feature to the IVI. It’s not only that we’re adding a thing that’s driver visible.

Some things, driver visible. You might provide a usage-based insurance that lets the driver see feedback to help them be a safer driver and then by extension get lower insurance rates. That might be driver visible. Then you might have something that’s slightly less visible like tire wear, where you’re monitoring the tire wear and maybe talking to ADAS system to say, hey, the tread’s getting a little low, you better increase the braking distance, you better be a little bit more conservative with the acceleration you permit.

And then later, sometimes you warn the driver, hey, by the way, your tires are getting low, you should replace the tires and here’s a great way to get OEM brand tires there. We know we’re gonna work for your vehicle. So that’s sort of like semi driver visible, sometimes visible, maybe sometimes invisible. And then you have something like the virtual sensors you mentioned, which might be completely driver invisible. Driver never knows, but that we’re able to drive BOM cost out of the vehicle, which maybe allows the OEM to have maybe better margins or maybe be more cost competitive with their their vehicle or perhaps combination. So I think there’s a spectrum of things that are very visible to not visible at all. And I think that’s a for me, it’s a key observation, and I think for the industry, not many people appreciate that it can be very diverse like that.

Absolutely. And in fact, I think when I when we saw this analysis, we kind of looked at three buckets.

One is directly impacting the bill of materials for OEMs. And then one increases the operational efficiency and then gives them the ability to continuously monetize it over the life cycle of the vehicle and then creates this personalized experience for the users, end users.

So that I believe is a significant opportunity and that accelerates OEMs towards the SDV objective.

That’s I love the way you’ve broken that into a slightly different taxonomy than I did. But it’s interesting because in some cases, like, example, the BOM cost. If you’re able to take costs out of the BOM, it’s immediate revenue or cost benefit, immediate top line or bottom line margin benefits.

And then if there but there’s ongoing benefits, well, that’s also useful and it may provide some ongoing services revenue or ongoing parts sales revenue, which OEMs are always looking for.

But in a way that possibly drivers may find attractive. Because there’s we always talk about this. We’ve talked about this many times on the podcast is that people say, “Oh, well, getting getting money, getting subscription from drivers is tough”. But the reality is it’s not tough if you deliver them something value creating for them. If you you deliver them a feature they want, they’ll be happy to subscribe to it. It’s when they feel that they’re pushing something on them that they don’t want or they feel like they should be getting for free.

You know, we have conducted several research talking to end users, understanding what are they willing to pay in vehicle. You know, looked at features on demand, looked at safety and security, vehicle performance. So anything that can enhance their in car experience and create a more personalized experience. If you look at why Xiaomi is doing so well. Now they have this connected ecosystem, right? They have appliances in the house that are connected.

Right.

And then you get in a car and they see all these as connected entities.

It’s a continuum of your digital life.

Right. Mean, see the appliances are robots inside your house and the car is a robot that’s outside your house. And then, you know, you add to this concept, the humanoid, then that becomes a connected ecosystem, right? I mean, again, the ability for OEMs to make the life seamlessly connected for the end user is what the end user is looking for.

That’s really exciting examples you mentioned. And you also made me think about something, your last point that we as an industry, we don’t know what algorithms are coming next. We don’t know the kinds of compute requirements coming next. And so I think it behooves us and I think it’s what Sonatus is trying to do, but I think the industry collectively is trying to do is we have to provide the infrastructure for things that are coming next.

16:13 Future of AI in Automotive and Continuous Improvement

So that OEM doesn’t ship a vehicle and the moment they ship it, they’re they have some sort of handcuffs that prevent them from delivering the next exciting innovation, but that they come with some flexibility, some headroom, but not necessarily only headroom, but especially flexibility to deploy things. And or even models that are already there, realizing the models are going to improve, the models can be refined. How can we make sure they’re upgradable both through OTA but even through other means. Our Sonatus AI Director product, for example, AI Director, is designed to allow deployment of models very easily in a very lightweight fashion that’s easy to do without the kind of the heavyweightness that you might associate with a full OTA.

Because AI models, we expect and I believe, could evolve more rapidly, could evolve more frequently as data tuning exists. You’re seeing that with autonomous driving. You know, autonomous driving cars, imagine you consider Tesla or Riviera or something like that. Yes.

They have feature updates. You talked about that earlier. But they’re also doing continuous updates to the driving algorithms. Continuously. And not only when you request some update, rather they’re continuously tuning the parameters.

I believe that from many of these AI models that will be like that, where there’s continuous improvements in AI to make them smarter and make the models more accurate.

Instead of only relying on heavier OTAs that maybe happen every month or every two months or every six months.

I agree. And in fact, I’ll bring up two points to your first point. The high-power SoCs on ADAS and cockpit, the growth of high power SoCs has been significant. From Qualcomm to NVIDIA to Horizon Robotics, AMD, if you look at ten percent of the Chinese vehicles that get manufactured today have high performing associates from Qualcomm and NVIDIA.

Right. The NPU engines compute-intensive AI workloads are increasing. And this is key to seeing agentic AI models use cases within vehicle like the virtual personal assistant. Now you need a single orchestration AI model that manages the end to end life cycle of AI in cars and this is key to acceleration towards the SDV objective like we discussed at the start of this conversation.

18:13 Sonatus Vehicle AI Products

The solutions from Sonatus is critical to achieving the SDV objective.

The Collector AI, the AI Director, Automator AI and the AI Technician. It’s end to end lifecycle AI management tool chain and runtime environment that provides secure containerized runtime environment for a lot of these deployments.

Yeah. Thank you for saying that. And we’re we’ve been showing so many different ways that that all those products you mentioned work together. You know, in some cases, OEMs may have the data they need to manage their models and then we can use our AI Director product to deploy them. In other cases, some of our model vendors are saying, you know, actually, I really would love to be able to use your data to tune the model on the vehicle. In some cases, you know, in some models, you actually need to optimize it for the specific vehicle’s weights and tires and things like that. So actually, some of our partners are wanting to use Collector AI to gather in-vehicle data for this specific vehicle to adjust its parameters to make it the most accurate. So So that kind of continuum is exciting for us and we’ve had great response here at the show.

The interesting part about some of those when we analyzed a little bit more about the different solutions, it’s end to end, right?

I mean manages the entire lifecycle, which is key, which gives the ability for OEMs to collect, analyze, deploy, train and you know, monitor, optimize, and automate.

That’s right. This idea of a closed loop, it’s not like fire and forget. It’s not like you ship the vehicle and that’s it. The reality is that understanding it, tuning it, continuously optimizing it is going to be a requirement I think going forward.

And so we’re excited to be able to participate in that. And I think it’s it seems like the sky’s the limit. The the number of different applications people keep suggesting seem seem significant. Yeah.

20:43 Conclusion and Call to Action

So look, we’re so excited to have worked with you on this report. We highly recommend people take a look. It’s it’s like, is it forty, fifty pages? It’s some it’s very long report, but really documents the value proposition and the potential revenue impact.

There’s lots of dollars… That’s correct. Lots of numerical breakdown that we couldn’t possibly cover here. So encourage people to check that out. I’m so excited to have you on the the show.

You have such an incredible wealth of background over your seventeen years at Frost and Sullivan. So really, you for joining us.

Yeah. Thank you for having me. It was an absolute pleasure.

We had a a lot of fun doing this white paper with your team and you’re right. I mean, there are a lot of revenue impact that we couldn’t cover in this conversation. It’s a very quantitative report. I mean, looked at real world use cases. I mean, it’s time to go beyond pilot cases to to actual real time implementation for OEMs to see this value at an enterprise level. And that’s where we are getting at with this white paper, and we want everyone to read it.

It’s impactful. I strongly recommend it. Look for the show notes. Take a look at it, and thank you for coming on the show.

Oh, thanks for having me.

If you like what you’re seeing in today’s episode of The Garage, please like and subscribe to see more episodes like it. And we look forward to seeing you in another episode of The Garage very soon.