Automotive Outlook 2025: Ecosystem Pivots Around SDV

The automotive industry is deep in the throes of a massive shift to software-defined vehicle architectures, a multi-year effort that will change the way automotive chips are designed, where they are used, and how they are sourced.

Creating a new vehicle architecture is no small feat. OEMs need to figure out who to partner with and which aspects of their current architecture to include. This shift involves moving away from the mechanical, hardware-centric point of view that has evolved over the course of a century, as well as the more recent single-feature-driven electronic control units and domain-driven architectures. The future is all about developing software first, and then building the hardware to tightly integrate and support that software.

“This revolutionizes the way we think about vehicles, and it provides flexible and scalable solutions that otherwise would not have been possible,” said Rob Fisher, senior director of product management, automotive segment at Imagination Technologies. “It also allows us to integrate new technologies and capabilities quickly into the vehicle.”

This is especially attractive to automakers because an SDV approach can significantly reduce manufacturing costs. “SDV reduces the complexity of the vehicle from a physical point of view,” Fisher said. “It also gives the option of upgradability over the air. This is a concept driven from the consumer industry in mobile phones, digital TVs, and applications like that where you have over-the-air updates, keeping your device current and relevant. That is very relevant [in automotive] since vehicles are evolving very quickly these days.”

The shift to SDV will become increasingly evident at OEMs this year, even though the tiered automotive supplier structure remains firmly in place. OEMs still want to migrate to owning more of the supply chain, but they don’t have the ability to make it happen with their internal teams. That makes partnerships more important than ever.

Robert Schweiger, group director automotive solutions at Cadence, noted that a number of significant technological developments are happening in software, virtual platforms, and chiplets in the move to SDVs. “In particular, the software-defined vehicle triggers a chain reaction of things that need to happen in a car, ideally at the same time, to make the whole architecture scalable, software-defined, flexible, and secure. What is needed is a software framework that can orchestrate all of the applications within the vehicle and can manage multiple operating systems.”

Fig. 1: Software-defined vehicle. Source: Cadence

That framework in many respects resembles any PC-based environment. What it’s replacing is a distributed architecture that includes multiple electronic control units (ECUs) with lots of microcontrollers, all running individual pieces of software that execute a specific function.

“The software-defined vehicle, by definition, means I can define software-based functions, but I also can update the software over-the-air,” Schweiger said. “A distributed architecture is obviously not ideal to have a central software framework because it cannot be upgraded easily. What you need is a zonal architecture. The zonal architecture is a centralized architecture, with a very powerful central CPU on which the software stack is running. This makes it a lot easier to update software by having all the functions on a SOAFEE (sustainable open architecture for embedded edge) framework.”

Bruce Franklin, senior director of automotive at Micron Technology, views the shift toward zonal and centralized automotive architectures from traditional domain distributed architectures as one of the most significant trends in automotive. “McKinsey estimates that by 2030, the global share of vehicles with zonal architecture will reach around 18% and continue to grow,” Franklin said. “This transition is driven by the need for more efficient, scalable, and cost-effective designs. Today’s vehicles are increasingly becoming like data centers on wheels, containing around 100 million lines of code and projected to grow to 1 billion with the rise of AI. However, current automotive architectures are not equipped to handle this data surge, necessitating an evolution in both architecture and storage solutions.”

With existing domain approaches, systems are grouped by function within the vehicle, such as in-vehicle infotainment (IVI), connectivity, powertrain, and more. This used to be sustainable, but the number of sensors, cameras and electronics have multiplied exponentially in recent years. In some models there are as many as 150 ECUs, which adds complexity and requires more wiring, driving up the cost as well as the weight of vehicles.

“Zonal and centralized architectures simplify vehicle design by consolidating multiple electronic control units (ECUs) into fewer, more powerful computing platforms,” Franklin said. “This not only reduces the complexity and weight of vehicles, but also enhances their performance and reliability. Additionally, centralized architectures support the integration of advanced technologies such as AI and autonomous driving systems, paving the way for smarter and more intelligent vehicles.”

As such, vehicle architectures will transform over time, with some features going into vehicles sooner than others as each OEM determines its best path forward from either a distributed or domain-based vehicle architecture to a zonal architecture. “It’s an evolutionary process to replace more and more of those different domains and have it centralized,” Schweiger said.

How quickly things move forward depends on what type of OEM is working to adopt the zonal approach. A new electric vehicle OEM that is designing a car from scratch can start with a centralized architecture right away and build up the car on this architecture. But a traditional OEM with a lot of legacy and many different models needs to scale up. “They need to see how they can bring their next generation of a certain series within their brand into production as an evolutionary process,” he said. “They cannot redesign each and every car model that they have. They have to apply a more non-disruptive, evolutionary approach. So it varies from OEM to OEM.”

BMW, a traditional OEM, this year is expected to release its Neue Klasse model. Christoph Grote, BMW Group senior vice president electronics, publicly showed a picture of the wire harness of a previous BMW car compared to the wire harness of the Neue Klasse, which is greatly simplified due to the move to a zonal architecture. While this car already has a centralized compute architecture, for BMW’s traditional 3 series or 5 series it will be step by step as they get closer to this.

One of the benefits of the software-defined vehicle approach is software before hardware. “You can have the software applications ready by the time the hardware is ready,” said Ron DiGiuseppe, automotive IP segment manager at Synopsys. “Validating the software on each of these applications is a big workload for all the developers, the OEMs, or the Tier Ones — or even the second-tier suppliers developing software. The SDV approach really will speed up the industry, allowing all that software to be developed well ahead of the hardware. Software-first development will speed up the introduction of these new technologies tremendously, so the clear benefit of SDV is accelerating these systems’ introduction because of the very early software development. Software-first is a different approach from hardware-first, then retrofitting the software onto the hardware. The SDV approach says, develop the software first, then port it to the different hardware platforms.”

This is a very different skill set from what the OEMs traditionally have done, but the industry is embracing it. “It’s very critical, and has benefits,” DiGiuseppe said. “If you’re developing software first, rather than waiting for hardware, the more traditional approach is that you get shipped the hardware development board, and you develop software based on that hardware development board. But with software-defined development you could run that code, develop the code in the cloud, on servers. It’s much lower cost, and quicker than waiting for that traditional hardware board to do your software development. So it’s really providing a lot more flexibility. You could locate your software engineers anywhere, because they’re all doing the software development on the cloud, as opposed to a board-based hardware approach. This provides a lot of flexibility and scalability for the whole software development community.”

SDV also enables an advanced user experience, Imagination’s Fisher said. “Having all of your controls in the car, including the operation of the car, defined by software, allows you to very easily tailor that user experience to a very specific brand that you might want to project. Also, it enables you to combine a lot of functions in the car for greater driving assistance and safety. So what’s the actual impact of having a software-defined vehicle? It’s driving a centralization of processing resources in the vehicle, where previously you might have had functions distributed throughout the vehicle, very close to the sensors, or maybe separate discrete infotainment systems. With SDV, these resources are being centralized into fewer and higher-performance processors. There’s a definite tension here, though. OEMs are finding that their previous competencies of mechanical expertise are no longer appropriate when the car is predominantly driven by a software architecture. That’s a massive pressure on OEMs and Tier Ones, which are having to change and become software experts and technology experts.”

Another aspect of SDV, which NVIDIA CEO Jensen Huang touched on in his recent CES keynote, is that for most autonomous vehicles, the Orin chip/autonomous system from NVIDIA is already a reality and is what most companies are using. He also announced in his keynote the successor of Orin, called Thor, a very small and much more powerful device. This means that for autonomous driving a centralized architecture is already in place, while other functions are still running on a separate infotainment cluster.

“You can see how things are converging,” Cadence’s Schweiger said. “To make this possible, a scalable E/E architecture for the whole car is needed that allows you to scale it from an entry-level car to the super luxury car, and this can be done with a zonal architecture and a number of zonal controllers. And since this system is very complex and a lot of software is running on top, you cannot design the hardware first, then say to your thousands of software designs, ‘Okay, now go and develop the software that is supposed to run on top.’ This means you need a virtual platform, which is the shift left approach, to start early with the software development and simultaneously design hardware/software, do early testing of the software, and so forth.”

Chiplets drive virtual platforms
Huang’s comments also signal that the shift to zonal architectures is happening more quickly in some areas. That, in turn, makes a virtual platform essential, particularly for chiplets.

“As you move down from a scalable E/E architecture to a system and SoC reference architecture at the chip level, it also needs to be scalable,” Schweiger said. “How can we do this? Today in typical premium cars, we have an ADAS domain, maybe NVIDIA-based, and we have an infotainment domain, maybe Qualcomm-, Samsung- or Renesas-based. So we have those two domains, and those chips are the most advanced chips in a car in terms of process technology. Maybe the ADAS chip is 5nm, maybe the infotainment chip is 7nm or 5nm, depending on the vendor. So you have those two chips, but what you would like is a single, centralized compute unit that is capable of running ADAS and infotainment in parallel. Since the individual chips are already fairly complex, semiconductor vendors are trying to first build a two-die version of the chip in a single package as an intermediate step. Maybe this chip has additional interfaces, such as UCIe, so you create a new chip in an advanced package, for instance, at 3nm with those two die, since you have an ecosystem of chiplets available that allows you to mix and match different chiplets from different vendors into this massive chip. Virtual platforms allow you to visualize this.”

Dipti Vachani, senior vice president and general manager of Automotive Line of Business at Arm, also sees increasing adoption of virtual prototypes to transform the silicon and software development process. “Virtual prototypes are accelerating silicon and software development cycles, with companies able to develop and test software before the physical silicon is ready. The benefits are particularly relevant to the automotive industry, where the availability of virtual platforms is accelerating automotive development cycles by up to two years.”

In 2025, Arm expects more companies to launch their own virtual platforms as part of this ongoing transformation of the silicon and software development process. “These virtual platforms will work seamlessly, with Arm architecture offering ISA parity, ensuring uniformity in architecture in the cloud and at the edge,” Vachani said. “With ISA parity, the ecosystem can build their virtual prototype in the cloud and then seamlessly deploy at the edge. This saves significant time and costs, while giving developers more time to extract even greater performance from their software solutions.”

AI in vehicles on the rise
AI in vehicles is growing, as well, particularly in ADAS for enhancing camera or radar data. “This is happening right now,” observed Adiel Bahrouch, director of business development for silicon IP at Rambus. “Let’s say the camera is able to recognize patterns and speed labels and those kinds of things, but in order to recognize them you need to train the system. The training part is where you will find AI technologies to make that happen. For radar, there is some research around how to train that data, to learn from whatever history and information we have in place and predict what’s coming. The same is true for camera pattern recognition, and a lot of training is happening there. Then, when we combine those different technologies, I’m sure there will be also a lot of training there.”

This means premium AI-enhanced features will become an automotive industry standard. “Technology becomes more affordable over time, and the automotive sensor and computing sector is no different,” said Wayne Lyons, senior director of marketing, automotive segment at AMD. “Premium safety features such as blind spot detection used to be expensive and found only in high-end vehicles. With smart sensors and in-vehicle computing capabilities becoming both smarter and more affordable, we’ll see advanced AI-enable features, such as parking assist or autonomous assisted driving, reach the mass-market. For the AI-enabled cars of tomorrow, these will become standard features required in all vehicles.”

Further, end-to-end AI is expected to enhance automated driving systems with generative AI technology rapidly being adopted in end-to-end models that promise to address scalability barriers faced by traditional automated driving (AD) software architectures. “With end-to-end self-supervised learning, AD systems are more capable of generalizing to cope with previously unseen scenarios. This novel approach promises an effective way of enabling faster scaling of operational design domains, making it quicker and cheaper to deploy AD technology from the highway to urban areas,” Arm’s Vachini said.

AI also is set to enhance the user experience in vehicles by learning from driver behaviors, Bahrouch noted. “This will include your daily routine, your favorite places to go — all those kinds of things are collected. There are some privacy discussions to be had here, but it’s known that when we want to enhance the driving experience, we need to understand the profile of the driver to understand the mood, the patterns, the routine, to then start predicting the behavior of the driver, and giving suggestions. For example, if the battery of a vehicle is about to go low, and the car knows your daily routine and the routes you take, it can combine them to suggest a route to a charging station close to your favorite restaurant, and maybe because you have a team meeting around that time, if you’ve already prepared the tools that you need in order to take that meeting from your car, the car can put all the different things together and propose an agenda, including navigation, to make your journey as smooth as possible. So that charging time you need will be efficiently used to take you to your preferred coffee shop and have a meeting right after that in your car while the vehicle is being charged. And all of these types of AI services will be delivered audibly in a voice that sounds like a human assistant is talking to you. This is one area where a lot of OEMs are spending effort and energy to make it happen.”

Additional AI capabilities in the vehicle will equate to more hands-off driving, but more driver monitoring, too, Vachini said. “Progress in the harmonization of vehicle regulations for L2+ hands-off DCAS and L3 ALKS will accelerate wide deployment of these premium features worldwide. Leading automakers already are investing in equipping vehicles with the hardware necessary to up-sell these features through subscriptions throughout the lifetime of the vehicle.”

And to prevent driver misuse of driving automation systems, regulations and New Car Assessment Programs are focusing on increasingly sophisticated in-cabin monitoring systems like driver monitoring systems. In Europe, for example, EuroNCAP 2026’s new rating scheme will incentivize deeper integration of direct-sensing (e.g. camera-based) driver monitoring systems with ADAS/AD features to provide adequate vehicle responses to different levels of driver disengagement.

All of this adds new security concerns, as well. Automotive is very sensitive to hacks, which is why all automakers are taking security seriously. “All OEMs are working on their secret sauce as to how to protect the overall car, so that no one can take control remotely and do weird things with the car. That would ruin the credibility of an OEM,” Cadence’s Schweiger said. “This is why security is at the very top of OEM concerns, and people are building hardware-based security systems, not only software-based. It’s part of the SoC where this security IP goes in, and there’s also a lot done on top of it. You have not only one security IP, but rather a multi-layered security strategy protecting the network, protecting the chip, protecting the software, since there are all kind of security measures that you need to really protect the car.”

Conclusion
From connected, autonomous to electrified vehicles, the current state of the automotive ecosystem is all about enabling personal mobility of choice. Nand Kochhar, vice president of automotive and transportation strategy at Siemens Digital Industries Software, views all these technologies together, whether it’s coming from the amount of computing capability that continues to increase, the amount of electronics that continues to get introduced in the consumer products. All that is flowing into every industry, and more specifically, into the automotive industry.

“Software-defined products are the best example of this,” Kochhar said. “They’ve got everything, and it spoiled people in terms of the user interface and ease of use. People have come to expect that. A lot of the technologies underneath are the same or similar technologies coming into automotive, as well, whether it’s from the telecom side, the connectivity technologies, the amount of computing power needed in cars, the amount of access everyone wants in their vehicle, or just being able to do shopping. Ten years ago, when I was working in the industry, we always were debating whether or not we should put a modem in the car because this cost money. Now, cars are getting over-the-air updates, so what part of technology is getting touched on? I would ask, what’s not being touched on? Is it software? Is it electronics? Is it connectivity? Is it cloud computing?”

Bringing software-defined products to reality in the automotive industry is the most far-reaching development to look for this year. There are more ADAS features in new vehicles, and increasing levels of autonomy. “L4 is where robotaxis fit in, because they’re still constrained in that city, in that route, but they’re going without a driver,” Kochhar said. “What could be more exciting than that? These vehicles are delivering not only technology, but business value as well.”

Additionally, connected devices are making users’ lives easier. “They’re getting accustomed to the same things from their home devices,” Kocchar added. “They get connected homes, connected appliances, connected AI assistants. And underneath, from a technology standpoint, are all these AI-based technologies in every phase, in every part of what we do — not only from a product development perspective. Day-to-day, our lives are going to be touched by generative AI technologies. That is the exciting part. We don’t even know how far that could take us and all of where it could take us, from all the robotics and other AI-based solutions. It’s going to be exciting in 2025, 2026 as to how these things keep unfolding.”

Related Reading
How Software-Defined Vehicles Change Auto Chip Design
SDVs are booming, but designing hardware that can run software updates a decade in the future is a huge challenge.
Auto Ecosystem Begins Shift To Software-First
Carmakers and traditional tiered automotive suppliers are still grappling with how and when to move to software-first vehicle architectures.
Moving Software-Defined Vehicles Forward
Automotive OEMs must undergo a change of mindset and a willingness to embrace open standards for SDVs to become a reality.