Automakers Grapple With Fundamental Tech Changes

Automotive OEMs are wrestling with a stack of changes that affect every part of their business and technology, from threats of tariffs and shifting geopolitical alliances, to new vehicle architectures, tighter market windows, and a fundamental reordering of relationships and priorities between OEMs and their suppliers.

There is no consistency to these developments or a best path to solving them. Some OEMs are attempting to bring everything in-house to more tightly integrate chips, software, services. Others continue to engage tiered suppliers to help deliver their products. And still others are on the hunt for new relationships to supplement or supersede the old ones, tearing down silos where it makes sense and rethinking go-to-market strategies.

Controlling the vertical is the goal, but how best to do that is incredibly complex. There is more guesswork as to what customers are and will be looking for, and much more focus on chips, IP, and software — areas where many carmakers are relative newcomers with a lot of legacy baggage. At the same time, core relationships are shifting as new technology is implemented. A Tier One provider may not know what their OEM customer is trying to do at the device level, or how chips will be used throughout a vehicle. They may know the OEM is adding some level of ADAS functionality, but the depth of that knowledge typically can vary greatly from one OEM to the next. Likewise, the OEM is unlikely to know details about which sensor arrays they need, or which specific camera resolutions or frame rates will be used.

Some OEMs are managing this transition better than others, and some are approaching it more aggressively than others. After multiple architectural shifts in short order, it’s not surprising that established carmakers are taking a more risk-averse approach to entering a domain, mapping out plans to tighten control over the vertical while maintaining higher-integrity customer-supplier relationships in the meantime. As a result, there still may be an OEM, Tier One and Tier Two in the loop with a knowledge base that demands close partnerships today, but plans are being developed that will change those relationships in the future.

“There’s a huge ecosystem challenge,” observed Andrew Johnston, director of quality, functional safety, and cybersecurity at Imagination Technologies. “This is essentially modern science and modern engineering, and philosophically that includes big data, machine learning — and training, inference, etc. It’s a very modern dynamic. Then, there is a lot of modern, complex technology, and it is very software-intensive. Traditionally, the automotive sector doesn’t do software very well. It’s still on a learning curve, even for traditional control algorithms with discrete and simple functions, to develop high quality software. Standards like ISO 26262 have helped deliver quality into these reasonably conventional control systems, but now we have to extrapolate that philosophy into more complex control systems. At the same time, the technology is inherently complex to deliver, whether it is the transducer at the front end, or a camera or lidar or a radar sort of transducer. It’s the same for the signal processing chain — the fundamental computational processing, whether it’s CPU, GPU or a combination of both, or NNAs and DSPs. That includes all these interesting things, and they’re all complex.”

Others point to similar developments. “All of a sudden, the OEMs realize, ‘Oh, I can eventually create my own chiplet as a secret sauce to differentiate from the rest of our competitors,'” said Robert Schweiger, group director automotive solutions at Cadence. “The Tier One can also say, ‘I can do one or two of these chiplets, because we believe we have a lot of expertise in this domain, and we should probably create a chiplet.’ You do not need to create a full chip, because you not only disaggregate the SoC functions, but you also divide the whole monolithic SoC into pieces from a cost point of view. But the cost is lower if you design one chiplet, rather than doing a fully integrated and packaged SoC, so the burden to do the chiplet is lower. And in particular, you do not need to manufacture at 3nm for each and every function. There are other functions that do not scale. PHY is a perfect technology. It does not scale very well. Maybe it could be in 28nm.”

Much of this is based on the OEM’s expertise. “For some companies that have a lot of PHY expertise, they can use a very mature technology node, which is a lot cheaper, has higher yield, all these benefits that you get from that,” Schweiger said. “Then, only the guys doing the AI processing part need to go to the bleeding edge, because they need the highest performance, and a PCIe PHY interface is not available in 60nm or in 7nm. It’s only available in 5, 3, or 4nm. That means if you want to have the fastest UCIe interface, and one of the protocols that UCIe supports is PCIe, that’s why you need to go to the most advanced process nodes — to have all these interfaces in the highest-performance version available. The interesting question will be who is providing which chiplets within the automotive supply chain and who is integrating everything.”

The integration challenge of making sure all of these systems work seamlessly together, alongside a huge complex software stack, is extraordinarily complex. It even has its own ecosystem. “The OEMs suddenly must bring all this together somehow and make things work,” said Imagination’s Johnston. “That’s hard, in the sense that just making it work is hard. But then getting it to be high-integrity, safe, secure, reliable, available — that’s orders of magnitude more difficult. So everyone globally is on a learning curve right now. At the semiconductor level we’re trying to predict where that market will go, and really understand what customers need, whether that’s Tier Twos, Tier Ones, or OEMs, and try to join some dots and make a product that best fits for high market success and customer satisfaction. It’s really challenging.”

Underlying at least some of this are the traditional electrical/electronic (E/E) architectures, which are in the process of being replaced by newer approaches.

“The history of ECUs started back in the 1970s as a single ECU in a car,” said Nand Kochhar, vice president of automotive and transportation strategy at Siemens Digital Industries Software. “The goal was to control the emissions at that time. Then it started to grow up. We brought in the brake functionality, we brought in safety features, airbags, and the ECUs kept multiplying. The hardware and the software was coming from suppliers into the OEMs, which built on that relationship. By the 2000s, some luxury cars already had started to introduce Level 1 and Level 2 ADAS features, and there were dedicated ECUs for that. I was a safety chief at Ford, and I would never let anyone touch the safety control modules. I would tell the infotainment guys, ‘You can do whatever, but this is what gives the airbag when needed, or seat belts when needed, and nobody else is messing around with it because every function is dedicated to that.’ Now, when you have 150 ECUs in a luxury car, combining all this starts to become impossible to manage from a complexity standpoint.”

Having that many ECUs is also inefficient. “The amount of wiring needed, which function needs to go where, and managing all of that — that’s what evolved to the next generation of architecture, which a lot of companies are in the process of transitioning into,” said Kochhar. “Automotive companies starting from scratch don’t have all the legacy and the history, and have been able to quickly jump to domain controllers and central compute units as an architecture, which forms the basis of the vehicle. Every company decided differently. It comes down to assuring the hardware and software ask is being met when all these functions are called out.”

This evolution is changing the fundamental relationships between OEMs and tier suppliers.

“There’s no one answer, and no clear answer,” Kochhar said. “When the Continentals and Bosch’s bundle the hardware and the software, the business models will get charged with that bundle. But now you can imagine 7 or 10 different companies bringing integration across those. Everybody’s trying to protect their IP and keep it separate, which is another nightmare for OEMs. Even if you start to work out the coordination, given the time it takes across the projects, it’s not a surprise to hear in the news that Company X’s launch got delayed, while Company Y has these many recalls. All of these situations contributed to the reasons why they have switched to this model of domain controllers, where they are working with the same suppliers, but now with a new approach. Not every OEM will decide to develop their own CPUs, but they’re very clear on the chip requirements, and the chip developer will develop that. On the level above an ECU, they’ll work with that supplier. But at a vehicle level, OEMs will control that. So now the transaction and the business model are changing. What do they get charged for? Previously, hardware was being sold, but they included software for free, so the price was baked in. Now there are different roles and responsibilities. That’s the transformation industry is going through.”

Vehicle sales slowing, but EVs hold promise
How quickly this organizational transformation happens is not entirely clear due to a global slowdown in vehicle sales. “All the technologies that enable ADAS, autonomous driving, ADAS connectivity, and so forth, are related to vehicle [sales] growth,” said Adiel Bahrouch, director of business development for silicon IP at Rambus. “We are facing a demand-constrained market, which is putting pressure on prices, margins, R&D efforts, and so forth. And we are seeing announcements from some OEMs reducing their design team or getting rid of certain divisions. But at the same time, we see that electric vehicle sales are on the rise globally, and the expectation is this will continue.”

On a regional basis, China is leading in electric vehicle adoption, followed by Europe, with the U.S. lagging in third place. “Ten percent of the U.S. market is electric vehicle-focused, and the rest is still the traditional conventional vehicles,” Bahrouch said. “There are projections where U.S. and Europe are expected to catch up by 2030 or 2035. So within 10 years, more or less, the adoption rate [in the U.S. and Europe] will rise to 60% to 80% — all driven by the zero emission regulations.”

Electric vehicles are important for semiconductor growth. Hybrid and full electric vehicles contain more than twice the semiconductor content by value compared to internal combustion engine vehicles, according to numerous studies.

“This is good, because it means the electric vehicle market will support semiconductor growth,” Bahrouch said. “Connected to this are battery management systems to monitor the health of the battery, and the state of charge to optimize the charging and discharging to keep the battery at a good state of health and protect the vehicle from overcharging or overheating, which is also safety-related. Those systems are unique for electric vehicles, along with motor control systems to control the flow of electricity between the battery and motor in order to maximize a vehicle’s range, which is one of the important criteria for OEMs to differentiate their vehicles from others.”

Partnerships more critical than ever
No one company can possibly do all of this alone, so OEMs continue to maintain strategic and symbiotic relationships within their ecosystems. “If you go back to September 2024, we saw the perfect example of this,” said Michael Munsey, vice president of semiconductor and electronics at Siemens Digital Industries Software. “It’s like the intersection of where automotive and semiconductors come together. For example, Analog Devices, an established company that has an automotive division and develops ECUs, realizes off-the-shelf ECUs are declining. Also being a semiconductor company, they don’t want to necessarily invest a lot in building new fabs, because not everybody has $19 billion lying around. Instead, they signed an MOU with Tata Group, which just got a big grant from the Indian government to build fabs. So now ADI has access to next-generation fabs. Also, Tata Motors doesn’t want to take the step yet to develop a semiconductor design group to do ECUs, but they still need ECUs. Guess who they’re going to have the ECUs custom designed for them? Analog Devices. And where’s Analog Device going to build those? In the fabs that Tata is building. So you’re going see these symbiotic relationships happen. If you’re a Tesla and you’re a newer company to the market, you probably could invest in building up the infrastructure that you need internally. But there are other ways to be creative about this as you look at the supply chain and start building more of these relationships between companies to be able to get this technology delivered.”

Rambus’ Bahrouch agrees, noting that he sees OEMs actively looking for partners. “The OEM used to be an integrator, where they would specify their needs, the Tier Ones’ support, and provide all the systems needed for the vehicle. Then the OEM would integrate all those systems into the car. But OEMs had no clue about the content or what technologies were used. This has changed. You now see OEMs changing their role in this value chain. Some of them are starting to build their own in-house software department’s capabilities in order to control, for instance, the value of that software and to differentiate their brand by means of that software and all the services they will be providing. The one who owns the data, or who controls the data or the services, owns the business model. Now you can charge money for those database services. Other OEMs are even bypassing Tier Ones and Tier Twos and starting to build an SoC themselves. That’s another level. So not only software, but hardware, as well, and that means their business model is changing.”

Change also is underway at the Tier Three level, which is where semiconductor, IP and EDA tool providers sit. “What we see in our industry is that nowadays we also have direct communication with OEMs to understand their roadmap and their semiconductor strategy, and make sure that our hardware silicon IP matches their future requirements for next-generation systems-on-chip,” he said. “So not next-generation, but next-next-generation, like five years from now, to make sure that when the need is there, our roadmap supports the vision of the OEM. This is new. It’s changed in the past two, three years.”

As relationships change throughout the automotive ecosystem, roles and responsibilities will have to follow, breaking down functional silos that have been in place inside automotive companies for decades.

“The people who are really going to have the power now — the traditional OEMs that are building new architectures — are building vehicles to sustain that software and all those things and don’t deal with the legacy,” said Marc Serughetti, vice president for system software at Synopsys. “This is a big challenge for the existing OEM. The challenge comes down to how they want to deal with the transformation they have to go through now to stay competitive, and set a path forward that makes them relevant. How do they keep track of all the legacy that they have? Today we see they are having major issues in that transition because they are trying somehow to bring legacy and new things together, which makes the problem much harder than looking for new things and a blank sheet. It’s an integration problem at the end of the day. It’s about how to bring the new stuff alongside the old, and validate all of this together.”

Moving to SDV
The future direction is a software-driven vehicle architecture, but the migration from a hardware-driven architecture is no easy task.

“The OEMs are grappling with it, and have quite a hard time on that side,” Serughetti said. “It’s a pendulum swinging, still. Two years ago, you had OEMs saying, ‘I’m going to do everything. I’m going to do semiconductors. I’m going to do software. I’m going to do X, Y and Z. I need to own the software. I need to own this.’ What we’re seeing now is maybe that pendulum swinging back a little bit the other way as they realize, ‘How am I going to do this? What needs to change?’ It’s a digital transformation that needs to happen. It just doesn’t happen in one day. You need SoCs that are more powerful, with more capabilities. You need a software development infrastructure that is in place. And while there’s a lot of technology to this, there’s also the methodology aspect. There is a people and organizational aspect. That needs to evolve, as well. That transformation cannot just happen. It’s not just a technology transformation. It is an organizational and methodology transformation. How do you start integrating the tooling that goes with that? There may be more enterprise software. Also, we’re still in an embedded world within the car. How do you start looking at the embedded aspect and the enterprise software development aspect, and bring those things together? There are things that have been solved in other industries. You just have to figure out how to apply them to this specific domain of the car, an embedded system that’s distributed, that has safety, security requirement, all of this together. This is a mindset that must change as the evolution occurs. It’s how the industry transforms. Standards are one way this happens, but there’s also collaboration. Previously, the automotive ecosystem was very siloed. Now, there are questions around where the OEMs should differentiate, and where they should collaborate.”

Bringing new tech into the vehicle
One of the biggest challenges for automotive OEMs is how to infuse new technologies into a well-established flow while also navigating the necessary partnerships to move the entire company forward with technology. All of this must happen while still heeding customer demands, and it all has to be delivered in a timely manner.

“Time-to-market is key because all those systems are becoming very complex, software is becoming very complex, hardware is becoming very complex,” observed Rambus’ Bahrouch. “Redundancy is needed. There are safety concerns and security concerns. Additional safety and security technologies are needed. At the same time, we have a customer that is expecting a faster application lifecycle, meaning the lifecycle of the vehicle, new models, new technologies. The objective of having software-defined vehicles is having the capability to add features to a car through software over-the-air, on-demand. So if a customer wants to have a feature, he just adds it through an app, pays for it for a month, then disables it if he doesn’t need it, because it’s all there. The hardware is there, so the foundation is there. The software can be added. Then the customer can enhance his driving experience. But in order to do that, time-to-market is becoming very key.”

Some of this already is happening. Bahrouch pointed to SoC design as an example. “SoC development could easily take three to four years, and at least two to three years. Then this chip development, being part of an ECU, is yet another year. Then we have some safety and security testing and compliance, after which this ECU will be integrated into your car. This means the next model will easily take five to seven years. That needs to change. The expectation of customers is now different. That development time needs to be reduced to three or four years, max. So the chip and software development must be two years faster. That’s also one of the reasons why you see software nowadays. The software development doesn’t wait till the hardware is ready. They are leveraging digital twins. They are leveraging hardware models, cloud capabilities, and so forth, to start software development while the hardware is not even there. Once the hardware is there, the expectation is the software and the hardware work smoothly, seamlessly, and all of them shift left. And additional technologies are supporting this faster time to market and trying to support or meet the expectations of the customer. If one OEM decides not to do this, and another one decides to do this, whether it’s a Chinese OEM or a European OEM, that would be a key differentiation for that OEM to be able to support and manage and meet customer requirements in terms of features, updates, deployment, release, and so forth. This is why time-to-market with reliable, robust systems for safety and security is a big issue, and a big concern.”

The big challenge for the automotive ecosystem is staying on top of all these things, from the global industry dynamics that impact supply chains, to how that affects vehicles getting into production.

“We have to be keeping a close eye on these things,” said Siemens’ Kochhar. “The good news is we all have some experiences from past events to draw on — for example, the tsunami in Japan many years ago, when we could not get the material that goes into developing black paint, and we had to stop the production because we couldn’t produce anything with that color. That was a disruption of one nature. Then we went into the pandemic disruption, and how that has evolved, bringing a lot of manufacturing back to the U.S., bringing a lot of investments back. Which factory is planned, and how is it going to grow? Which battery technology is going to come home? Which one is going to switch to solid-state technologies versus lithium ion? How are we going to handle battery recycling? These are all related to the business challenges that OEMs face, and which they must keep in front of them.”

Related Reading
Automotive Outlook 2025: Ecosystem Pivots Around SDV
OEMs and suppliers are beginning to move in lock-step, linking software design with chip development to speed time to market, reduce costs, and help future-proof vehicles.
Automotive OEMs Face Multiple Technology Adoption Challenges
The path to fully autonomous vehicles may be clear in concept, but fully realizing that development environment is another story.