Chiplets in the Automotive Industry: The Future Has Only Just Begun

Mr. Singer, how important are chiplets for E/E architectures in the cars of tomorrow?

Stefan Singer: Chiplets offer an impressive combination of flexibility, scalability, and high performance. Unlike traditional monolithic system-on-chip (SoC) solutions, chiplets enable different semiconductor functions to be combined in a modular fashion. This allows manufacturers to quickly and efficiently customize vehicle models and their equipment. I would particularly like to emphasize that chiplets enable significantly higher bandwidths between components. This offers enormous advantages, especially in demanding automotive applications such as advanced driver assistance systems (ADAS). 

Chiplets also promise better resource utilization, for example through a shared cooling system and power supply. Another advantage is the ability to implement very large and complex functions that would no longer be integrable in a monolithic design. In addition, OEMs and suppliers can respond flexibly to changing and special requirements – new chiplets can be added or existing ones replaced without having to initiate a complete hardware redesign each time.
 

What about the cost efficiency of chiplets?

Stefan Singer: Here, I would like to correct a widespread misconception: chiplets are not automatically cheaper than traditional SoCs. Although cost efficiency can be achieved in the long term through flexible scaling and improved manufacturing processes, it should not be forgotten that chiplets require complex and therefore relatively expensive packaging techniques. The actual cost advantages only arise when scaling effects take effect and integration costs can be reduced. In this respect, the long-term effect of potential cost savings still needs to be carefully weighed up.
 

Japanese manufacturer Honda has announced that it will use chiplets in a production vehicle for the first time as early as 2028. Is this the starting signal for widespread adoption of the technology?

Stefan Singer: Honda's announcement definitely represents a milestone – after all, to my knowledge, this is the first confirmed series production project in the automotive sector. Nevertheless, we will not see widespread market penetration immediately, but rather a gradual spread over several years. Experience has shown that the automotive industry needs time, especially when introducing technologies that bring about fundamental changes. That's why I don't expect chiplets to be widely adopted before 2030.
 

A frequently mentioned desire within the industry is for a chiplet marketplace that allows components from different manufacturers to be combined. How realistic is this vision?

Stefan Singer: Technically and conceptually, we are still in the early stages here. Initial evaluations of combining chiplets from different manufacturers are available. However, practical experience shows that many questions, such as those relating to common standards, interfaces, or liability issues, remain unresolved. The Universal Chiplet Interconnect Express (UCIe) is an open industry standard that currently only defines the basic communication level. It will be some time before we have a functioning marketplace that could truly enable plug-and-play. I would like to address one issue in this context: in order to achieve a global understanding of the standards, numerous players with different interests must first be coordinated.
 

What challenges must be overcome when integrating chiplets into future E/E architectures in vehicles?

Stefan Singer: Integration brings challenges that should not be underestimated. These include compatibility issues between chiplets from different manufacturers. When integrating various vehicle functions, questions arise regarding a common software strategy, clock synchronization of the various chiplets, and risk management. In addition, chiplet-based systems require more complex testing and validation processes.
 

How important are chiplets for the software-defined vehicle (SDV) of tomorrow? Is it impossible to do without chiplets?

Stefan Singer: No, not necessarily. The key to SDV lies primarily in the software architecture. Of course, chiplets can help increase the flexibility and scalability of vehicles – but at the end of the day, software remains the decisive factor. However, it is noticeable that the automotive industry is focusing very strongly, perhaps too strongly, on hardware when it comes to SDV. While this is correct and necessary, a well-thought-out software strategy is even more crucial in order to leverage the full potential of software-defined vehicles.
 

Finally, let's take a look into the future: What medium- and long-term developments do you expect to see in chiplet technology in the automotive sector?

Stefan Singer: Chiplets will undoubtedly play a central role, especially for performance-intensive and complex automotive applications. I am convinced that this technology will come in the long term – albeit still associated with a number of challenges and learning processes. A broad, globally coordinated standard would be desirable to enable faster progress. I firmly believe that a significant number of applications and perhaps even standards will be achievable by the end of this decade. However, the path to get there is challenging and requires the cooperation of many players on a global level.