An Automotive Value Chain In Flux

When companies view suppliers from inside their specialized niches, it is tempting to imagine the business world will continue as-is, with just minimal improvements each year. But in the automotive value chain, this no longer holds. The rapid pace of innovation around intelligent systems in cars is disrupting the business flow. Back in simpler times, semiconductor companies would work with Tier-1 companies like Bosch, Denso, Delphi, and Continental. The Tier-1s would provide requirements to the semis, and the semis would build chips to those requirements. In many cases, this might be in an ASIC flow – the Tier-1 would provide requirements and the semi would implement and manage development and manufacturing. The overall chip architecture could be “owned” by either the semi vendor or Tier-1 depending on the business arrangement.

Now that electronics account for 40% or more of the total cost of a vehicle, Tier-1s and OEMs are paying much closer attention to the ownership of automotive SoC architectures. Nobody wants to see their contribution shifting increasingly towards just assembly, leaving them to become a Foxconn of cars. That’s stirring up competition in the supply chain and a lot more interest in developing and owning high-value components in the system, especially chips and system architectures.

What are companies doing to achieve this? They need to build world-class electronic design teams, a process that’s already underway. For example, Bosch already has a well-known strength in design around sensing; they have also been building up capabilities in higher-level SoC architecture and development. GM Cruise on the OEM side is aiming for full autonomy solutions, which takes a lot of design talent. We’ve already seen how cash-flush hyperscalers like Amazon and Google have been able to boot chip design capability from scratch. I don’t think the automotive teams are quite as far along, but they’re on the same path.

As with the hyperscalers, automotive teams have no interest in reinventing anything that isn’t essential to their singular goal of capturing value in the automotive system. They use industry-standard commercial IPs for most of the chip and build their own IP subsystems for highly specialized and differentiating functions such as AI accelerators. Otherwise, all their differentiation is going to go into how they choose to architect the SoC and optimize its system level differentiating metrics (performance, bandwidth, guaranteed latency, functional safety level, power consumption, cost, etc.). Given that most of the IP licensed by a design team is also licensed by their competitors, how that IP is assembled in a unique system-on-chip (SoC) architecture is critical to create value through product differentiation. The network-on-chip (NoC) interconnect is key to product differentiation because it contains the “knobs and levers” that architects use to create the SoC’s architecture and optimize it for power, performance and cost metrics.

Because our NoC technology is the foundation of many automotive SoC leaders’ most complex chips, we get to see a lot of architectures. These designs are diverse at the block diagram level. We are used to seeing block diagrams for application processors or digital baseband modems for mobile phones – they all look similar. High-end automotive block diagrams are wildly different, some purely vision-focused, some with a big emphasis on fusion between vision, LIDAR, radar, and other sensors. Even in lower-tier architectures, there’s a lot more diversity than we see in mobile phone and consumer electronics chips.

This trend isn’t a short-term phenomenon. Innovation is spurring not only a plethora of new chip architectures but also changes in how system software is developed. System architects are thinking ahead, designing platforms to scale across multiple generations and price points. For Tier-1s and OEMs, the software is costly. They must be able to leverage as much as possible, from premium to value-priced features and across generations, to support a compelling long-term roadmap in the hyper-competitive automotive market. Just consider features we already see, not only in safety and semi-autonomy but also in the cabin – driver attention monitoring, sterile cockpits, road noise cancellation, personalized audio bubbles, and more.

Where do the chip makers sit in all of this? There’s a wide span of possibilities. On one end, you have a company like NXP offering an open solution, reference systems with an SDK – no need to get involved in the complexities of chip design. I’m sure there will continue to be a strong market for solutions like that – the Tier-1s and OEMs don’t want to take over ownership of every chip in their systems. On the other end, you have companies like Mobileye who span up into the car with hardware and software. And they want access to driver data, navigation and other types of data, which will go to their cloud to help them refine and expand the customer experience they can provide.

There are going to be a lot of struggles over who owns what in this value chain. The pendulum will continue to swing between the familiar but less certain horizontal split in functions. I predict more verticalization as automotive and semiconductor suppliers find more opportunities for innovation in the system as a whole rather than just in those individual functions. Semiconductor vendors will need to stay on their toes, continuing to reinforce direct relationships with the automakers – GM, BMW, Mercedes and others. They’ll also need to bring on more staff with direct expertise in these markets.

You can watch my video interview with Ed on this topic here.