Why Auto Ecosystem Relationships Are Changing

The automotive industry is in the midst of rapid change on many fronts. OEMs are exploring new functions and features to add to their vehicles, including chiplets, electrification, autonomous features, as well as new vehicle architectures that will determine how vehicles are going to be designed from the foundation up. But all of this is dependent on the relationships between all of the ecosystem players involved. Semiconductor Engineering sat down with David Fritz, vice president of hybrid and virtual systems at Siemens EDA; Kamal Khouri, vice president and general manager for automotive at GlobalFoundries; and Frank Schirrmeister, vice president of solutions and business development at Arteris IP. What follows are excerpts of that discussion, which was held in front of a live audience at the recent AutoSens Conference in Detroit. To read part one of this discussion, click here. Part two is here.

SE: How do better relationships within the automotive ecosystem create value for the end consumer?

Khouri: As a contract manufacturer, it used to be that our customer would be a semiconductor fabless or fab-light semiconductor company. They’d build a chip. That chip would go into a Tier One. That Tier One then would build a module or ECU for an OEM. But that hindered the ability for the OEMs to advance technologies and break that chain down into a more interconnected ecosystem/supply chain to help them achieve the features and capabilities. However, there’s also an economic model behind that, and we’ve looked at it very carefully. The concept of margin stacking is the way our capitalist society works, where I build something, I pass it on to you, you put your profit margin on it, you pass it on to someone else. There are components of things that are valuable and that each company in the supply chain is adding. But there are components that are not valuable, and are just getting passed on. And the end consumer who goes and buys a car ends up paying for that. So there’s an optimization of the supply chain from an economic perspective, as well as the technological perspective.

Schirrmeister: For the consumer, it boils down to confidence in safe technology and in convenience — the consumer’s experience. Interestingly enough, you actually need more than one OEM. There’s the car OEM, but if you look at some of the OEM presentations, they look at the mobile OEM, the health OEM, the neighboring OEMs, and they all need to interoperate. I want my phone to work with the car as seamlessly as possible when I enter it. Ideally, by the time I’m five feet away from the car, it should know everything. My car should be synchronized. That’s the convenience bit. Assuming we figure this out as a design chain, which I’m confident we will, that results in confidence that things are safe. The foundry technology needs to be ISO 26262,  and the semiconductor IP and chip design tools need to be TCL-qualified. That’s the confidence bit. Then there’s a convenience bit that the users experience, which they and are used to with their phones. That can be replicated, and is not interrupted when they enter the car.

Fritz: I see it slightly differently from that. What’s important to the end user, the consumer, is that we actually deliver on our promises. If you look at the hype curve for autonomous vehicles from four or five years ago, it hasn’t changed a whole lot. The consumers are saying, ‘When are we ever really going to see an autonomous vehicle? I see a couple of driving around San Jose or San Francisco, but I’m not ready for that yet. And it was supposed to be here by now. Elon said we’d have them in 2024.’

Khouri: Four years ago, you couldn’t be a naysayer. People looked down at you.

Fritz: We need to set better expectations. Those of us  in technology know you could take the worst case and then you double that, and that’s about when it really happens. Autonomous vehicles are that way, as well. They were overhyped, over-simplified, and now we’re understanding the complexity behind it. We know it’s going to take new technologies like chiplets, like high-end modeling, like high-speed interconnects that do quality of service, all those sorts of things. We need to be down at 3nm, for example, to make it all work together. We’re not there. People see announcements with GlobalFoundries by an OEM looking at a billion-dollar relationship between this OEM and that AI company. All those things happen, but we’re not really seeing a lot of benefits from that. The important thing is, let’s be honest and say this is when it’s going to happen, and then we make it happen. We don’t let it happen with the understanding that the OEMs have a really tough job, because there tends to be a lot of DNA that is resistant to change. They still must deliver the current model years, which is a big enough job. And yet, they need to find when in their roadmap they’re going to intersect with something that’s Level 4, Level 5. How does that happen? Understanding all of that when we’re engaging with those customers, that’s part of what we do. We can say, ‘Here’s how you make that transition. Here’s how we work together as a partner to help you make that change in your organization, and skill sets, and culture, and everything else.

SE: If an OEM and an AI company partner together and develop a chiplet, does that means it will be proprietary and you will be producing less volume because it’s proprietary to them?

Fritz: Part of the reason or justification behind chiplets is you built that one AI accelerator chiplet and it goes to many different sources to help get your volume up.

Schirrmeister: The idea here is that you can actually apply that particular accelerator to more people, and then because of the composability at the OEM level, they can differentiate and provide their unique value at the OEM level together with the software — without having to do the $100 million investment and do a specific proprietary custom SoC run. That’s the intent. Now, there are steps to get there, and we’re not quite there yet. We’re still figuring out the right protocols. Is it Bunch of Wires or UCIe? We’re figuring out the protocol stack. Do I use AXI over UCIe? Do I use CHI over AFS or UCIe? There are very interesting challenges that we, as an ecosystem, need to figure out. The fascinating part is, it goes through everywhere. It comes from the OEM functionally, it goes down through the Tier 1 integration aspects through the IP. Do we support the right protocols? Is it cost effective from a manufacturing perspective? There’s job security for at least a decade or two as we figure it all out.

SE: What is a Tier 0.5, and what are the ramifications?

Schirrmeister: The GSA and McKinsey came up with the term. When I asked them the same question, they said it is a new, very tight cooperation between the OEM and the Tier 1, in which they have what used to be just a requirement coming down, and then coming back up in waterfall model has now become agile. People are really interacting very closely on that. That’s the implication of the Tier 0.5, and it’s an adjustment process. There might be companies spun out that become these Tier 0.5s, which are very close between the OEM and the Tier 1.

Fritz: Aptiv and Magna are two examples of Tier 0.5, where they want to do the whole integration of the whole vehicle. You slap your logo on the chassis, and off it goes. That’s the goal.

Schirrmeister: Get the popcorn ready and watch the design chain change, and how the power dependencies change over time. That’ll be the popcorn consumer for the next 5 years, figuring out how this all falls out. Some will win, some will lose. It’s an interesting process over the next 5 to 10 years.

SE: When do you see chiplets coming into automotive?

Schirrmeister: While there are long timelines, we see the discussion at our level as the IP provider. We see that going on today because people are asking which protocols they need to have three years from now so their chiplet will be able to talk to other chiplets, because they need a certain amount of time to implement it. To this point, Arm gave an excellent presentation at the GSA forum in Europe a couple of months back talking about how, as an IP provider, they need to have their deliverables ready at least five years ahead of the next product cycle. You have to think in new ways. You will get certain things, you will make mistakes that need to be corrected. As far as when chiplets will become imminent and become solutions, it’s probably a three-to-five-year range, but the discussions of the problems and solutions need to happen today.

Fritz: Chiplets don’t apply to all parts of the automobile. It’s really the high-end compute central compute type things — IVI, ADAS, AD, that sort of thing. If you want just a gateway, or if you just want an ECU to control steering or transmission, it doesn’t make a lot of sense there.