Automotive Growing In 2023

Automotive has to be one of the most fascinating industries where semiconductors and the semiconductor ecosystem are making huge strides. From the evolution of increasingly autonomous vehicles, to more immersive driver and passenger comfort and infotainment experiences, along with additional safety-related features, it’s a rich development environment.

I recently had the opportunity to discuss the current automotive development trends with Robert Schweiger, group director, automotive solutions marketing at Cadence.

Here are excerpts from our conversation.

SE: Looking at the automotive industry overall today, what are you seeing as far as big picture trends?

Schweiger: In general, if you talk about the industry landscape, the semiconductor industry for automotive is still growing quite nicely, according to the forecast from Yole, and it really speaks toward leading edge technologies. Another issue that we see is that the supply chain disruption still exists, which is triggering a variety of activities within the automotive ecosystem.

SE: Where does the industry stand today when it comes to addressing the demand for increasingly autonomous vehicles?

Schweiger: There’s still a lack of an adequate SoC for autonomous driving. There are, for sure, chips like Nvidia or Qualcomm, but those chips are super expensive if you are not in mass production. If you want to have a scalable chip that allows you to go from a premium level car down to an entry level car, then these two chips as an example are possibly too expensive. That’s why there are a lot of activities where automotive ecosystem customers are looking at how they can scale this chip, and how they can create a chip at a price point that better fits their offering of various costs. This is a big issue, and we expect a lot of design activities around this, not necessarily just from the traditional semiconductor companies but also from other companies.

SE: How are the OEMs and Tier Ones responding to this challenge?

Schweiger: The OEMs and the Tier Ones would like to go into SoC design because they realize all the functionality is moving into a chip so they need to better understand what can be done as part of an SoC. That’s why they would like to go into chip design, but they still need to find their role, and what they want to do. For instance, they could get to the virtual prototyping stage, which would help them to create a specification, and then pass on the rest to a design house or to a semiconductor partner. They could also say, “Let’s not only do virtual prototyping, let’s also do RTL design,” and then the implementation is done by somebody else. This varies a bit from company to company. The issue for OEMs and Tier Ones is they have no experience in chip design, and they do not have the chip design teams, so some of them are ramping up teams. This is a challenge right now because designers are very relevant nowadays in the market, and you need to pay a lot of salary to get them on board. There’s also a shortage on designers, so that’s really tricky for them.

SE: Automotive continues to be a market that newcomers definitely want to be a part of. What are you seeing there?

Schweiger: The traditional semiconductor companies that are very experienced in SoC design coming from the mobile or consumer markets now realize, “Oh, automotive is growing quite nicely. Let’s get into this market.” There are many examples where you see activities from companies that have never done any automotive chips. One example that is public is Samsung, that has developed an automotive infotainment chip at 8nm, but there are many more companies. They know how to do complex SoC design, but what they do not know is how to deal with functional safety, as well as the typical automotive requirements centered around reliability, 15 year lifecycles and also quality. That means if you do the test of a chip at the end, they expect a zero PPM failure rate. That’s a problem for them. Then there are startup companies that are developing new chips so their problem is that they are small companies, and they want to know how they can bring a chip to the market as soon as possible, and augment their existing team with capabilities that they do not have. We’ve had various engagements with these companies, and we added certain team, SoC or functional safety capabilities, so they could develop an automotive SoC and get it certified for ISO 26262.

SE: What also appears to be happening is that the OEMs seem to have realized that they do not have to do ADAS Level 5 concurrently with electrification, that those could be parallel efforts, rolled out at different times. At what point did they realize those adjustments needed to be made?

Schweiger: If we look at the Nvidia and Qualcomm chips, these are super high performance chips that are water cooled, and they are water cooled for good reason. They consume a lot of power. That is also the reason why they are so expensive: because they are very complex. They are complex not only in number of transistors, but also in the way they need to be cooled, and so forth. That’s the problem for EVs: the power consumption. In one of Nvidia’s chips they had in production, which they provide a whole ECU with, and it’s already the most used prototyping environment for autonomous driving. The problem is that it consumes 500 watts. In an EV, you can imagine how far you can get with that. I have friends that say they switch off the heat in the winter to get something close to a decent range. That’s the reason I said earlier there is a lack of an adequate SoC for ADAS. It’s due to pricing and power consumption.

SE: What’s happening today with the automotive supply chain issues?

Schweiger: When it comes to the supply chain disruption, if you look at the foundry or at the process technologies, where is the actual shortage? The shortage is at mature nodes. It’s not at the 5nm or 7nm nodes. The high volume for components that are used all over the place, like voltage regulators, microcontrollers, and the like — these are the chips that are done at 65nm and above. This is the issue. And for advanced nodes, there is a lot of investment by the foundries so there is no shortage. If a customer says they would like to do this or that chip, and asking what we would recommend in terms of process technologies, we tell them right away if they go to mature nodes be prepared that if you have a low volume, you will run into trouble because of fab capacities. The investments are really happening predominantly in advanced nodes. There are some fabs coming online for mature nodes because of the supply chain shortage, which is another reason why automotive companies are trying to do their own chips. If you have a very generic microcontroller, you basically compete against anybody on the market that is building a washing machine or a small device or an automotive ECU. These are all standard components. But the minute you create your own chip where you have put your stamp, you are the only customer. In terms of volume, if you look at the semiconductor volume, then automotive compared to other markets is fairly small. With that volume, automotive companies could never compete against a mobile company. This is a whole different level but they are always at the bleeding edge. Maybe they do not necessarily interfere. In terms of pricing, the key thing is getting to a certain volume that you can charge a decent price. Sometimes there are chips, like a camera chip that is $45, is that price because they don’t have the volume. But in a certain segment, they are the market leader and this was the price point to get to a decent ROI. And people are willing to pay it. It’s a very complicated situation.

SE: What are some of the biggest challenges today in automotive technology?

Schweiger: In automotive today, there are application specific challenges that are tightly connected. There are also generic automotive challenges that go across the board for each and every application. What I mean is, zonal architecture will be the next big thing for OEMs. And the reason is you have these couple of chips [zonal controllers] in the central compute unit that have the capability to significantly reduce the ECUs. I.e., going from 100 ECUs down to, for example, 30 ECUs or 40 ECUs. You can imagine the cost savings and also the simplified network as a result. This approach has a whole range of advantages and the OEM will try anything to make this happen. And that’s actually why it will happen. But in order to put the zonal architecture in place, you need to have fulfilled some prerequisites, which is the in-vehicle networking bandwidth. There must be a high speed network. There is, for the inter-ECU communication/the ECU-to-ECU communication, the backbone will be Ethernet based. But the Ethernet that is currently in production is 1 Gigabit, and we need to have 10 Gigabit for this to work. The roadmap goes up to 25 Gigabit, so this is one thing that needs to be sorted out. And even though 10 Gigabit in production PHY is not to yet available, it will happen.

SE: What does this mean for in-vehicle computing performance?

Schweiger: Since the zonal controllers are multifunctional devices, they need to have quite significant computing performance. That means a zonal controller needs to eventually process radar data, camera data, audio data, LIDAR data, all the typical processing tasks, and also provide the advantage for ECU consolidation. Before they had one box, one function: one ECU, one function. Now they have one box many functions because of these zonal controllers, so computing performance is a big thing along with power consumption and anything that is related with it.

Beyond all of these topics, there are a number of other technical challenges facing the automotive ecosystem including hardware/software complexity, cross-application requirements, thermal issues, reliability, safety and security, as well as cloud computing and 5G, which is how cars will be connected to the outside world.

For another industry perspective, check out this blog from Synopsys, which also discusses some key technology drivers happening now.

In my next blog, I will share the expectations for the automotive industry from another industry leader.

Thanks for reading.