Much of the automotive industry has begun repositioning and retrenching over the past few months, pushing back the projected rollout for fully autonomous vehicles and changing direction on power sources and technology used in the next-generation of electric vehicles.
Taken together, these shifts mark a significant departure for traditional automakers, which find themselves playing catch-up to companies like Tesla, Waymo and NIO. Determined not to be left behind, they have begun zeroing in on ways to close existing gaps. Among them:Development of new fuel sources for electrification, including hydrogen fuel cells and plug-in hydrogen fuel cell hybrids. Utilization of new sensors and other low-power technology. Development of solid-state batteries. Shifting focus to more in-vehicle processing. Deployment of fundamental changes to business models.
All of these developments require a big increase in semiconductor content, and that trend is expected to continue as vehicles are increasingly electrified to the point where they become, in effect, smart phones on wheels. In a just-released report, KPMG estimated these changes could boost the automotive semiconductor market from roughly $40 billion today to between $150 billion and $200 billion over the next two decades, and those numbers do not include V2X infrastructure or increases in tooling and manufacturing equipment.
The most recent research shows electric motors will overtake internal combustion engines within a decade, although numbers may vary by region. Alongside that changeover will be an increasing number of chips to control and monitor what goes on inside a vehicle, as well as how the vehicle interacts with the outside world. There also will be changes to how many of those vehicles are powered, how that power will be optimized, distributed and stored in a vehicle, and how and where data generated by an increasing number of sensors will be processed.
“Traditional companies are now shifting their focus to software,” said Burkhard Huhnke, vice president of automotive at Synopsys. “Tesla made a statement several years ago that the best teams are not being used for electric vehicles. This will happen now.”
Electrification of vehicles is a given for competing in automotive over the next couple decades. Whether those vehicles use an auxiliary engine, fuel cell, or run 100% on battery power, the simplicity, responsiveness and reduced emissions from an electric vehicle are compelling. An electric motor has roughly 200 parts, compared to about 2,000 for internal combustion engines, which helps to significantly cut costs and improve reliability. At this point, virtually every carmaker is increasing electronic content in the vehicle, with many opting for electric motors as the primary powertrain.
The challenge so far has been to balance the benefits of electrification with ease of use. One option increasingly cited by big carmakers — excluding Tesla, NIO and others, so far — is hydrogen, which they claim will become competitive with other fuel sources over the next decade. They contend that it will overcome some of the biggest problems with electric vehicles today, such as how far they can travel on a charge, how long it takes to charge them, and how to develop a charging infrastructure in less populous areas.
Charging infrastructure and charging time are two of the obvious drawbacks with battery electric vehicles. With insufficient numbers of chargers, people may have to wait in line for extended periods of time. Moreover, a person who lives in an apartment, for example, may not have access to a charging station. And in areas prone to power outages, there may not be electricity available to charge vehicles during emergencies, which can be life-threatening in case of floods or fires.
“Battery technology is not ready for long-haul driving,” said Luke Gear, technology analyst at IDTechEx. “What’s holding back electric vehicles is the upfront cost of batteries and charging infrastructure. About 50% of the electric vehicles in the United States were sold in California. According to our research, in 2028 we will reach the peak of internal combustion engines, and then sales will decline.”
Carmakers are betting that hydrogen fuel cells will be a key part of that shift. The vision is to be able to pump hydrogen like gasoline from new or existing filling stations, with an infrastructure to store and move liquified hydrogen. But it’s not that simple.
“With an FCEV (fuel cell electric vehicle) you get high-performance and long range,” said Monterey Gardiner, senior hydrogen and eMobility Technology Engineer at BMW Group. “The challenge is additional hydrogen infrastructure and high system cost. Right now you need to make 500,000 vehicles per year to get to a competitive cost.”
BMW plans to introduce a fuel-cell version of its X5 next year. Honda, Toyota and Hyundai already offer fuel-cell powered vehicles.
“The powertrain fuel cell is complete,” said Byung-Ki Ahn, senior vice president of the Electric Powertrain Division at Hyundai MOBIS. “What it needs now is thermal management, so we have created direct oil cooling to a hotspot. In addition, we used to use double-sided cooling. Now we use multiple layers and silicon carbide modules.”
Ahn believes fuel-cells will be a critical component in vehicles over the next decade, as prices drop and economies of scale kick in.
The big question for automakers is whether this is enough. Tesla’s rapid expansion has sowed panic among long-established carmakers with its construction of a manufacturing plant in Shanghai and announced plans to build another one just outside of Berlin.
Relying on hydrogen requires significant production and infrastructure investment, as well. Liquid hydrogen is difficult to manufacture in volume, transport and store.
“Hydrogen makes metal brittle,” said K. Charles Janac, CEO of ArterisIP. “There are three ways to store it. One is cryogenic. The second is using metal hydrides. And the third is using organic materials. None of them is satisfactory.”
A fuel cell also is expensive, at least for now. One reason is that it requires platinum, which is used as a catalyst. Current research involves alloys of platinum and ruthenium to reduce that cost. But some of those costs also can be offset by reductions of precious metals in other areas.
“Right now, platinum and palladium are used in catalytic converters for internal combustion engines,” said William Crockett, vice president at Tanaka Kikinzoku. “We literally ship catalytic converters to Japan today to recycle those materials. That cost goes away with hydrogen fuel cells and electric vehicles.”
New business models, new approaches to technology
Alongside of these shifts is a recognition that the old way of doing things doesn’t work anymore. Volkswagen disclosed in June that it would develop most of its software in-house, increasing its share from less than 10% to more than 60% by 2025. The company said it plans to use about 5,000 experts to create a single operating system across all of its brands.
“What they’re all seeing is that incremental improvements are not enough,” said David Fritz, senior autonomous vehicle SoC leader at Mentor, a Siemens Business. “To compete against Tesla they need different methodologies, tools and architectures. Virtually all of the automotive companies have gotten to the point where they realize they can’t keep doing things the same way. But the big challenge is how to make that leap in technology and skill sets. This is the model that Apple pioneered, where you bring in-house everything that can differentiate you.”
ArterisIP’s Janac agreed. “The Tier 1s either become software- and silicon-savvy, or long term they have very little future.”
Still, all of these changes have prompted something of a renaissance in innovation. One such development involves sensor fusion, whereby logic is added at the sensor level so that processing can be done at the source of the data rather than centrally.
“One of the new sensors is an inertia sensor,” said Fritz. “When you apply your brakes, you expect the car to slow down in a certain distance on a slope. But that’s not always accurate. If you can sense intelligently, you can apply braking only as you need it, save power, and regenerate energy.”
That has other benefits, as well. By doing processing locally, the amount of data that needs to be sent to the central brain of a car is significantly reduced. That reduces the wiring harness requirements, the overall weight of the vehicle, and it improves range per charge.
Fraunhofer’s Engineering of Adaptive Systems Division introduced an image sensor that only saves and processes the important data in an image, rather than the whole image. This achieves the same kinds of benefits in cameras, where the largest amount of data is collected today.
An alternative approach is to to more with existing technology in order to reduce the number of components and systems required ultimately for autonomous driving. LiDAR, for example, has been improved to operate in more weather conditions than in the past, and also now can measure velocity and classify objects.
The same approach is being applied to improving energy efficiency across a vehicle, and this is particularly obvious with efforts to utilize in-wheel or hub motors. So rather than a single motor propelling a vehicle, there would be two or four motors.
The idea of hub motors has been around for about a century, and the big problem has been unsprung weight — that portion of a vehicle not supported by the suspension. Using this approach requires changes in the vehicle design because it can affect the overall ride and handling. But there are big advantages to making this work. The motors are smaller and they connected directly to the wheel, meaning less energy required to propel the car forward, more opportunity for harnessing energy directly for propulsion, and less weight.
It appears that that at least some of the previous concerns have been resolved. Hyundai MOBIS’ Ahn said development has been underway since 2011, and the technology is expected to be commercialized within the next few years. But there are hybrid versions of this approach, as well, such as Citroën’s dual-motor 19_19 concept car, unveiled earlier this year, which can travel 800 km (497 miles) on a charge and recharge 80% in 20 minutes.
Fig. 1: Citroën’s 19_19 concept car, with tires developed in collaboration with Goodyear, at VivaTech conference earlier this year. The central hub on the wheels remains fixed. Photo: Semiconductor Engineering
A big challenge with electrification is in-vehicle energy storage. Lithium-ion batteries are expensive, heavy, and volatile. Scaling the technology involves more batteries, more weight and more potential problems.
This is where solid state batteries fit in. Based on solid electrodes and electrolytes, they are far easier to handle without fear of starting fires.
A number of research projects are underway involving solid-state batteries, such as an all-solid lithium battery developed by Deakin University in Australia. Problems that still must be solved are reduced conductivity, particularly in cold weather, compared with liquid electrolytes. There also have been questions about whether there is enough lithium to power all-electric vehicles.
“We don’t think there’s enough lithium on the planet for 1 trillion IoT devices plus automotive batteries,” said Eric Hennenhoefer, vice president of research at Arm. “A significant amount will have to be derived from energy harvesting.”
Work is underway across the industry to make that happen. Alongside of that, reprocessing of lithium as demand increases is expected, although that still may be years down the road.
“The main source of lithium is in Australia,” said Tanaka’s Crockett. “Production last year was more than 20 tons. There is also one lithium mine in Nevada (roughly 200 miles from Tesla’s battery factory), and there’s more lithium in Chile, Argentina and Bolivia, which are known as the ‘Lithium Triangle.’ China invested $4 billion in a stock deal to get into the Lithium Triangle for its electric vehicles. But you’re also seeing companies like BMW, Volkswagen and Japanese carmakers investing in solid state battery technology. There are a number of different chemistries being developed.”
What makes this so important is that solid-state technology is more modular, stackable and smaller. It uses a prismatic design rather than a cylinder, lasts longer, and charges faster. They’re also more easily replaceable, Crockett said.
More in-vehicle data processing
Another big change involves data processing. Several years ago, the general consensus was that 5G would be so ubiquitous that data from all vehicles could be sent back and forth to the cloud with enough speed to eliminate the need for on-board processing.
That proved untenable for several reasons. First, signals in the millimeter wave spectrum for 5G signals are highly susceptible to interference, and they attenuate quickly. Second, there is too much data to send back and forth from the cloud, and even in the best of circumstances that communication takes too long to avoid obstacles on a roadway. And third, the cost for building and maintaining a network of repeaters would be prohibitive across thousands of miles of roadways.
“5G represents an improvement, but not good enough to offload computing, particularly in the sub-6GHz range,” said Scott Jones, advisory semiconductor practice leader at KPMG. “In fact, one of the reasons the opportunity for semiconductor content is so high is because 5G is not good enough.”
This is a key factor in sidelining autonomous driving. The amount of compute power needed to address all of the possible interactions and corner cases is enormous, and far too expensive to put into a vehicle today. Current estimates are somewhere in the range of three supercomputers.
A second factor involves the sheer size and cost of the compute resources needed to deal with all of the potential corner cases and interactions with human-driven and autonomous vehicles, as well as pedestrians, bicyclists, inclement weather, potholes, lane shifts, and so on.
“All of the prototypes are trunks full of data centers,” said Robert Day, director of automotive solutions at Arm. “Those are not deployable. They’re too big, too power-hungry, and too expensive. What companies are demanding from us are server chips that provide the same performance at a fraction of the cost with better thermals. We’re also being asked for high-performance CPUs with built-in functional safety, which is new. The other thing that is new is the Autonomous Vehicle Computing Consortium. Part of the charter of AVCC is mass deployment in the 2025 time frame, although full autonomy will be much later than that. But what’s new here is they have decided they’re not going to differentiate with that technology.”
Fig. 2: Amount of semiconductor content and value at each driving level. Source: KPMG
This represents a significant change in direction, because companies have been viewing autonomy as a differentiator. With everyone working on it together, it becomes a base-level standard, and it also comes to market more quickly. That, in turn, generates far more demand for all kinds of electronics.
“This is a revolution in thinking among the Big Three automakers (Volkswagen, Toyota and Hyundai),” said Synopsys’ Huhnke. “The question is whether there is enough talent to make this work. The biggest problem is that the thinking process is different. It will take time. It’s a question of who can shift their focus to a hardware and software stack and scale up. That is the new battle. The Big Three have enough financial power to invest in this and there are a lot of customers worldwide, not to mention a lot of employees whose jobs depend on this working. But it’s not going to happen overnight.”
Transportation as a service
Another big change involves how vehicles are used. Most analysts believe there are significant opportunities for different ownership and use models.
“What we’re seeing is that for very specific trips that people take, there are massive opportunities for mobility as a service,” said KPMG’s Jones. “What that means is vehicles will be utilized for up to 50,000 to 60,000 miles per year. So even if the car costs $120,000, in three years at 50,000 miles a year you will make money off the vehicle. We’re also seeing real interest in this in retirement areas. That’s a huge opportunity, even at low speeds of 30 miles per hour.”
IDTechEx’s Gear agrees, noting this will have a big impact on car usage models. “The upswing for autonomous ride hailing begins in the 2025 time frame,” he said. “But it will take awhile to get going because of a reduction in government subsidies and the U.S. government relaxing restrictions on cars. Subsidies end in China and Norway next year.”
Subsidies have long sustained green energy. But as costs increase, and as new regulations such as Europe’s tough new emissions target of 95 grams of CO²/km roll out in 2021 — the current standard is 130 grams/km, and the new standard will be phased in starting in 2020 — that could help drive mobility as a service. Internal combustion engines will have a tough time meeting those standards, and electric car ownership is less convenient for people who don’t have place to charge their vehicles.
This is a different way of looking at vehicle sales, however, and it requires a different model for design.
“You need to look at next-generation cars as a floorplan,” said Mentor’s Fritz. “This is a different way of approaching the problem. So you design for Level 5 and you get there one step at a time. It seems as if a lot of companies have lost their way.”
All of these changes raise questions about what automotive companies will look like in a decade. Will Tier 1 or electronics companies become carmakers, or will automotive giants become chipmakers and develop their own chips? At this point, no one is certain. And while Tesla and new BEV startups may have the lead today, there are no guarantees things will remain the same. In fact, the future giants of automotive may look less like automakers today than electronics companies.
What is clear, though, is the semiconductor industry will benefit from more electronics, more services, and more tools required to build those chips and systems. The automotive electronics industry feeds other parts of the design through manufacturing chain, as well, and that is likely to drive a boom that will last for decades.