Giant Auto Industry Disruption Ahead
The move to self-driving vehicles over the next decade or so will result in a massive restructuring of entire segments of the global economy that have evolved to create and support automobiles and the people who drive them.
The shift will create many new jobs-particularly for semiconductors and electronic systems-and conservatively it will eliminate hundreds of thousands of existing ones. It will reshape entire communities and ecosystems. Perhaps even more consequential to existing automotive companies, it has the potential to upend established companies and replace them with others, including some that have played only limited roles in the automotive market.
The general consensus is the path to fully autonomous vehicles will precipitate one of the biggest disruptions in the history of business. Just how vast and deep the changes will become is difficult to grasp because there are so many interconnected pieces to the global automotive industry, not to mention scores of companies entering this market. But consider that in an age of autonomous vehicles, revenue from speeding tickets will evaporate. Speed traps will be non-existent. Billboards will be ineffective on long-distance trucking routes, and truck stops may cease to be viable. Rental car agencies at airports could cease to exist.
Car insurance, a thriving industry today, will likely cease to exist as we know it. It may be included with a car-if people even own those cars. Accenture estimates the loss to insurance companies will be $26 billion by 2035. But it also points out there are new opportunities alongside autonomous vehicles, including comprehensive coverage models for cybersecurity risks, software or hardware failures and infrastructure risk and fleet operation liability.
The economic impact of autonomous vehicles ripples out everywhere. A commonly used metric in economics is that money changes hands five times from the first dollar spent. In the case of automobiles, that adds up to a significant amount of money that will move in completely different directions than in the past.
Consider the following statistics from the Alliance of Automobile Manufacturers:
- Half of companies listed in the Dow Jones Industrial Average depend on automobiles for revenue.
- Of the G20 member nations, every country except Saudi Arabia manufactures automobiles.
- Automotive manufacturing accounts for $953 billion in revenue, and is responsible for $109 billion in R&D around the globe. In the United States, it represents between 3% and 3.5% of total GDP-and that’s just one slice of the automotive sector.
- In the United States, 7.25 million jobs are tied to the automotive industry, including 2.44 million at automakers, 1.65 million at auto dealerships, and 3.16 million at auto suppliers.
There is no shortage of data or opinions on this subject. Every major consulting firm has issued one or more reports on the impact of assisted and autonomous driving. (Accenture Mobility; PwC 2017 Strategy & Digital Auto Report; McKinsey’s Self-Driving Car Technology; Deloitte’s Fact, Fiction And Fear Cloud Future Of Autonomous Vehicles, among many others.)
There are a several reasons this shift has garnered so much attention. First, the automotive industry has been progressing almost linearly for more than a century. “The car industry went through a huge change in the 1950s,” said K. Charles Janac, chairman and CEO of ArterisIP. “For the last 60 years, it has been almost completely stable. It’s basically the same product being incrementally improved until mid-2010 (when Tesla issued its IPO). If you look at a legacy car company, the dealers make all their money from maintenance, the assembly people are skilled at mechanical assembly, and the purchasing people have relationships that have to be fractured. The implications are humongous.”
Lip-Bu Tan, president and CEO of Cadence, agrees: “This is a very old, established industry and it’s now going through a big transformation. It’s comparable to what Steve Jobs did with the phone. Apple came from nowhere, and suddenly everyone was chatting and sharing data. You can apply that same kind of change to the automotive industry.”
Second, this is a huge industry with large amounts of capital, and spending continues to rise.
And third, the automobile is a key component in people’s lives. It is essential for many people to get to their jobs or away from work altogether. Outside of cities, it is critical for getting groceries, getting medical treatment. And cars consume a significant amount of money earned, both for purchase and maintenance.
For all these reasons and more, there are plenty of companies vying for a piece of this market, particularly from the semiconductor side. The Consumer Electronics Show, which until several years ago was the primary venue for connected technology such as smart watches, 3D TVs and smart appliances, is now dominated by automotive technology. The GENIVI Alliance, which focuses just on in-vehicle entertainment, required two full ballrooms in Las Vegas’ Bellagio Hotel this year, rather than a handful of booths. And the number of companies with some connection to the automotive industry keeps rising.
Winners and losers
“We’re tracking 330 companies focusing on electric vehicles, and 107 on autonomous vehicles,” said Wally Rhines, president and CEO of Mentor, a Siemens Business. “Not all of them will survive. But the good news for EDA is everyone has to load up on design tools for at least the next five years. A big area for this market is fault inject and driver simulation. If you were to verify a driverless car, you would have to drive 14 billion miles. We’re now in the single-digit millions. The only way to get there with driverless cars is virtually.”
This includes everything from modeling and simulation to virtual prototyping, and for semiconductor tools companies this is contributing significantly to sales. “Some of the published numbers show this will be a $160 billion market in 2022,” said Rhines. “Today, semiconductor TAM (total available market) is somewhere around $375 billion. Next to sensors, this is the fastest growing segment for chips. The question is how much of what we already have will be able to migrate to autonomous vehicles. So if you look at speech recognition, it’s good enough for a cell phone, but it’s not good enough to turn your car.”
Part of the opportunity comes from the electrification of vehicles, which is step No. 1. The other piece involves assisted and autonomous driving.
“The move to autonomous vehicles involves two big changes, not one,” noted Aart de Geus, chairman and co-CEO of Synopsys. “The first one is the move from the combustion engine to the electric motor. One shouldn’t underestimate how big a change that is, because many of the skills of the past are based on 100 years of harnessing the combustion engine. Using an electric motor is super-efficient and super easy to control. There are many benefits. But instead of a gas tank you have a battery, which is a very sophisticated electrochemical device. The evolution of those is consistent with, but not as fast as, the evolution of Moore’s Law. And with that come other issues. One of the things people don’t appreciate is that a gas tank, in itself, carries no energy. It carries part of the energy, and it only delivers that energy when combined with oxygen, which isn’t in the car. A battery carries all the energy. It’s stored in a chemical form that can be released very rapidly.”
This is a manageable problem, but it is still a problem because it requires a separate system to recharge the batteries. It’s not so easy to add high-voltage lines everywhere and just replace gas stations. But all of this has to happen in parallel with the electrification of vehicles.
The move to fully autonomous vehicles adds yet another set of challenges.
“The amount of computation that will go into autonomous vehicles is enormous,” de Geus said. “You have an enormous amount of data being taken in from an array of cameras on a car (including LiDAR and radar). There is an enormous amount of work going into collecting data and actually doing the learning, which is being done in the cloud. There is a whole field developing with specialized computation for that. There are companies designing their own processors for specialized algorithms. If it is 10 times faster, there is enormous value in that.”
Who reaps that value isn’t obvious, though. “As we move to autonomous vehicles, incumbents may not be incumbents anymore,” said Cadence’s Tan, noting that consolidation will lead to more startup activity. “With autonomous vehicles there are more electronics, more software, and higher-speed data. The whole business model is changing, and that’s a big change for an established industry. This is equivalent to the way Amazon has changed retail, where now you deliver products to the home. It’s the same with Netflix. And as business models change, so do the leaders.”
There is plenty of agreement on that point, and an equal amount of uncertainty. “Self-driving vehicles raise the bar significantly for automotive electronics,” said Joerg Grosse, OneSpin Solutions‘ product manager for functional safety. “Some manufacturers will rise to this challenge. But others may not, leaving room for more innovative new players. Consumers and legislators will demand a level of functional safety and reliability previously required only for military and aerospace applications. For a start, functional verification must be much more thorough. A missed corner-case condition can easily lead to someone being killed. The use of formal-based technologies will expand greatly, since only formal provides mathematical certainty. “
Correctness is not enough, though. “Owners will expect their vehicles to continue to operate safely even when a stray alpha particle flips a memory bit,” said Grosse. “Automotive chips must have error detection schemes, and the coverage of these safety mechanisms must be verified by injecting faults and proving that they can be fixed or detected in time to take corrective action. Finally, automotive electronics have to conform to relevant standards, especially ISO 26262. There are many unanswered legal questions around liability for self-driving vehicles. Releasing a product without all proper standards certification would be unthinkable. Automotive manufacturers and their electronics suppliers must improve their verification, implement safety schemes with known coverage, and obtain certification. Anything less is commercial suicide.”
Timing and limitations of technology rollouts
So when exactly will autonomous vehicles begin showing up on roads? That isn’t clear yet, either.
Maarten Sierhaus, director of Nissan Research Center Silicon Valley, breaks it down into two pieces. “Eyes on, hands free, will happen in 2020 in cities. Eyes off, robotaxis, will show up in 202x. But we need to develop a system that is accessible to the way we drive. People behave differently in different places. In Amsterdam, bicycles rule. In San Francisco, pedestrians rule. And cars need to know the difference.”
This requires a combination of social science, AI and mechanical and electrical engineering. At least three of those four specialties are new for the automotive industry.
“An AI system can negotiate with other systems or pedestrians,” said Sierhaus. “But how does it let a car go or behave the way we do as human drivers? The first step is human-robot teaming, where you still want humans in the loop. Nissan has been working with NASA on the SAM concept (Seamless Autonomous Mobility), where you have one or more individuals helping multiple cars. The question is how many vehicles can one person handle. You want intelligence in the vehicle, in the cloud, and human intelligence when it’s needed. That’s particularly important for a fleet management system.”
All of this will take time. Estimates typically range from 10 to 15 years for fully autonomous driving in place where human-driven and autonomous vehicles need to interact. It is expected to happen faster, such as more remote cities in China, where there is no existing infrastructure, and much slower in older cities that are crowded with pedestrians and bicycles, or in areas where charging is not available. Even within relatively new urban areas, infrastructure needs to be developed to charge electric vehicles. At this point, most of the electric vehicle sales are commercial fleets and suburban homes, where residents can connect a charging cable from their car to an electrical socket.
Side by side with this is a question of cost of technology. The electronics that carmakers are hoping to add into vehicles involve sensors with a range of price points, depending upon whether they include basis accelerometers or LiDAR or advanced radar. There also will be artificial intelligence systems, which will likely be developed at the latest process node and require massive compute power. And there will be a variety of processing elements, accerators, I/O systems and memories scattered throughout these vehicles.
Economics of autonomous technology
How much all of this is supposed to cost, and how much it will likely cost, are not clear. What is clear is there is a gap in expectations between automotive OEMs and chipmakers. Chipmakers expect the costs will be in the tens of thousands of dollars, while carmakers want the cost in the range of $3,000 to $8,000 for cars to continue to be affordable.
But it’s not clear how many autonomous vehicles actually will be sold or who the buyer will be. If it is a company such as Uber, which will run autonomous vehicles round the clock, an extra $20,000 over the life of a car may not matter. If the same patterns in car ownership continue with autonomous vehicles, where cars basically sit idle 90% or more of the time, cost will remain an issue. But metrics for quality over time for each of those models are widely different, and may have significant implications for how quickly this technology gets adopted. So far, none of this is worked out.
“The baseline differs, depending on who you talk to,” said André Lange, group manager for quality and reliability at Fraunhofer EAS. “And with autonomous cars, the central processing unit may have to decide which information is right and from which sensor, because one or more might get dirty or stop working. What might happen to a sensor and how do you cope with it? This requires that you know where the defect is in a system.”
This requires much more than a collection of inexpensive sensors, though. It requires constant communication among the various parts, with data being analyzed at the edge, in the vehicle’s central logic, and in the cloud. The only way to make that work is for the entire industry to standardize on a single set of integrated technologies and to leverage economies of scale and commoditization of components.
PwC came to a similar conclusion in its 2017 Automotive Trends report: “If auto makers expanded their cooperative efforts, the industry would essentially be smart-sizing, the way the airplane manufacturing sector has over its long history.” Another option, according to the consulting firm, is greater consolidation.
But either of those options make it much harder to differentiate one car from another, because with autonomous vehicles branding is no longer about horsepower or cornering ability. It is strictly limited to the body shape and the electronics and amenities inside the cabin, and even those may be of limited value if people don’t actually own their cars.
Global reset button
This is what keeps automotive industry executives awake at night. BMW pitches its cars as “the ultimate driving machine.” Porsche’s ad for a Cayenne Turbo is “sportscar together.” Volvo has built its brand on safety. General Motors describes its lineup as “a wide array of vehicles and brands to fit your individual driving needs.” And Ford’s home page says, ‘We all drive, some just for the fun of it.”
None of those branding messages will work as autonomous vehicles begin rolling off the assembly lines. Past developments in mechanical engineering may be interesting from a historical standpoint, but electric motors are simpler to build, have fewer parts, and they are pretty much the same from one car to the next. That ultimately leads to consolidation and a reshaping of the entire industry.
For autonomous vehicles, even the skill sets are different. “Many of these car companies are excellent at mechanical design, but now the core competency is software,” said ArterisIP’s Janac. “This is changing the chip industry, too. The programmers are now programming for a specific chip. We’re moving toward software-defined hardware, which means a car company will become an integrator of subsystems. But that’s not why you buy a car. So there are three models emerging so far. One is the Intel/Microsoft model, where they are the provider of IP and the system assembler does not do any innovation. The second is where Tier 1s like Bosch and Denso become more important system manufacturers and build their own chips. And the third is where the car companies take back control of their architectures and everyone else operates on low margins. It’s unclear at this point which one will win.”
With regard to software, automotive OEMs are competing with companies such as Google, Facebook/Instagram, Microsoft, and Amazon for engineering talent, which have a reputation for higher compensation, including stock options, and greater career mobility.
It’s also not clear which countries will win. The move from gasoline engines to electric motors is significant for a different reason-it reduces the number of parts to 200 from about 2,000. Moreover, almost nothing in an electric motor is difficult to master. In an gasoline engine there are various combustion modes, direct-injection technologies, or variable cam phasing, for example.
This puts companies in places such as China-latecomers to the combustion engine market-on par with automotive giants when it comes to electrification.
“China has not been a factor, but we see that they can quickly enter and possibly dominate this market,” said Sundari Mitra, CEO of NetSpeed Systems. “For China, this will extend well beyond the automobile to include smart cities. They are innovating, and Chinese companies do not have a cash shortage.”
Mitra noted that one of the biggest shifts underlying all of this is a verticalization of the semiconductor market, which makes it easier for smaller companies to compete. This is particularly evident in the automotive sector, where the number of companies competing for a piece of the market has grown significantly over the past year.
“This is all about the economics of autonomous vehicles,” said Mitra. “The spend on semiconductors in autonomous vehicles versus electric vehicles will double.”
Redefining who wins
There are other pieces to this giant puzzle that aren’t even on the horizon for most companies yet. One involves data, which today is largely limited to location-based services such as Google’s Waze or Apple Maps. This whole segment of the industry is about to explode in a variety of ways, some good and some bad. On the positive side, being able to collect data is extremely useful for improving safety and identifying potential problems.
“Right now we don’t see people using data for quality the way it should be used,” said Michael Schuldenfrei, CTO at Optimal+. “This is all being done at the component level today, and it’s not being shared across the supply chain. What we’re looking at is quality protection as a service, where you can share data without exposing IP. We believe that if zero defects per million is the goal, which is what the automakers are asking for, then you have to get over this problem. You have to understand failure from the system-level versus the components. The interaction of components is critical to avoiding the ‘no trouble found’ syndrome.”
Data becomes increasingly important in this new world of autonomous vehicles, and that has broad implications, a well. “At the end of the day, this is a data-centric world, and whoever owns the data owns the landscape,” said Cadence’s Tan. ” The key components are sensors, processing and the hyperscale cloud. So the first thing is how you collect data. Then it’s how you process data, which increasingly involves edge computing. And then the entire data center and cloud are changing.”
One of the comparisons that is often cited in conferences and presentations about autonomous vehicles-and almost every semiconductor conference today includes at least some talk about assisted and autonomous vehicle technology-involves avionics. Commercial aviation today is one of the safest forms of transportation available.
“At this point, we don’t know how this will impact safety,” said Louie de Luna, Aldec‘s director of marketing. “But what we have seen is that regulation itself has restricted the advancement of technology in avionics. There is a lot of paperwork. The avionics industry has made travel safe, but they have to document the process. The automotive industry is way more advanced when it comes to technology, though. In avionics, you’re probably not going to see 10nm or 7nm chips, but in automotive you will see the same kind of intensive simulation, verification and way more testing. This requires a layered approach, and it requires not only functional coverage, but structural coverage, as well.”
Put simply, there is a lot of work to be done on every level. While autonomous driving capabilities may be present in some cars in the future, it’s not clear when those cars will be considered safe enough to take over completely. But once those cars do begin showing up on highways and local roads, the automotive industry -and everything it touches-will look very different than it does today.
China’s Ambitious Automotive Plans
After years of playing catch-up to multinationals, the country is targeting an opening with assisted and autonomous vehicles.
Verification Of Functional Safety
Part 1 of 2: How do you trade off cost and safety within an automobile? Plus, a look at some of the challenges the chip industry is facing.
Reshaping Automotive Design
Conflicting goals, evolving standards and the need for new methods and tools make this an interesting market for chipmakers.
The New Road Warriors
Tech companies pour into automotive sector, vying for a piece of the assisted and autonomous driving market.
Radar Versus LiDAR
Racing to drive down sensor costs for the automotive market.
Auto Chip Test Issues Grow
Semiconductors used in cars have higher quality and reliability requirements than most chips, but they have the same cost and time-to-market pressures.