Welcome to the planet's grandest high-stakes poker game, one where the players at the table are auto manufacturers frantically raising the ante in their financial commitments, betting tens of billions and their very survival on their electric vehicle development decisions.
In early 2019, Reuters estimated that the top 29 global auto manufacturers had already pledged to invest more than $300 billion towards developing electric vehicles( EVs) and supporting technologies including autonomous driving capability. Since then Daimler, Ford Motor Company, General Motors, Stellantis and Volkswagen Group, have committed an additional $152 billion in investments, a sum four times their combined 2019 operating profits.
An additional $60 billion plus in investments have also been made in more than 70 EV start-ups wanting in on the game. Of course, all this money does not include the other untold billions being directed to the 400 plus start-ups in China, or the billions being invested by global auto suppliers like Bosch, Denso or ZF Friedrichshafen.
Deciding that the future of transportation, and therefore their survival, will be based on EVs, many of the top auto companies have promised to end production of their internal combustion engine (ICE) vehicles by 2035, if not earlier. Others like Audi, Daimler and VW have stated they are now stopping the development of new internal combustion engines altogether.
Yet, even with the global EV market share projected to grow from only about 2.6 percent of auto sales today to some 20 to 25 percent of sales by 2030, not all competitors will be winners in a world soon to be awash in electric vehicles. Consulting firm KPMG, for example, estimates that in addition to the 179 new battery EV model launches and refreshes between 2016 and 2020, at least another 251 are expected between 2021 and 2023 by traditional auto manufacturers, or original equipment manufacturers (OEMs) as they are referred to in the industry, as well as start-ups.
These numbers above exclude all the existing hybrid electric and other alternative fuel models available or in the works, which number at least another 265 models. As a comparison, in 2011, there were only two battery electric and 29 hybrid electric vehicle models offered for sale in the US.
With all these EVs chasing the same relatively small number of current EV-inclined buyers, how does an EV maker differentiate itself in such a crowded market?
While EV driving range and price are obviously key market differentiators, an EV's software suite is quickly becoming as, if not more, important. Even now, if an EV's software is not perceived as being advanced, buyers for it become harder to entice. For instance, in China, which boasts the world's largest and most competitive EV market, VW has had problems selling its new flagship EV ID.4 model, according to a report by Reuters, because its features are not as sophisticated as those offered by other EVs already available in the country, which have self-parking, advanced-voice control and self-driving capabilities.
While VW says new software-updates are planned to provide these features, unless the company can not only provide the same experiences but ones more advanced than its competitors fairly quickly, VW may find its ID.4 is already considered passé, at least for many potential EV customers in China.
VW's experience highlights a looming problem for all competitors in the EV poker game. A recent study by consulting firm McKinsey & Company asserts, "Advanced electrical and electronic (E/E) capabilities… and the ability to make design upgrades during the (EV) vehicle life cycle are becoming crucial to remain competitive, both in China and globally." Falter at one or the other, and an EV player's betting position might be wiped out quickly.
Further, if the EV range gains slow down over the next decade as expected, an EV's software-driven features along with their affordability, will decide the winners of the potential $1.1 trillion in annual EV sales, while the more profitable software-related revenue worth billions will be up for grabs.
Where rubber+software meets the road
Not only are EV makers chasing the same limited number of customers, they are also pursuing an even more limited supply of software and systems engineers with smart mechatronics and robotics expertise. Software's complexity in current ICE vehicles is staggering, with many vehicles having 150 million lines or more of code. However, future EVs will likely have triple or more the lines of code as advanced autonomous driving features become available.
Complicating the expertise issue further is that new EVs (and ICE vehicles) are increasingly "cyber-physical systems." Simply put, cyber-physical systems unify the physical world with the world of information technology. Vehicles as cyber-physical systems are not mere self-contained and isolated entities but are ones that will evolve in capability over a decade or more often in response to other evolving systems such as transportation infrastructure, manufacturer monitoring or dealer management systems, the Internet, and other vehicles.
Today's connected vehicles create up to 25 gigabytes of data per hour, a small portion of which is being shared outside the vehicle. However, by 2030, when vehicles could be interacting with scores or more external systems over a range of communication channels, that amount may reach four terabytes per hour, all of which will be captured, analyzed and monetized by multiple remote third-party systems.
"Much of that future data, perhaps up to 90%, will be unstructured," observes Jeff Fochtman, Senior Vice President of Business and Marketing at data storage company Seagate Technology, given it will be originally generated by a vehicle's camera, lidar, radar and or ultrasonic sensors. The amount, type and usefulness of the data poses unique challenges for automakers in deciding which data to store, how to store it, and where to store it. Data privacy, safety and security compliance issues intensify the challenges involved. Underpinning all this data communication is the assumption by auto companies that sufficient network bandwidth with minimum latency will be available soon to support tens of millions of continuously communicating vehicles.
"However, bandwidth alone does not suffice to move—in order to use—all this data," Fochtman adds. "There will be a necessity to move massive data sets (100TB and over) quickly via data shuttles and arrays, for example, from fleet vehicles to data centers."
Creating "smart vehicles" that sense, think, act and communicate in real time within a large transportation ecosystem, that use and generate vast volumes of data, that are increasingly electric powered, and that need to evolve their capabilities over time via over-the-air (OTA) software updates and upgrades represent a radically different system design paradigm for traditional auto manufacturers. The systems and software engineering challenges are materially more demanding. And creating reliable, interconnected, open, updatable and secure system-of-systems at scale is, to put it mildly, non-trivial.
Hence, separating EV winners from losers will depend not only on financial wherewithal, but on each competitor's "relative strength in their cyber-physical systems engineering," states Chris Paredis, the BMW Endowed Chair in Automotive Systems Integration at Clemson University. He adds that vehicle complexity "has pushed beyond what we can handle traditionally. Automotive systems have risen to a level of complexity where formal systems engineering approaches are needed."
However, if it is performed well, "Cyber-physical systems engineering becomes the enabler of (necessary) complexity and innovation," which will be critical to those wanting to stay in the EV game, according to Paredis.
Furthermore, while the "trade-offs of economics and performance are all going the same direction, where it is now feasible to build an EV for the middle class," Paredis observes, there is little consensus over the best way to design smart, affordable EVs over the next decade.
There are three main electric vehicle drivetrain designs that an EV manufacturer can choose to pursue: battery powered, hybrid, and or fuel cell. Among these, there are several different types of energy storage systems as well as multiple alternative hybrid design options available. Each option has its own strengths and weaknesses. Which combination of options a manufacturer or start-up pursues is based on how they see the EV game unfolding over the next decade.
"There will be a necessity to move massive data sets (100TB and over) quickly via data shuttles and arrays, for example, from fleet vehicles to data centers."
Among the numerous multi-billion-dollar cyber-physical system engineering decisions that are being hotly debated within the industry, two tend to stand out presently. The first revolves around the tradeoffs and timing involving which EV drivetrain option should be pursued, while the second concerns the systems and software architecture that should be the foundation for providing the intelligence in smart cars.
The consequences of these decisions are profound. One or more top tier OEMs will likely get their choices wrong and disappear or be acquired in the next decade, consultant firm KPMG predicts. On the other hand, the number of EV start-ups that are forced out of the game might be counted by the score.
Modify ICE or Go EV? Drivetrains Force Hard Choices
"We are in the midst of a revolution, the ongoing digital redesign of the historically analog automobile," says Venkat N. Krovi, Michelin Endowed Chair of Vehicle Automation in the Departments of Automotive Engineering and Mechanical Engineering at Clemson University, "which has led to the unraveling of the original vehicle form."
The elimination of the ICE powertrain along with the combination of affordable electric motors and control software creates "a fundamental flattening of the automotive industry," Krovi observes, thereby allowing everyone, "to gain insight into the formerly opaque world of the car and has lowered the barriers to entry to new automotive competitors." This has catalyzed the explosion of EV start-ups, each of which fervently believes they will be the one to fundamentally disrupt the automotive industry.
Importantly, Krovi points out, the EV drivetrain is much simpler because of the reduction in system complexity. "Think of the efforts needed to make building the controlled explosions in an internal combustion engine safe, powerful and free from pollution," he says.
Additional vehicle elements are also simplified. For example, when financial services company UBS conducted a teardown (PDF) a few years ago of both the Chevy Bolt, the world's first mass-market EV and a VW Golf, which it considered an equivalent ICE vehicle, it found that the Bolt only had 35 moving and wearing parts compared to the 167 in the Golf. The Tesla EV has only around 20 such parts.
"Unlike an ICE vehicle where the steering column requires a physical mechanical connection, in electric vehicles, software becomes the controlling element now. You can control an EV with a joystick, inside or even outside the vehicle, for instance. Once you decouple the physical connections required, your design freedom explodes," declares Clemson's Krovi.
How OEMs and EV start-ups choose to use their newfound design freedom varies greatly. For example, VW, following Tesla's lead, has decided to create an EV specific, or native, electric battery platform, called the modular electric driver matrix (MEB). This so-called "skateboard-like" approach which underpins its EV ID.4, is simpler, more flexible, and less costly than the modular transversal toolkit (MQB) approach that VW previously used for its battery powered e-Golf. For that car, the MQB platform was highly modified to allow batteries to be placed throughout the vehicle, while the MEB approach allows for a larger battery pack and thus longer driving range, which is a major advantage of creating EV specific platforms.
Yet for now some manufacturers, like BMW, Jaguar Land Rover, and Stellantis have held out against introducing an EV-specific drivetrain architecture, although each has plans to do so in a few years. They believe that their optimal approach is to continue to build EVs on flexible drivetrain platforms that support both ICE and EV vehicles, albeit with some EV driving range limitations. Udo Hänle, BMW's Head of production strategy is quoted in Automotive News Europe story as saying, "In our view, market forecasts are too uncertain to warrant inflexible, electro-only platforms."
In addition, it is expensive to build EV-specific platforms (about €1 billion), and given the uncertainties of just how quickly the EV market will grow, some carmakers believe it is better to hedge their financial risk for a while still. Furthermore, BMW does not believe customers will purchase an EV based on its underlying EV platform alone. "It's not relevant for a buying decision," argues BMW Group CEO Oliver Zipse.
Larger automotive suppliers are also challenged to place high-stakes bets. For example, ZF has stated it will no longer develop any traditional ICE transmissions without a hybrid or electric variant and pledged to spend $14 billion to develop electric and autonomous technologies. "The coming transformation in the industry is clear and the bets are huge and will vary so much by region—the rate of EV and autonomy adoption in developed Western markets or China will be very different from that of say India or South America," says Andy Whydell, ZF's Vice President of Product Planning for Vehicle Systems.
As always, the market will decide which EV platform approach, dedicated or ICE retrofit, is ultimately correct. What no manufacturer disputes, however, is their ever-growing dependence on the software executing in their EVs computing systems to provide a competitive edge. And here too, people passionately argue about the best approaches to provisioning, architecting, owning and executing that software.
Supercomputers on Wheels
"EVs are a total reset opportunity, with both hardware and software architectures being revisited and rethought," states Chet Babla, Vice-President of the automotive business for semiconductor design firm Arm Ltd. "The variety of computing elements is becoming important and driving a lot of software complexity that everyone is trying to get their head around."
A variety of computing elements are needed to support infotainment, connectivity and battery management. For example, while the elimination of an ICE powertrain simplifies a major source of software complexity, efficiently managing an EV's batteries presents its own complexity that software must manage. Efficient battery management systems are not only important for EV range and safety considerations, it is especially crucial for EVs with autonomous capability.
The variety of computing elements also stems in large measure from the desire to evolve current advanced driver assist systems (ADAS) into ones that can eventually provide full autonomous driving capability. An autonomous vehicle needs multiple sensors, such as cameras, LiDAR, ultrasonic sensors and/or radars to provide the 360-degree information required for safe navigation. For example, a Tesla currently uses eight cameras and 12 ultrasonic sensors to provide the inputs needed for Level 2 autonomous driving capability, meaning the vehicle can take over steering, acceleration and braking in specific circumstances, but the driver must still have their hands on the wheel to take over if necessary (see SAE Table). Even more sensors will likely be needed to permit full Level 5 autonomous driving, where the vehicle can drive everywhere in all conditions and driver intervention is not needed.
Of course, this voluminous sensor data need to be processed in real time, not to mention the data required for all the other vehicle electronics, increasing the need for extremely fast and powerful computing processors. As a result, "Standard compute is no longer acceptable, we need specialized compute," states Suraj Gajendra, Arm's Senior Director of Technology Strategy for the automotive business, which has "moved computing requirements from an inside-out to an outside-in approach."
"The coming transformation in the industry is clear and the bets are huge and will vary so much by region—the rate of EV and autonomy adoption in developed Western markets or China will be very different from that of say India or South America."
In other words, manufacturers must define their EV computing requirements from a cyber-physical perspective of their vehicles, which possess some level of autonomy, interacting and evolving within a larger, highly connected system-of-systems ecology that itself is also evolving. As Gajendra explains, "The software applications and services that run in a vehicle are developed in the cloud and deployed over-the-air directly onto vehicle. Given the versatile nature of these applications, specialized compute elements like CPUs, GPUs, image signal processors, and neural (network) compute engines are required to execute them efficiently and accurately."
"These applications are 'mixed-critical' in nature. Some of them need hard real-time response within milli-seconds and some don't, some of them need high degree of safety and security, some don't," Gajendra states. "Understanding the needs of applications at a higher level of abstraction and then deriving the processor requirements from them is very critical."
Whereas the past approach to providing computing power was to spread electronic control units (ECUs) throughout a vehicle for localized processing, the strategy currently in vogue has been to consolidate the processing of multiple ECUs into more powerful domain control units (DCUs), or even to consolidate most of a vehicle's computing into a handful of central processors. With this latter approach, vehicle computing resembles more of a generalized computing platform in terms of hardware and software architectures, but with the processing power of a supercomputer.
VW decided with the development of its EV specific platform MEB that it was also getting late to start with a clean-slate approach to vehicle electronics and software, especially if it was going to develop autonomous driving capability. VW felt engineering trade-offs are easier to manage and less costly when building a system with a desired future capability from the beginning than trying to retrofit it in afterwards.
Admittedly emulating Tesla, VW has moved to consolidate the functionality of dozens of ECUs (and multiple electronic architectures) into a small set of central servers running its own core operating system/middleware to create a service-oriented architecture for controlling vehicle application software. This approach allows a better separation of hardware from software development, which can reduce the complexity of producing both. It also allows VW to more easily add or update software-driven vehicle features, including software from open source suppliers rather than only from its traditional supplier base.
Another reason VW chose to build its own OS instead of using something like Google's Android OS (which Ford and GM are doing) is to keep tight control over vehicle data that can be monetized in the future. VW believes that it will generate "triple-digit-millions" in sales in the future through selling customized services, like offering autonomous driving on a pay-per-use basis. It envisions customers would be willing to pay 7 euros per hour for the capability. Tesla has recently offered a monthly subscription to owners of Tesla's who have a Full Self-Driving computer 3.0 or above, plus Basic Autopilot or Enhanced Autopilot, a subscription for FSD capabilities from $99 to $199 per month depending on the vehicle's configuration.
VW is not alone in rethinking its vehicle computing architecture. BMW, Mercedes and Volvo are but a few OEMs who already have or are planning to move to more centralized computing architectures, as well as taking over more of their own software development. Like VW, they all hear the siren call of future software-driven feature subscription revenue.
Chip maker Nvidia's CEO Jensen Huang takes the subscription idea a step further. He envisions auto manufacturers offering future EVs at cost which provide basic driving features, and making their profit on the sale of customized services via subscriptions. If this indeed occurs, those auto manufacturers who have delayed major upgrades to both their EV and digital architectures could find themselves in an eroding competitive position.
Coders Will Clean Up
In this EV high-stakes poker game, winners will be determined both by player skill and more than a bit of luck. Much of the decision-making occurring today is predicated on a multitude of assumptions about EV versus ICE vehicle sales, as well as which type of EV will sell the most, pure electrics and/or hybrid electrics. For example, GM and VW see no future for hybrids, but Toyota, Ford and BMW disagree and see a market for both. EV start-ups disagree which is the best strategy to pursue as well.
It also remains to be seen whether a compelling case for EVs can be made to customers still worried about cost, range and recharging infrastructure, not to mention their hesitancy to spend much for autonomous driving capabilities. Government investments and incentives like those proposed by the Biden Administration, or those in China, the European Union, and Russia may help overcome some of those concerns, but if the EV market grows slower than predicted or hoped for, there may be a lot of lost bets. Paradoxically, if the EV market grows faster than predicted, EV laggards could be wiped out.
Then there is the question of obtaining the software and cyber-physical systems engineering expertise needed now and in the future. "OEMs built everything around the engine and chassis. Electronics were never core, but now they are all in a rush to do it," notes Uwe Class, ZF's Vice President of Advanced System Development at ZF Friedrichshafen AG and one of the world's largest suppliers of automotive components.
For instance, VW has set out the ambitious objective of developing more vehicle software internally, up to 60 percent by 2025 from the less than 10 percent it does now. To achieve this, it has created an independent business unit Car.Software that will eventually be the central group developing the software across VW Group vehicles. The unit has about 5,000 digital specialists working in it today, double that by 2025.
Chip maker Nvidia's CEO Jensen Huang takes the subscription idea a step further. He envisions auto manufacturers offering future EVs at cost which provide basic driving features, and making their profit on the sale of customized services via subscriptions.
Similarly, Toyota has created a holding company, Woven Planet Holdings, Inc. to "focus on a more agile 'software-first' development process and Software Defined Architecture for future Toyota vehicles," which will bring more software development inside the organization. BMW and Volvo have also publicly said they plan to take over more of their own software development.
One question is whether these and other OEMs can really transform themselves into software-first companies, especially given vehicle software and electronics expertise has traditionally resided at their suppliers. The other question is how these same suppliers will respond to OEMs increasingly turning from customers to competitors? Suppliers are not sitting still in their development of advanced EV technologies, especially in software.
Another potentially disruptive wild card includes the entry of a digital dynamo like Apple into the fray, which it is always rumored to be on the verge of doing. If it does, it could make the EV poker game a lot more interesting. In the past year, Chinese internet giant Baidu and Taiwanese electronics manufacturer Foxconn have both decided to get into the EV game. Sony is sniffing around, which may spur Apple or perhaps other tech companies to finally join in as well. Even then, as Tesla found out, transforming from a tech company into an auto company can be hell.
And speaking of Tesla, against which nearly every EV player measures themselves against both financially and technologically, it continues to innovate. Recently, it has decided to move away from its archetypal skateboard chassis to a structural battery pack design, which will undoubtedly will cause some soul-searching in other EV competitors who imitate Tesla.
But if the hype is to believed, Tesla may be finding itself in the future challenged by EV start-up Lucid Motors beginning next year. If true, it could also mean a second "existential threat" OEMs will have to confront. It is likely there are other worrisome rivals within all the EV start-ups popping up around the world.
How the planet's largest, high-stakes poker game will play out in an industry undergoing a once-in-a-century revolution is anyone's guess. One thing is guaranteed; it will be the most intriguing contest to watch as players vie to create an automotive industry radically different than today's.
- How Software Is Eating the Car - IEEE Spectrum ›
- Electric Vehicles Aren't Taking Over Our Roads as Fast as Hype ... ›
- Tesla Inside: The Ultimate Vintage Land Rover Retrofit - IEEE Spectrum ›
- The EV Transition Explained: Reshaping Labor Markets - IEEE Spectrum ›
- Hyundai Making (Tesla) Killer EVs on the QT - IEEE Spectrum ›
Robert N. Charette is a Contributing Editor to IEEE Spectrum and an acknowledged international authority on information technology and systems risk management. A self-described “risk ecologist,” he is interested in the intersections of business, political, technological, and societal risks. Charette is an award-winning author of multiple books and numerous articles on the subjects of risk management, project and program management, innovation, and entrepreneurship. A Life Senior Member of the IEEE, Charette was a recipient of the IEEE Computer Society’s Golden Core Award in 2008.