In the collision between consumer technology, cloud computing and cars, the legacy auto companies are lacking key ingredients for the future of the automobile. Here are the vectors for auto disruption.
This post originally appeared on Medium.
The potential is so exciting, you can almost taste it — a world of self-driving electric cars, eliminating 1.25 million annual traffic deaths, reducing massive amounts of climate-altering emissions, urban parking lot real estate turning into parks and ending the dangers of drunk and distracted driving. We all seem to know what we want. But who will bring it to us?
I believe this inevitable transition creates a very clear set of vectors for disruption of the traditional automotive industry and, like other industries transformed by technology (newspapers, travel agents, the music industry, traditional retail, etc.), many new winners are likely to emerge. Let’s examine why.
As every industry becomes a technology industry, its pace of innovation must accelerate. Those of us used to Moore’s Law as the single most-defining characteristic of the information age know that rapid innovation is a constant. If you don’t quickly innovate, your competitors will, and you will get lapped in one tech cycle. Cultures of rapid innovation, while common in technology, are less common in industrial age companies. Think about the length of product cycles in the car industry—new car models take three to five years to develop and launch and then sit stagnant for six years in the market, devoid of meaningful improvement or new innovation. The only way to get new feautures? Buy a new car.
The technology industry will most certainly take a different approach to cars. We think about building and deploying hardware platforms into the market, on top of which we frequently update the OS. The OS enables developers to build thousands or millions of apps which bring new functionality to the user. We know the user expects the product to meaningfully improve over time with over-the-air updates. And when we do need to rev the core hardware, we do so rapidly, typically in one-to-two year cycles.
A good example of this is the phone. Today, consumers are getting rapidly updated mobility features for their cars (navigation, for example), but they are getting it on their phones and not from the traditional auto companies.
Complicating the car industry’s ability to rapidly innovate is the fact that they have largely become system integrators. They develop very few components of a car, instead sourcing almost all key components from tier one suppliers such as Bosch and Continental. Take a look at this graphic:
When so much of your product is designed and built by suppliers, you may lack key engineering leadership within to substantively and quickly produce innovations not generally available to your competitors. While new entrants to the car industry will certainly utilize tier one and two suppliers for much of their components, they will likely choose to produce proprietary innovations in some areas where neither today’s car companies nor their suppliers typically excel. We will get to a few of those areas below.
(I am confident several readers will point out that Apple sources tons of its iPhone components from third party suppliers. While true, Apple designs the most important components themselves—the CPU, the OS and many default apps. This article helps explain why that is so critical. In addition, Apple works deeply with its suppliers to direct development around areas most likely to benefit the end user. The same cannot be said of the majority of the existing car companies.)
Tesla, Google, Apple and other expected new car entrants are centering their designs around electric vehicles. This is true for many reasons, the biggest being that electric motors are much simpler than internal combustion engines. Removal of this complexity, in some ways, is a resetting of the core expertise required to produce a great automotive product — goodbye combustion engine, carburetor, transmission, exhaust, emissions systems and fuel economy management; hello batteries, power optimization, charging systems, and engine controllers. Some of these new capabilities are native to consumer electronic manufacturers who have been dealing with battery life and power optimization for some time, albeit on a very small scale. A bunch of car companies have produced an electric car. But building an entire organization around this and nothing else will produce a focus and an innovation trajectory likely to outstrip that of companies who do it as a side project.
Perhaps the most significant shifting of the automotive tectonic plates is the move to software. The future of the automobile will largely be built by software developers. Yes, existing combustion engine cars have embedded systems with lots of code in them to handle everything from HVAC to automatic transmissions. In fact, the complexity in integrating these many layers of software together is causing lots of consternation at the traditional car companies, given this is not their main areas of expertise. In addition to this, future cars will utilize software in profoundly different ways.
Of course we know that Tesla (currently) and Apple (future) are trying to re-imagine the interface between the driver and car, and their dashboards are (likely to be) gorgeous and vastly improved over the mostly superfluous dials and gauges car manufacturers think we need to see (when was the last time you had to check your RPMs or engine temperature?). Good hardware, software and UX designers will be behind all of that. But future vehicles equipped with ADAS systems and eventually autonomous capabilities will need to make trillions of driving decisions based on lots of sensory data.Vision, LiDAR, sonor and other sensors will combine with real-time streams from the internet, from other vehicles and even from municipal environmental data sources (our portfolio company INRIX is one such data supplier). These inputs are analyzed in real-time, likely with a combination of local on-board and cloud-based compute resources to make driving decisions. Such complex AI systems will be adaptable machine learning systems which continuously refine their decision-making models.
Understanding this makes it less of a surprise that Google leads the way in autonomous vehicle development today. Google’s search engine is an at-scale example of just such a system and much of Google’s core development expertise is in cloud-based predictive systems.
There are two main reasons why legacy auto companies are unlikely to excel in these areas. The first, is that very few of the world’s best AI engineers, data scientists and cloud computing experts work at auto companies today. And while there are certainly talented engineers at these companies, despite the many Silicon Valley-based research centers opened by the car companies in recent years, companies like Google, Tesla, Apple and Uber have been a bigger draw for the extraordinary technology architects and data scientists looking to disrupt the auto industry through software. The second reason is data.
Connecting some cameras and sensors to a Mobileye chip and doing some lane-centering or adaptive cruise control is the easy stuff. To reach truly autonomous driving is much, much harder because the system first has to learn. There are no existing set of rules we can program into a car which will prepare it to anticipate and avoid all hazardous situations it may eventually encounter. Effective autonomous driving systems must use machine learning to develop sophisticated models which can adapt to many different circumstances. Machine learning systems require large data sets to reach optimality.
Do you remember when Google offered “Free 411”? They weren’t doing it to be generous. They did it in order to capture millions of different voices and speech patterns to train the speech recognition systems now used by Google Now. Google is accustomed to using data scale to reach performance levels others cannot match. This is precisely the network effects which allow Google search to still out-perform competitors. Google, with 63% search marketshare, just has more data than everyone else. They see more searches and more clicks than anyone and can train their algorithms accordingly.
The same data scale advantages will affect self-driving cars. This is why Google has driven their current fleet of 48 self-driving cars more than 1.2 million miles — to gather data and train their systems. That slow-moving pedestrian between two parked cars? That’s a hazard. Sunlight reflecting off a puddle of water to your left? Not a hazard.
Tesla, too, is heavily focused on this. Check out this passage from VentureBeat.
The best self-driving cars will be those that are part of the largest network or fleet, sharing data and learning among them. This is a problem for the existing automakers. They have no data. If they were onto this, they would be equipping their current automobiles to gather this data and to train systems back in the lab. (I have heard Uber intends to do this, as their driving footprint is very large.) In addition, almost all of the components for self-driving cars built by legacy auto companies will come from the tier one suppliers. Those guys also have no data. In fact, they are likely to have even bigger problems gathering it, since they have no direct relationship with drivers—our data would not be available to them.
5. Direct consumer relationships
I have written before about how the shift of attention from mainstream media properties to social media platforms requires brands to now have a direct relationship with their customers. One of the biggest vectors for disruption in cars is the existing manufacturer/dealer model. The model where auto companies sell to dealers who, in turn, sell (with terrible experiences) to drivers is an idea who time has come and gone. Tesla, as the first successful direct-to-consumer auto brand, has this right. They don’t have dealers, they have showrooms. They don’t haggle with us over price. (I remember once hearing car companies defend this practice by stating that consumers actually prefer to haggle over price.) How does the industry react to this modern model of engagement, allowing one to actually know and understand one’s customer? They sue to protect dealers.
Modern car companies will not use legacy dealer networks. They will sell directly to consumers and engender long-term relationships with them.
6. Executive Dismissiveness
The final signal that the auto companies won’t bring us the future can be heard from the mouths of the most influential executives at those companies today:
“I think, like so many Silicon Valley techies, that they believe they are smarter than the world’s automobile business, and that they will do it better. No way.”
– Bob Lutz, former vice chairman of General Motors
“There is absolutely no reason to assume that Apple is going to be financially successful in the electric car business. Electric cars are generally money losers. If I were a shareholder I’d be very upset.”
– Bob Lutz, former vice chairman of General Motors
“I have no idea who will be first to market with an autonomous vehicle.”
– Mark Fields, CEO Ford
“We’re in the car business today, and they’re not.”
– Mark Reuss, Product Development Chief, GM (speaking about Google)
You might remember these quotes from one of the co-CEOs of Blackberry:
“[Apple and the iPhone is] kind of one more entrant into an already very busy space with lots of choice for consumers … But in terms of a sort of a sea-change for BlackBerry, I would think that’s overstating it.”
– Jim Balsillie, February 2007
“As nice as the Apple iPhone is, it poses a real challenge to its users. Try typing a web key on a touchscreen on an Apple iPhone, that’s a real challenge. You cannot see what you type.”
– Jim Balsillie, November 2007.
It’s true that Apple, Google, Uber and the many new upstarts working on the future of cars know nothing about the car business as practiced the last one hundred and seven years. The key here is that the future of the car business is going to be dramatically different than the past.
I am confident many of the existing automotive companies will produce cars with autonomous features. And some of them will be quite good. And eventually they will produce some fully autonomous electric cars too. In the meantime, there are many super-strong, thoughtful entrepreneurs with lots of incredible hardware and software experience working to fundamentally redefine what consumers should expect from cars.
The history of innovation and disruption teaches us that, at the point of major technological platform change, new entrants can emerge and take meaningful marketshare and value away from the incumbents. I believe we are on the cusp of such a change. It might make sense to take that transition very seriously. I know we are.