It feels like a magic carpet ride. Sitting in the back of a self-driving car, I am dazzled by how flawlessly the vehicle negotiates freeway onramp merges, stops before a pedestrian crossing in a school zone, and smoothly avoids a careless human driver turning into its lane at an intersection.
In the 30 years that I’ve been in the technology industry, the challenge of making a car drive itself has been one of the most complex and multifaceted ones I have experienced. It takes an intricate blend of artificial intelligence (AI), physics, semiconductor technology, mechanical engineering, optics, and software that all come together to do something we humans find almost to be second nature.
When innovators get this technology right–and they will–the impact on society will be huge, and the mammoth businesses will be transformed. The automotive industry represents 3.5% of U.S. GDP, and the freight transportation sector is a whopping 9% of the entire U.S. GDP. The combined revenues of U.S. auto makers and suppliers alone are nearly $500 billion, and that excludes manufacturers based outside the U.S.
But this transformation will take some time. As magical as my most recent driverless ride was–and in just one year the experience has improved dramatically–consumers will not trade in their SUVs and sedans for driverless cars anytime soon.
I’ll get to the ongoing technological and business-model hurdles in a minute. Geography poses a more practical and immediate challenge: In places such as Southern California or Arizona, where roadways are built in grids, pedestrians are scarce, and the weather is generally good, consumers could see self-driving cars materialize first. In Boston and other older cities, driverless cars literally face more roadblocks. And since cars are designed to move across multiple geographies, it is hard to sell a consumer vehicle that only works in, say, Phoenix, but can’t take a road trip to San Francisco.
Most consumers’ first experience in a driverless car will probably be part of a network run by a ride-sharing company such as Uber or Lyft, which could program the cars to operate strictly within localities that are suited for autonomous vehicles, forming a dense network of units that could operate around the clock.
The economic benefits to these ride-sharing networks, and ultimately to users, is huge: Companies pay $1 to $2 per mile to their drivers–about 75% of revenue. In contrast, self-driving vehicles cost about 10¢ to 20¢ per mile to operate. If operating costs could be reduced by some 90%, transportation could become cheaper for consumers and more profitable for companies. And the ride-sharing companies’ challenge of finding qualified drivers could be alleviated.
But who will build the technology that will drive these new cars? I’m an investor in Aurora, a young autonomous vehicle company that is developing software and hardware to build a robotic “driver,” so I’m inclined to think startups–and not big auto makers–will win the race to develop the systems needed to make driverless cars go. These so-called “full stack” companies–Aurora rivals Waymo and Zoox are also building solutions that include hardware and software–historically have thrived in instances where control of the entire architecture is a competitive advantage. Think back to the early mainframe and minicomputers: the Macintosh, the Blackberry, and the iPhone. All of these products combined various technologies, but one company was the primary architect of the overall system.
Building a commercial full-stack system requires the combination of massive resources, a huge range of technical talent, and a dose of (very scarce) experience in the sector. While the major players won’t disclose the money they are investing in research and development, it’s pretty clear that they are spending hundreds of millions per year. So capital is, in fact, a major competitive advantage, and there will not be many companies that will be able to afford to play in this market.
The auto makers themselves have the resources to invest in a full stack of hardware and software, and indeed, General Motors and Tesla are among some of the car makers developing autonomous vehicles in-house. These giants certainly have relationships with many of the component makers who help make up the driverless ecosystem. But I think the economics are tricky: Car makers won’t be able to amortize the cost of development (which will be in the billions) across other car makers; after all, no one really wants to buy technology from a competitor. Some manufacturers, such as GM, are also expanding into ride sharing, perhaps both as a way to build direct relationships with consumers, and to create a market for their own driverless cars. This is another place where the disruptors have the edge: GM’s network, now in a handful of cities, will have a hard time competing with established players such as Uber and Lyft.
My prediction: Car makers will continue to build vehicles—only a growing portion of their portfolio will be trucks and cars that integrate with autonomous technology. After decades of selling cars based on their performance and appeal to drivers, they’ll need to figure out how to build and market cars that appeal to passengers. And they’ll have a huge opportunity to build businesses around the very important work of servicing and replacing autonomous vehicles. Some in Silicon Valley are underestimating how complex this piece of the puzzle is, but the task of designing, building, and servicing millions of vehicles that live 10-plus years in tough environmental conditions is an industrial feat that will take time to replicate.
The timeframe for all this to become a reality remains difficult to predict. (We haven’t even dived into the complex safety and regulatory considerations–how do states distribute a “driver’s license” to a robot?–that could slow or accelerate the mass introduction of driverless cars.) There are still many variables and twists in the road. Some market players are committing to “commercial” offerings this year. While marketing professionals at these companies will surely paint a pretty picture, the more likely scenario is that we will really start to see deployments in “easy” cities in the 24-month horizon. The more complex cities for autonomous vehicles, like New York or Boston, may have to wait as long as three to five years before the technology is ready for them.
It is important to remember the magnitude of the impact of self-driving on society. In the U.S. alone, 37,000 people died in automotive fatalities in 2017–a number that has been rising because of texting and driving. Self-driving cars have some challenges, but they are not distracted by texting. As urban populations increase, so do commute times, and we will be able to reclaim an enormous amount of productivity and fun when we are not driving. The cost of moving around both people and goods will be significantly reduced. We’ll likely experience less traffic, and just generally be happier. So while the road ahead for autonomous vehicles is longer and more complex than we’d like, this is a once-in-a-century kind of transformation that is happening right in front of our eyes. Any person who believes that technology makes our lives better should be keenly interested in its progress. I know I am.
Mike Volpi is a partner at Index Ventures, where he invests in autonomous vehicles, artificial intelligence, and open-source companies. Mike is an investor and board member for Aurora and also sits on the board of Fiat Chrysler Automotive.