Will there be a significant "first-mover" advantage in self-driving cars? Will the first company (or group of companies) to bring the technology to market enjoy significant long-term advantages over rivals that lag?
For a long time, I didn't think so, for reasons I'll explain in a moment. But General Motors (GM -2.90%) president Dan Ammann recently made an argument that has me thinking I'm wrong -- that the first (and maybe the second or third, but probably not the tenth) company to put lots of autonomous vehicles on the road will indeed have a big advantage over its rivals for years to come.
Rethinking the advantage for the first-mover in self-driving
Almost a year ago, I argued that the first company to bring self-driving vehicles to market probably won't enjoy a big advantage. That argument held that because there were several serious, credible efforts underway to develop systems for self-driving vehicles, the technology was likely to become a commodity soon after its introduction -- and thus there wasn't much value in being first to market.
I thought that was a strong argument at the time. It makes sense if you look at how new automotive technologies (antilock brakes, adaptive cruise controls) have become available in the past. But it missed something important: Like other machine-learning systems, autonomous vehicles learn from experience. The more hours the system racks up on the road, encountering real-world conditions and challenges, the better it gets.
And -- this is the key -- these are connected vehicles: Lessons learned by one vehicle are instantly shared with all of the other vehicles using the same system. The more vehicles on the road, the faster the system learns.
As Ammann sees it, that means the first company to deploy self-driving vehicles at scale will have the opportunity to build up a huge competitive lead. Here's why.
Why GM thinks the first to market will win big
For several years now, GM executives have said that the company will bring its self-driving system to market "when it's ready," refusing to supply a specific date.
In a presentation last week, Ammann explained what "ready" means: GM's system will be ready when it's safer than a human driver in a complex urban environment, while reliably getting passengers to their intended destinations. Right now, GM thinks that could happen in 2019 -- if the system continues to improve at its current rate:
But as GM sees it, "safe" doesn't mean that the system is a finished product; there will be plenty of room for improvement after that safety goal is surpassed. Kyle Vogt, CEO of GM's Cruise Automation self-driving subsidiary, used this slide to explain:
GM's plan is to bring its self-driving system to market in ride-hailing services operating in dense urban environments. In that context, Ammann argued, the quality of the self-driving system used by a given ride-hailing service will be a significant competitive differentiator. As Ammann explained, that's a big change from the way we think of services like Uber and Lyft today:
Today, the [ride-hailing] product is relatively undifferentiated and it's relatively static. It doesn't change much over time and it's not much different between different providers. There's actually a common pool of supply for the various networks. Most drivers are driving for a number of networks, there are pretty low customer and driver switching costs on either side of that, and the cost model is relatively static.
That's why, GM believes, ride-hailing with human drivers didn't turn out to be a winner-take-all business. But Ammann argues that things will be different when the cars are driving themselves, because the quality of the ride and the cost of the service will both improve as the system improves.
And which service will improve most quickly? The service that is able to deploy a whole lot of self-driving vehicles before anyone else.
"If you have a product that's continuously improving," Ammann said, "you can open up a significant gap versus competition."
What this means for GM and its rivals
Of course, GM's argument is that it is better positioned to "win" on these terms than any other company, because not only does it have an advanced software effort, it also has hardware -- the car, and key sensors -- ready for mass production immediately. GM's self-driving ride-hailing vehicle is a modified version of its Chevrolet Bolt EV; it can be built by the thousands on the Bolt's existing assembly line as soon as GM's self-driving software is deemed ready.
While it's possible that a rival effort has more advanced self-driving software right now, companies like Alphabet's (GOOG -1.48%) (GOOGL -1.38%) Waymo subsidiary don't have the ability to mass-produce a self-driving car on their own. The companies that do -- other automakers -- either lag on the software front, haven't yet completed development of a suitable vehicle, or both.
Could Tesla (TSLA -5.88%) beat GM? While Tesla was seen as an early autonomous-driving leader, its recent struggles suggest that it may need to rethink its approach (notably, its refusal to use lidar sensors) before it can deploy a safe, credible, fully autonomous system. Given Tesla's growing list of other challenges, I don't see that happening before 2019.
If you buy GM's argument, there's a strong case for believing that GM could be a dominant player in the self-driving world of the not-too-distant future.