What the PC and Smartphone Revolutions Can Teach Us about Autonomous Vehicles
John Blair
Automotive industry leadership is going all-in on technology that could reshape their businesses. How? History provides some clues.
Experts and auto industry insiders argue that autonomous vehicle (AV) technology will soon not only become a key and differentiating feature in automobiles, but also potentially transform the industry’s basic dynamics by augmenting the rise of Mobility as a Service (MaaS) providers and thereby reducing the demand for personal car ownership.
That has set the stage for an increasingly competitive race to market. Automotive original equipment manufacturers (OEMs), MaaS providers, and Tier 1 suppliers are working intensively to develop systems that deliver AV capability to cars. It’s too early to predict who will win that race, but we can look to previous technical revolutions for insight into the likely competitive dynamics.
The technical architecture of AVPs—a distributed software system
Physicist Sir Earnest Rutherford is reputed to have said, “There is only physics; all the rest is stamp collecting.” I propose a modern-day equivalent: there is only software; all the rest is infrastructure. This is certainly the case for AV platforms (AVPs).
AVPs require advanced hardware, including sophisticated active sensors such as Lidar, fast processors, and advanced wireless communications. The necessary software, however, presents an even greater technical challenge—for example, accurately and reliably synthesizing an understanding of the vehicle environment, both in the moment and projected far enough ahead to navigate the road, other traffic, and pedestrians safely.
Automobile OEMs and their Tier 1 suppliers have long had world-class engineering capabilities, as shown by the remarkable improvements in fuel economy and engine power since 1980, achieved while average vehicle weight has increased.
Given this track record, why aren’t AVPs simply a new engineering challenge for the industry? There are three crucial differences:
AVPs are distributed systems. AVPs must include complementary cloud-based support systems providing local, metro and global support—delivering data on everything from traffic light status up to high-level maintenance and management. This requires a distributed software system of global scale.
AVPs are ecosystem platforms. AVPs have multiple cooperating elements—architecturally, they have much in common with operating system platforms, where system software, application software, and hardware vendors come together to deliver the system’s functionality.
AVPs have high regulatory and certification barriers. Level 4 and Level 5 AVP capability will require new certification programs and regulatory systems. Level 4 allows the AV to drive the car within specified restrictions (e.g., in designated areas or on designated roads). Level 5 removes those restrictions.
These three elements differ in design and deployment from traditional automobile software, which are largely in-vehicle systems, proprietary to OEMs in close collaboration with one or a few Tier 1 suppliers, and subject to well-established and well-understood performance and certification metrics.
But we have seen evolutions of this kind of platform ecosystems before: the personal computer revolution that began in the 1980s, and the transformation of the mobile phone industry since 2007. In both cases, competitive dynamics were fundamental drivers to the ultimate industry structure.
Lessons from previous software revolutions—PCs and smartphones
In both PCs and smartphones, network effects—the mutually reinforcing feedback between user adoption and developer adoption—plus enormous fixed costs rapidly led the competitive landscape to collapse into global duopolies: Windows and Intel (Wintel) versus Apple in PCs, and Google Android versus Apple in smartphones.
Those same competitive forces will buffet AVPs—but with a twist. The same drivers to global scale will apply, but certifications and regulations (and traffic rules) will be locally idiosyncratic.
Additionally, profits will be highly asymmetric. In the Wintel world, Microsoft and Intel accrued the great bulk of PC industry profits; PC hardware vendors struggled to differentiate themselves and earn sustained profits. Similar forces have been at work in today’s smartphone industry.
Finally, a winning ecosystem strategy will be crucial. Unlike traditional automobile software, AVPs will require enlistment of ecosystem partners at multiple levels in the technology stack. The resultant ecosystem business models are as yet undetermined—and will be critical to vendor success.
Implications (if history is any guide …)
Perhaps business strategists—like army generals—are doomed to fight the last war. But lessons for AVPs can be gained from previous software platforms.
First, and crucially, bottleneck assets—assets that are in fixed supply and controlled by one ecosystem participant—will emerge. In the Apple smartphone ecosystem, for example, the iOS operating system is the key bottleneck asset that anchors Apple’s profits on the iPhone hardware and, increasingly, on its network delivered services. Participants in the emerging AVP ecosystems will need to identify and focus on the bottleneck assets that will emerge.
Second, first-mover advantage has historically been weak until a dominant design emerges. Apple was not a first mover in smartphones; Intel was not a first mover in PC processors. Until dominant designs emerge, the AVP design must “float”—it must evolve rapidly based on constant user feedback and rapid prototyping.
Finally, it’s likely that no more than two AVP ecosystems will be successful, and as a consequence those ecosystems will likely be cross-OEM. Smaller, more atomic AVPs may emerge and provide advanced AV services—for example, for one of the large global OEMs. But the platform characteristics discussed above suggest that even those will need to fit within larger, cross-OEM ecosystems—and develop the appropriate complementary business relationships.