hey all you people

Frameworks for Estimating Large-Scale Autonomous Vehicle Deployments

August 10, 2019

Now that autonomous vehicle (AV) providers are deploying their vehicles on limited routes, the natural question for industry observers to ask is: when will I be able to call an AV to pick me up from wherever I am and drive me wherever I want to go? or when will I be able to purchase an AV for myself?

In other words, how long until there is a general purpose Level 4+ AV that can navigate long distances and unfamiliar routes?

One of the ways I’ve begun to think about a timeline for this kind of large scale deployment of Level 4+ AV’s is in terms of different macro-frameworks centered around big, unresolved questions. The reasons I think it’s useful to think about Level 4+ AV deployment timelines in this way is:

  1. It can help illustrate why, even if the technical challenges are solved, there are still several hurdles left before we transition into an “AVs-everywhere” state of the world.
  2. It can help those of us involved in the AV industry think about where to focus our efforts in order to expedite the deployment process.
  3. It implies there might be a particular sequence of events that must occur and that the arrival of ubiquitous Level 4+ AV’s will necessarily be much more incremental than some might think.


In my mind, the most relevant frameworks for estimating a timeline for the large-scale deployments of AVs are as follows:

  • The Engineering Framework (Technical). Does it work?
  • The Cost-Per-Mile Framework (Economic). How much does it cost to use at scale? Does Waymo have to charge me $2.50/mile for a ride that otherwise costs $0.50/mile if I drove myself?
  • The Replacement Cycle Framework (Behavioral). Do consumers care enough to abandon their existing vehicle before the end of its full “functional” life? Will people be willing to write off their largest non-home asset?
  • The Consumer Confidence Framework (Psychological). Are people actually comfortable riding in them?
  • The Protectionism Framework (Trade). What happens if Honda deploys a general use Level 4+AV before GM? Will countries allow access to the clear winners?
  • The Infrastructure Framework (Fiscal). What in the current transportation system needs to be repurposed? How much will it cost and how quickly can funds be approved?
  • The Detroit-Silicon Valley Framework (Industrial). Will the two epicenters of AV production work together or in opposition?
  • The Regulatory Framework (Governmental). What are the laws and who owns them? How long before the Federal Motor Vehicle Safety Standards (FMVSS) are updated to reflect Level 4+ autonomy?

Simultaneous vs. Sequential Development

One of the interesting things to think about is that while there is an assumption that the above frameworks and the answers to their associated questions are evolving simultaneously, some might more appropriately be viewed as sequential gating mechanisms that have to be addressed in a particular order.

For example: local governments won’t be able to fully repurpose infrastructure to service ubiquitous robotaxis until the technology and performance of the AVs is sufficiently developed. Likewise, OEM’s won’t want to engage in mass production until they’re confident that the regulatory approvals are in place.

In this sense, the above frameworks could be transformed into a series of gating questions, each of which depends on the resolution of the preceding question. Here’s what I think that sequence might look like:

Sequential vs. Simultaneous Development

Another way to look at it would be as follows:
Engineers are working on perfecting the technology. But regulators might not want to move until the engineers have reached a certain stage of development — and manufacturers don’t want to move until the regulators have written all the regulations.

For their part, consumers will have to wait for the manufacturers to start mass production before they make up their minds about how much they trust AVs and whether it makes economic sense to rely on them. And only after the first set of large scale deployments in one country will other countries decide whether they’ll allow AVs that compete with their own domestic auto industry.

As a result, if we assume that each stakeholder in this “AV deployment chain” relies on the preceding stakeholder, the answer to the question “how long until ubiquitous autonomy?” is not as straightforward as “maybe n years.”

Instead, the answer is closer to“however long A has to wait for B, who depends on C,” and so on down the line.

This exercise is in no way supposed to be a “bear” case for autonomy. I’ve spent the better part of the past two years working on AVs, and I continue to think they will be the most consequential unlocking of economic and social value we’ve seen in several decades.

Instead, this exercise is more meant to illustrate that while there is an assumption that the large, unresolved questions around AVs are being resolved in parallel by different stakeholders, there are reasons to think that a certain sequence will have to be followed.

This has the natural effect of potentially slowing down the deployment and commercialization process for the ubiquitous, robotaxi-like vehicles that we all like to imagine. Luckily, some of the smartest teams in the world are working to make this future a reality. And I for one can’t wait to see how it all plays out.