Does Uber have any special advantages in the world of autonomous vehicles?
I received the following question in response to my article on Uber’s Master Plan. I reference the model developed in my answer and so I encourage you to read the original piece:
I was talking with [name of famous economist omitted] the other day about whether it’s clear Uber has any special advantages once the world moves to autonomous vehicles. We weren’t so sure it would. What do you think?
This is a tough question to answer because of how unclear it is how the intervening years will unfold, and because everyone has their own ideas about how long it will take. What do we know with reasonable certainty?
- Self driving cars will be rolled out
- Regulators will regulate
- A lot of players, big and small, will be competing or will have competed for a piece of the autonomy value chain
Pretty straightforward. What about the known unknowns?
- The number of self-driving car tech platforms
- The state of common automobile design (regarding its relevance to manufacturability, maintenance, lifecycle, upgrades, and recycling)
- The prevailing fuel source (also regarding its relevance to fleet operations)
- The state of the value chain (where is the most value captured, what is commoditized, and what doesn’t really matter?)
- The influence of regulators
- The consistency of regulations
- Public perception of self driving technology as a whole versus specific perceptions about each self driving platform
I could keep going, but we are going to experience diminishing marginal returns pretty soon. But do check out this list:
With so many unknowns I view this question like asking whether the internet will give retailers a sustainable competitive advantage, if the question was posed right after Al Gore invented it.
It’s crucial to understand that the more transformative potential a technology has the more uncertainty there is about how it will evolve and change the competitive landscape.
However, all of this prefacing is not a cop out. I’ll answer the question with an initial question and two scenarios, which will hopefully be insightful in the context of the uncertainty. Just keep in my mind my own opinions on this are in constant flux, and I am eager to hear others.
Question: Does owning the autonomous driving technology provide a sustainable competitive advantage?
Let’s start with this retweet of Tren Griffin.
In the value chain, and particular in platform economics, value can accrue to bottlenecks where substitutes are limited. This is the lesson from Microsoft’s success in (particular) the 90’s. They identified a section in the value chain that sat above a commoditizable supply where they could leverage network effects to drive out substitutes, capturing significant value.
Is this an analog for autonomous driving? Maybe, but I don’t think so.
Arguments for:
- Fleet learning provides network effects that improve the quality (safety) of an autonomous driving platform
- Information coordination is easier inside a closed system, increasing efficiency of autonomy within an autonomous driving platform
Arguments against:
- Closed fleet learning will likely see diminishing marginal returns as machine learning algorithms, neural network hardware, and datasets improve, and sensors proliferate
- Regulations will be placed on platforms for competitive and safety reasons which may require levels of information sharing that diminishes certain competitive advantages
The lists are obviously bigger, and please send me your suggestions, but these are the most instructive in my mind at the moment. The arguments against are compelling, and lead me to believe telecommunication and internet routing protocols are a much more relevant analog than operating systems.
I don’t know enough about the telecommunications / internet routing value chains to be definitive on this, but I doubt it bodes well for autonomous driving protocols in the end state.
Why do I say end state? As autonomous driving technology develops over the coming years those who are in control of the leading platforms will likely have a temporary competitive advantage which they could use to build a defensible position, which brings me to my two scenarios.
Scenario 1: Uber as everything we know about it today. Essentially this napkin sketch with investments in autonomy and maps thrown in:
Let’s make the following assumptions around this model:
- Uber has begun to replace the majority of its driver-partners with autonomous cars by partnering with at least three automotive companies to purchase cars with their tech integrated at cost in exchange for those companies being allowed to sell the tech to consumers (but not licensed for use in non-Uber fleet operations).
- There are several autonomous tech platforms, some of which can interface with cloud service providers so that anyone with a little bit of capital can purchase and configure their own self-driving fleet. For the sake of argument lets say there is at least one that is bolt on (think comma.ai), one that is licensed to automakers by a tech giant (Alphabet), and one licensed by a consortium involving HERE, Intel, and Mobileye. Tesla has their own but does not license it.
Does Uber having strong existing operations and consumer base and the ability to source autonomous vehicles at cost give them defensible competitive advantages when local fleets can be set up by players with sufficient capital and cojones?
No, I don’t think so. This represents an “as a service-ification” of ridesharing, which inevitably drives down barriers to entry. Yes, the unit costs may be higher for micro- or local fleets, but entrepreneurs have a tendency to ignore poor unit cost models.
On the flip side, does this ability to pull software and hardware off the shelf to build fleets pose a structural threat to Uber? For the reason outlined in Shane Parrish’s tweet, I don’t think so. I would not be surprised to see Uber remain dominant in this scenario. But, I don’t see them achieving a strongly defensible position from which they can sustainably harvest value as Amazon has started to do, which brings me to scenario 2.
Scenario 2: Uber as it might be, if it executes like I think it should.
Let’s start with my model:
Let’s imagine in this scenario the following has happened:
- Uber has begun to replace the majority of its driver-partners with autonomous cars by partnering with at least three automotive companies to purchase cars with their tech integrated at cost in exchange for those companies being allowed to sell the tech to consumers (but not licensed for use in non-Uber fleet operations).
- There are several autonomous tech platforms, some of which can interface with cloud service providers so that anyone with a little bit of capital can purchase and configure their own self-driving fleet. For the sake of argument lets say there is at least one that is bolt on (think comma.ai), one that is licensed to automakers by a tech giant (Alphabet), and one licensed by a consortium involving HERE, Intel, and Mobileye. Tesla has their own but does not license it.
- Uber has developed mapping capabilities in that each of their cars is constantly mapping its surroundings, maintaining maps that the Uber platform uses.
- Uber has rolled out its Uber Prime product, driving a majority of its users to single-home within the Uber network and shift some of their consumer goods purchases away from Amazon Prime. The sophistication of the Uber network and the resulting benefits to consumers make it very hard for municipalities to regulate Uber out of their market (what electorate would be ok with their elected representatives barring Amazon from operating in their state/city?).
- Uber is operating subsidized transportation networks for local municipalities all across the country, and has built an harmonious relationship with many legistlators/regulators as a result.
- Uber manages service and maintenance of its fleet internally by building service centers in each market.
- Uber manages fuel operations through its service centers, and is driving down those costs by negotiating large contracts. Fuel type in this prediction is unknown, but interesting to think about.
- Uber is operating an autonomous logistics fleet using technology acquired in its acquisition of Otto. This means transportation of food for Uber FRESH (future grocery delivery service), consumer goods for Uber ESSENTIALS, and parts needed for service and maintenance of their fleet are all being transported by Uber autonomously. Warehouse on wheels concepts are relevant here.
Capturing this much of the value chain may seem like a tall order, but Amazon’s to do list would have seemed pretty incredible back in 2000. I am not suggesting this will happen, but for the sake of argument it is not outside the realm of possibility.
If Uber is able to evolve into this system then I do think autonomous cars provide a key piece of a very defensible network. More than anything else, the cost advantages of owning so many parts of the value chain are significant. Additionally, much of an autonomous network relies on information coordination, which can be easier within a closed network.
At the moment, I think two of the biggest competitive threats in the scenario are Amazon and regulators.
It should be clear that my proposed master plan for Uber involves pulling business away from Amazon, which has steadily moved closer and closer to local “just in time” delivery of consumer goods. If I were Amazon I would (first wait until Lyft’s market value dropped as far as possible and) think strongly about acquiring Lyft to integrate them into the Amazon logistics network. I suspect they have several creative alternative ideas floating around, but this is certainly a good one.
Regulators are also a major uncertainty, for many reasons. The one in particular I want to point out is with the concept of transportation access as a right, not a privilage. Being barred from Amazon due to abuse (ex. too many returns) would be a major inconvenience. Being barred from using the dominant local transportation system would be much more than an inconvenience.
Uber, like any platform, must self-regulate. As its dominance grows, and particularly as substitutes diminish, Uber will be challenged to effectively self-regulate when the consequence of access bans exceeds inconvenience. For more insights into how platforms self-regulate, I highly encourage you to listen to this Andreessen Horowitz podcast:
There is one last point I would like to make regarding execution.
I’m not yet the history buff that Marc Andreessen is, but I’ve been unable to think of any compelling historical examples of large asset light companies that have transformed themselves into asset heavy ones. While they may find creative ways to get around needing to do this, odds are they will load their balance sheet with cars, parts, and service centers. This kind of shift is massive, and will stress Uber’s leadership and structure to the limit. I am very interested in any thoughts on this.
There are many places to pick holes in my assumptions and analysis, and I encourage you to do so (and send them to me!). As already mentioned, my thoughts on this are constantly evolving, as they should considering the changing technological and competitive landscape. I hope this analysis has been thought provoking.
