Part III: Integrated vs modular approaches to AV development: comparing Tesla, Waymo & Uber.
Tesla, Waymo and Uber are starting to show their cards. Here’s a closer look at their business models and how each fares in terms of interdependence versus modularity.
With the AV race in motion, Tesla, Waymo and Uber are among the leaders when it comes to deployment readiness. Tesla has already collected data from over 50,000 vehicles and released semi-autonomous features through its existing fleet. Waymo has announced partnerships spanning OEMs, ridesharers and a likely fleet manager / service partner, while opening up test ride seats in Phoenix. And Uber, before it’s fall from grace, inked one of the first pilot deals in Pittsburgh and has built a massive driver / rider base with 5 billion regular rides under its belt.
Over the past few months, we’ve seen these players develop 3 distinct strategies in a market that clearly seems to favor first-movers. With software, hardware, mapping, revenue models and more changing, what are the most important things for these companies to succeed?
- First, we’ll start to unpack each player’s model in terms of strengths and weaknesses
- Second, we’ll do a dive deeper into Christensen’s interdependence vs modularity theory.
This is Part III of 4-part series. If you’re already familiar, skip ahead to Part IV for a look at what it all means for Tesla, Waymo and Uber deployments in the near term. If you missed them, see Part I for more background on AV fundamentals and Part II for thoughts on where AV will go next?
It’s All About Perspective
Each players’ path is very much rooted in their initial vantage point. Their model must address 3 layers of the AV stack: vehicle manufacturing, the AV technology that layers atop that (think lidar, sensors, software, mapping, etc.), and finally the consumer interface / demand aggregator to tap into the first two.
Tesla
Tesla was built as a fully integrated model, necessitated by the complexity of their mission to move to electric vehicles. Note, integration / interdependence denotes a proprietary supply chain from manufacturing hardware to distribution and customer interface, where these components are created in alignment and optimized around performance. Modularity instead optimizes around flexibility and efficiency with open industry standards, where components in the value chain can be interchanged easily.
Employees used Tesla’s interdependent nature to reimagine the battery and redesign car components and the overall car structure around it at a reasonable cost. Now they are trying to take that approach one step further, layering on AV tech with an eye toward the Tesla network in the future.
By moving from the car design to AV software and hardware, Tesla enjoys unique advantages around iterative loops and shared learnings / revenue. Tesla can leverage its existing, early-adopter base to encourage an intermediate wave of Level 3 adoption first (levels explained here) while most competitors are skipping to Level 4. This intermediate step involves pushing semi-autonomous software updates to its existing fleet while most cars today don’t even have this over-the-air capability. Then, the data (now including video clips) from that fleet will enhance Tesla’s Autopilot algorithms, creating a virtuous cycle that encourages further adoption. At some point, this advantage may create such a large gap in performance between Tesla and it’s competitors that others are forced to license to catch up.
In terms of future risks, Tesla may overindex in a decaying revenue model, as their shared fleet still relies on some level of personal ownership and car brand value.
Uber
Uber is approaching AV from essentially the inverse direction as Tesla. Uber has already built a demand aggregator with massive reach, so now they are trying to layer AV technology atop that to avoid disintermediation. They enjoy the advantage of a full fleet, for both data collection and a comprehensive consumer offering as AV slowly accounts for a fraction of total desired rides, backfilled by human drivers.
“It starts with understanding that the world is going to go self-driving and autonomous… So if that’s happening, what would happen if we weren’t a part of that future? If we weren’t part of the autonomy thing? Then the future passes us by basically, in a very expeditious and efficient way,”
— Travis Kalanick, former CEO of Uber
On the risks side, Uber appears to be offloading the vehicle manufacturing, which is widely viewed to be an incredibly hard capability to build from scratch. While that is capital efficient, it removes many of the integrated advantages Tesla will enjoy. Uber is instead beholden to OEM designs and presumes auto manufacturing will not be a key to success.
Waymo
Alphabet’s Waymo formed around building AV technology, and the company recently confirmed that they are permanently leaving the manufacturing game. Instead, Waymo has focused on driving over 3mm miles on roads and 1bn miles in simulation, deeply cutting the cost of their proprietary lidar and protecting their IP (against Uber, coincidentally) along the way. For all other parts of the greater stack, Waymo has assumed a fully modular approach, relying on OEMs for manufacturing at scale and ridesharers / service operators for the consumer-facing interface.
This allows all of these players to maximize brand value at each level and focus on local core competencies while staying capital efficient. In many ways, it most resembles the airline industry, where the market has separated component manufacturers, aircraft manufacturers, aircraft leasing agents and carriers. However, this modular model lacks the tight innovation cycles of Tesla and the clear consumer touchpoint of Uber.
That said, Waymo may be looking to monetize a modular world differently, when you think about the greater ecosystem of Google Ads, Google Maps and Waze. But more on that later.
To Integrate or Not To Integrate?
“There are two ways to make money in business: You can unbundle, or you can bundle.” — Jim Barksdale
These three models present stark contrasts: integration of various levels from opposing angles, with one player pushing full modularity. As such, now seems like a good a time to better understand the implications of these overarching decisions.
Currently, AVs demand a massive step-change in both hardware and software capabilities. In such scenarios where a perceived performance gap remains, an interdependent business model (where the components, manufacturing and go-to-market are under one roof) will typically allow for faster innovation and iteration to close that gap. With AVs, the gap stems from both technology shortcomings and consumer perception. Thus, this gap will likely persist for a long time as companies work to educate and build trust with regulators and consumers through various pilots.
During this time frame, integrated players will fare better than those opting for modular consortiums. This is true for a few reasons:
- Integrated players do not risk outsourcing the most valuable components. While the industry develops and certain elements become commoditized, value capture will likely shift across the supply chain. Case in point: lidar is currently one of the most expensive and scarce technologies to date, but most believe that lidar cost and availability will open up dramatically over time. Integrated players can focus on the holistic product without worrying about where value will shift.
- Integrated players can be agile because they own all of the capabilities. Digital Equipment Corporation used this strategy to dominate minicomputers, as did Apple Computer for the desktop. They reallocate resources internally as needed to push product and capture market-share and learnings. They quickly redesign operating systems, logic circuitry and hardware composition to fit evolving consumer demands. In a first mover market like AV, this advantage amounts to a scaling flywheel: More Level 3 cars on the road → capturing a ton of public data for mapping and algorithm learning → stronger software algorithms and higher safety → more Level 3 demand (cycle restarts). An integrated model is prepared to process all types of data and feedback to adjust hardware and software as needed. A partnership model is simply not integrated enough to realize many of the interdependent learnings and agility necessary.
- An integrated model means more touch points for educating consumers. Take GM’s OnStar’s release back in 1996. The company was introducing novel technology that many consumers did not understand nor realize they needed. OnStar utilized GM’s existing channels (dealerships) and brands to convey the value of their new technology and attract customers. This garnered them early adopters and engagement that they likely wouldn’t have gotten if they had simply opened the technology up to all OEMs. AV will involve even greater customer education and trust-building.
- Integrated players can capture more value through bundling. When all of the features sit under one roof, companies can capture more value for the bundled set than they can for individual components combined. They can use this extra cash to fund further innovation and deliver a truly differentiated product across multiple axes. For example, Apple bundled premium hardware development and a proprietary operating system for the iPhone together, allowing them to capture far more than the sum of the parts would a la the Android’s modular model.
Typically, once the performance gap is resolved — here that likely means safe Level 4/5 autonomy has been achieved and consumers trust it — modular companies reign by bringing down cost and adjusting performance to meet specific customer needs through supplier competition and exploitative innovation. This is indicated above by the shift from functionality and reliability to speed, responsiveness and convenience. For modularity to work, companies must also standardize the interfaces between components. For example, the combination of sensors, lidar and cameras must be clear enough to be specifiable, verifiable and predictable.
However, in AV the first-mover advantages may be so strong, and the time to modularity so long, that other players cannot catch up in time to earn public trust. This would mean that many would fail / consolidate, and others might simply start licensing technology.
Investors seem to appreciate the importance of interdependence in AV, as integrated player Zoox was valued at $1.55bn in their Series A while still in stealth mode. However, Waymo is not alone in the modular world. Most incumbents — OEMs and Apple for example — have opted for a partnership model to manage a dizzyingly dynamic field. Assuming AV is not a winner-take-all market (but may have a few winners-take-most), each model has a stake in the future. We will explore what that’s likely to mean and where the fault lines for integration lay in the final piece, Part IV.
Up next: Part IV to understand how those may affect successful deployment. See Part I for more background on AV fundamentals and Part II for thoughts on where AV will go next?
