Don’t Mistake Ridehailing for AV Ridehailing

When Uber introduced surge pricing in 2011, many condemned it as merely a price gouging tactic. But consumers eventually realized its primary benefit: unlike fixed supply networks such as hotel rooms or taxi medallions, Uber had the ability to add more vehicles to its network almost instantly as demand spiked. This meant that more customers could reliably get more rides when they needed them.

Having a dynamic supply of vehicles had a big impact on Uber’s business and, more importantly, became the dominant model for how other ridehailing companies managed their networks. The customer demand drove the supply, not the operator or medallion holder.

This fundamental idea also happens to utterly confound how we might design these networks with automated vehicles.

During a typical day, ridehailing’s low-demand ‘base’ and high-demand ‘peak’ loads ebb and flow (rising during commuting times and social hours, waning otherwise). Since this supply is made up entirely of driver-owned vehicles, Uber doesn’t pay for the downtime of unused vehicles — that cost is borne by the independent driver. The driving function is human-piloted on base load and human-piloted on peak load too (I describe this as Human Base, Human Peak).    

Do AVs break ridehailing’s dynamic supply?

But what happens when you introduce a fleet of fixed-cost automated vehicles into a ridehailing network? It means that the most complicated part about a Waymo isn’t the driving task but the ownership structure. As former Uber board member Bill Gurley recently asked, should Waymo build its network to base load or peak load? Perhaps a new model can emerge, one in which you have a base of robots that add humans during peak times: Robot Base, Human Peak.

In a ‘Robot Base, Human Peak’ network structure, you would want the minimum number of robot vehicles (base load) but no greater. Supplementing this during peak load times would be human-piloted (and critically, human-owned) vehicles much like today’s Uber supply. When demand wanes, the humans take their cars home.

Doing this jumbles the customer experience a bit (‘am I going to get a robot or a human once my ride is accepted?’). It has complex long-term challenges because it likely shrinks the total supply of human-owned vehicles, as a driver’s revenue opportunities decline. Many will churn out of the driver pool entirely as they look for work other than driving.

A third model: Tesla’s ‘Dynamic Ownership’

Tesla could disrupt this model significantly, assuming Full Self-Driving technology delivers on its promise.

Tesla could leverage a company-owned fleet of robotaxis while supplementing the network with individually-owned Teslas nearby. This creates an all-Tesla experience with automated vehicles handling both base and peak demand loads: Robot Base, Robot Peak.

By enlisting individual owners as future robot minders, Tesla could have the most durable ridehailing network of the robot era. What’s fascinating is that this competitive edge isn’t due to a deep technical breakthrough, but rather because its fleet would have a mixed ownership structure. A fleet with ‘dynamic ownership’ characteristics allows for peak load management of robots like we have in today’s ridehailing networks. Peak downtime is borne by an individual, not a fleet.

Aggregators like Lyft and Uber could, of course, also welcome individually-owned Tesla FSD vehicles to their networks. However, this assumes that Tesla vehicles will be allowed to join non-Tesla networks and that they would be incentivized to do so beyond the Tesla network. Tesla could lock in FSD vehicles to their network with a range of incentives (Supercharging credits, software unlocks, etc.) or contractually.

While the pending launch of Tesla’s company-owned robotaxi fleet is likely the most consequential AV moment of 2025, how the company handles a supplemental supply of individually-owned vehicles is the most impactful for the global ridehailing market.

Whoever can successfully build and manage a robot fleet with dynamic ownership can take over the world.

Thank you to Jeff, Alex, Sarah and Nick for helping me think through and edit this post.