Who will win the robotaxi race? Tesla, Waymo, or Uber?

Tesla's Robotaxi Day is almost here. On Thursday, Elon Musk will unveil a vision for a ridesharing service powered by his company's autonomous vehicles.

His main competition on the technology front is Waymo, owned by Google parent Alphabet. On the ridesharing side, Uber is the one to beat.

Who will win?

Strangely, the company with no autonomous vehicles could prevail.

Uber's flexible, inexpensive business model
Uber stopped developing driverless cars years ago. Instead, it redoubled efforts to create a dense, two-sided marketplace of riders and drivers in the US and several other countries.

It took well over a decade to get this up to scale. Today, consumers open the Uber app because they know there will likely be loads of cars nearby to pick them up. And drivers come to Uber because they know there will be many riders waiting and willing to pay.

Creating network effects like these is really difficult. But once they're established, they are very hard to compete with. This is reflected in Uber's market value, which stood at $160 billion this week, close to a record.

The other benefit of running a marketplace is that you don't have to pay for much. Uber doesn't buy and maintain cars. It doesn't really employ drivers in most jurisdictions. And if demand drops, Uber's costs fall too because it doesn't have to share as much revenue with drivers.

This flexible, capital-light business model works. In the second quarter, Uber generated $1.7 billion in free cash flow.

A robotaxi future
What happens if Tesla and Waymo driverless cars replace human drivers? One scenario is that these companies build their own ridesharing networks. Without having to share revenue with human drivers, they could offer cheaper rides and undercut Uber.

This understates how expensive and complex it will be to run a huge autonomous ridesharing fleet. It also undervalues Uber's existing network and its direct relationship with millions of users.

When investors started worrying about this potential threat to Uber earlier this year, Bernstein analysts did a deep-dive into the technology and the emerging business models.

What they found is that running a robotaxi fleet is expensive and loses some of the attractive elements of Uber's approach.

"There are real challenges around scaling a pureplay autonomous network, including the underlying technology, regulatory approvals, building a large-enough supply-base to run a high quality network (especially during peak hours when consumers need the service the most), and ultimately getting customers to adopt the service," the analysts wrote in a recent research note. "Along the way, we think Uber and Lyft can continue to play the role of demand aggregator and strategic partner, already operating networks at scale."

The fixed costs of robotaxi fleets
The Bernstein analysts estimated that about 350,000 to 400,000 robotaxi vehicles would be needed to effectively replace the current rideshare networks in the US.

These cars can cost roughly $150,000 each, partly because they are loaded with high-end sensors. That means just getting the cars ready would cost more than $50 billion, according to the analysts. (Tesla could cut this expense drastically because its robotaxis may be much cheaper, but Bernstein analysts worry the core technology might not be as good).

These vehicles will have to be replaced every few years, too. Then there are many other costs that, crucially, are fixed rather than flexible. With Uber's current approach, it doesn't pay most of these expenses:

Maintaining and cleaning the cars and all those sensors
Remote human monitoring assistants who can take over if anything happens
Cloud costs to store all the data from the robotaxis
Parking
Charging the cars
Deadheading
A bigger problem may be "deadheading." This is when robotaxis drive around with no passengers.

With Uber, when human drivers aren't giving customers a ride, it doesn't pay. With robotaxis, most of those fixed costs remain, even if no one is on board.

"The fixed fleet model has its drawbacks," Bernstein's analysts wrote. "We estimate Waymo's network was running with 60-70% 'empty time' in a day and 30-40% deadhead miles as of May."

The uber aggregator
One way to avoid deadheading is to have a steady supply of users ordering rides all the time. Who has amassed the most demand here? Uber by far.

For example, the Uber app was downloaded about 6 million times last month, according to Sensor Tower data. Tesla's app was downloaded roughly 500,000 times, while Waymo got around 200,000 downloads.

Uber is an aggregator of demand. Tech commentator Ben Thompson writes a lot about aggregation theory and how powerful it is on the internet. If you pull together enough demand, that attracts more suppliers, which improves your service even more. New users come, creating extra demand that pulls in even more supply. And so on.

Thompson even wrote a newsletter in 2017 about how Uber CEO Dara Khosrowshahi was an experienced aggregator from his days running online travel agent Expedia.

No matter how good Waymo and Tesla's autonomous technology is, they need distribution. They must get bums in driverless car seats quickly, regularly, and at huge volumes to cover the fixed costs of running their robotaxi fleets.

This is where Uber comes in. It has already struck deals with Waymo and Cruise, another AV company, to have their driverless cars be part of the existing Uber rideshare network.

"The rideshare platforms can be valuable strategic partners for AV players given the challenges associated with scaling a network," Bernstein's analysts wrote. "In particular, we believe that Uber offers valuable access to a large and active rideshare audience, with an already-scaled OpEx base, thus it can be a source of incremental demand and more efficient network management."

An unusual situation for Google
While Musk is used to dealing with expensive, complex hardware, this is a pretty unusual situation for Alphabet.

Google is probably the ultimate online aggregator. Its search engine aggregates most of the western world's questions and desires. When people type in a query, millions of online businesses jockey to provide the answer. Google doesn't pay for most of this content, it just sells relevant ads next to the results. This supports one of the most profitable businesses in history.

Compare this to the robotaxi situation, and the shoe is on the other foot. Waymo is doing all the hard work here. It owns the cars and develops all the sensors and other technology. It must pay to maintain the fleet. If anything goes wrong, Waymo is on the hook.

Meanwhile, Uber is getting new supply to add to its existing network. When this happens, the network usually becomes more useful and attracts even more users