Viewpoint: Robotaxis will test Tesla in a trillion ways

Tesla will launch its long-promised robotaxi service in Austin, the United States, in June because it has to.

The countdown began just over a year ago when news broke that the company had abandoned its cheap electric vehicle (EV) project and chief executive Elon Musk responded by announcing a robotaxi unveiling, scheduled for August.

Delayed to October, that event was a flop, with the vehicles confined to a studio lot and guests heckling Mr Musk on when the real robotaxis would appear.

With Tesla’s core EV business having slumped since then, it is Mr Musk’s repeated promises of a June roll-out that have pushed the stock back up to a ridiculously high multiple and trillion-dollar valuation.

Missing another deadline is not an option. But far from Mr Musk’s grand vision of self-driving Teslas running around everywhere, what will likely materialise in Austin is a minimum viable robotaxi.

For Tesla diehards, it will nonetheless be enough to bolster their faith – even as it reveals the risks to their favourite company’s entire autonomy project.

Mr Musk had spoken of launching 10 to 20 vehicles initially. Who they carry and when, where and how they do so are all important variables.

All about optics

“There are a thousand ways to game the optics of a launch,” says Mr Alex Roy, general partner at New Industry Venture Capital and an expert on autonomous vehicles.

The operational design domain can limit where driverless vehicles go, on which type of roads, at what speed and even during which hours. For example, when Cruise, the former robotaxi start-up backed by General Motors, first got going in San Francisco, paid rides were available only at night, when roads are emptier.

Waymo, the leading robotaxi company in the US, owned by Alphabet, has operated in San Francisco for years, but still does not yet serve that city’s airport. Tesla can also choose who gets these initial robotaxi rides. 

I strongly suspect there will be more Tesla-enthused influencers in the Austin area than usual in June, eager to post footage of their experience.

Tesla’s back-up will be critical. All robotaxi services require remote operators to help autonomous vehicles get unstuck from situations that confuse them or even, if necessary, take over driving.

Tesla’s support will tend towards the former, less interventionist approach, according to comments made on the last earnings call by Mr Ashok Elluswamy, head of Autopilot engineering.

Tesla will likely pack these remote ranks for the Austin launch, despite the expense. Building a profitable robotaxi business requires reducing the ratio of such support staff per vehicle, but this is not about Tesla’s earnings; it is about the multiple.

That means shoring up what Mr Roy terms “narrative command” – the sorcery by which Tesla somehow defines the conversation around autonomous vehicles, including with terms such as “robotaxi” and “autopilot”, despite the lack of autonomous vehicles.

Mr Musk must deliver something and it must be safe. It took only one bad accident, and a botched response, to derail Cruise.

Clearly, a couple dozen robotaxis in one city backed by a phalanx of remote operators would be closer to Waymo’s model than the generalised autonomy Mr Musk has pitched for years.

But investors have clearly made their peace with that already and, assuming the Austin launch creates buzz and avoids accidents, Tesla can take the opportunity to throw the story forward again.

Risks at hand

Indeed, Mr Musk said on the latest earnings call that he felt “confident in predicting large-scale autonomy around the middle of next year”. Belief is his chief currency. Yet the launch also unlocks three significant risks.

First, his politicking has turned Tesla showrooms and chargers into targets for his opponents, and robotaxis may also be in the crosshairs. 

Second, by definition, Tesla’s robotaxi effort will now transition from pure speculation to something more tangible and, therefore, measurable.

They will be like rolling advertisements for Tesla’s progress, and will also start the clock on how quickly the company expands coverage. Mr Musk predicts an S-curve, and he needs it, especially as Waymo appears to actually be on one already.

Finally, while Mr Musk has seemingly conditioned investors to accept a scaled-back robotaxi pilot in place of a national roll-out, the dissonance between the two remains significant.

Mr Elluswamy made a curious comment on the last earnings call about how robotaxis deal with emergency vehicles, noting that the driverless vehicles Tesla will deploy in Austin have an “audio input” – presumably to hear sirens – unlike the “customer-facing versions”.

Yet the foundational premise of Tesla’s autonomy ambitions is that every EV it sells with self-driving hardware installed is capable of eventually being a robotaxi.

Mr Elluswamy’s comment, suggesting the company-owned fleet has an added sensor, appears at odds with this. (Tesla no longer has a public relations department to help clarify.)

This gets at a broader question hanging over Tesla’s plans to scale up quickly from Austin: How will all those privately owned Tesla EVs really function as robotaxis?

Thorny issues of insurance, liability and even just cleaning up seats are well known, but there are other, less obvious ones. Mr Roy raises one logistical example, that of a private Tesla owner living outside an operational design domain who wants the EV to earn money as a robotaxi, but must drive it to the edge of the map first before it can do that, and then pick it up there later.

Then there is that critical remote support, which must be scaled up to presumably service the many thousands of privately owned robotaxis that are supposed to be integrated to the system soon. 

Tesla would no doubt counter that its sophisticated technology will soon overcome any remaining challenges, offering a generalised solution to autonomy that can operate virtually anywhere, and one that is safer than a human driver.

As an aspiration, it cannot be proven wrong, but the limited launch in Austin rather undercuts it. BLOOMBERG OPINION