Whatever one feels about Tesla’s controversial CEO, Elon Musk, one thing is clear: the EV company and its leader know how to sell a robotaxi vision.
As a testament: Tesla doesn’t have a single robotaxi on the road providing paid rides, years after Musk has promised multiple times that it’s just around the corner. The closest glimpse of an unmanned Tesla on public roads only came on June 10 with a short clip on X that showed a black Model Y appearing to drive itself on Austin’s road with another Tesla trailing behind.
Meanwhile, Alphabet’s Waymo is demonstrating its autonomous taxi in multiple cities and that humans are willing to pay for it. In May, the company announced that it crossed the milestone of 10 million paid rides.
Yet, Tesla is never left out of the robotaxi conversation. Some analysts have been pricing the mere idea of a robotaxi service into Tesla’s stock.
Seth Goldstein, a Morningstar analyst, told Business Insider he could even see Musk’s company pulling ahead.
“Tesla’s vehicles will be cheaper than Waymo. The camera-only approach means that, in theory, a Tesla Full-Self Driving system can operate on any road even if a Tesla has never driven there before,” he said. “Versus, if you look at the way Waymo’s system works, they have to spend at least a year in a new location, geo-fencing the entire area.”
The Tesla thesis is compelling.
What Tesla and its CEO promise is a robotaxi that has a lighter sensor stack — that is, only cameras — that will help the company’s ability to scale a robotaxi service at lower costs. In addition, Tesla says it has the manufacturing capability to rapidly produce a two-door, purpose-built robotaxi called the Cybercab. And yet another proposition is that Tesla owners also will be able to turn their cars into autonomous chauffeurs.
Combined with an AI-powered driver system that has been trained on “billions of miles of anonymous real-world driving data,” according to Tesla, Musk’s company is proposing what he says will be a cheaper and more intelligent robotaxi service that can scale faster than its competitors, like Waymo, at a fraction of the cost.
“I predict there will be millions of Teslas operating fully autonomously in the second half of next year,” Musk said during Tesla’s first-quarter earnings call in April.
Tesla’s success, however, rests on a lot of assumptions.
Alex Roy, general partner at New Industry VC and former director at Argo AI, a self-driving startup, is bullish on Tesla’s robotaxi vision — but he lists several hurdles the company has to jump over.
For one, Tesla has yet to show evidence of a significant operations arm — the boring but crucial side of the ride-sharing business that includes maintenance of the vehicles.
“If you don’t have a significant operations arm to maintain, store, and repair the vehicles and that is very close or inside the deployment geofence, you can’t be in the robotaxi business,” Roy said.
In this area, he said Waymo’s being first is crucial and makes the company a leader in the robotaxi game.
The other half of Tesla’s thesis — that is, Tesla’s proposal to turn any one of the millions of privately-owned cars into a robotaxi — is still at a wait-and-see moment, Roy said.
The VC believes that Tesla’s AI driver will only be as good as what Tesla’s suite of cameras allows the AI driver to see, regardless of the vast amount of real-world data the EV company touts. Roy is thinking about conditions where visibility will be an issue, such as foggy weather.
Waymo previously demonstrated at Google I/O how its lidar and imaging radar system are able to detect humans in the middle of a sandstorm or a pedestrian whose view is blocked by a bus, therefore undetectable by cameras.
Roy said that he “absolutely” believes Tesla can deploy with cameras only, but with limitations.
“It cannot drive faster or better than the limitations of what it can see,” he said. “Whereas, a Waymo has sensors with additional capabilities that a camera can’t see.”
Putting aside the limitations of cameras, Roy said Tesla will have to figure out two other components to deliver a robotaxi service powered by personally-owned vehicles: a good user experience and, again, an operations solution for managing the fleet of personally-owned Teslas.
The VC said Tesla can either support private owners with cleaning and maintenance through its own operations, or the owners themselves can do it. Turo, the car rental app, is an example of what the latter solution would look like, Roy said.
“It’s like Airbnb — a lot of hosts are great. They’re not all great,” he said. “Furthermore, we also know from Turo that if vehicles have to come from further away, that could be inconvenient.”
Related stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
The robotaxi race is an expensive endeavor
Cruise, once a formidable player in the robotaxi space, burned through $10 billion before it was shut down by its parent company, General Motors, last year.
Waymo, a subsidiary of Alphabet, announced last year that it raised $5.6 billion from outside investors. In addition, Alphabet CEO Sundar Pichai said last July that the company would commit $5 billion in a multi-year investment toward its robotaxi ambitions.
The consensus among analysts is that robotaxis will be a few-winners-take-most outcome, with data and capital being key to survival in a brutal business.
Companies with access to “data and capital (to fund compute) remain well positioned to drive the next leg of AV deployments,” Brian Nowak, a Morgan Stanley analyst, wrote in a note on Wednesday.
Roy and some AV players argue that the type of data self-driving companies have access to may be more important than the quantity of data they have.
Amnon Shashua, CEO of Mobileye, said at a Los Angeles conference in April that the amount of real-world data available has been “exhausted,” which is why simulated data has been helpful for exploring rare driving cases.
“People have said, ‘Oh data is the new oil.’ Well, actually no, data is the new unrefined materials that goes into manufacturing oil,” Roy said. “There are many types of data, and data varies in quality. There’s a lot more to it. It’s just not that simple.”
For Roy, the conversation around data or the “lidar versus camera narrative” is largely an “obfuscation of the real issues.”
“The only thing that matters is will people pay for this,” he said, adding that customers don’t want a robotaxi that pauses or hesitates or has “suboptimal remote operations and suboptimal customer service.”
Goldstein, the Morningstar analyst who is bullish on Tesla’s success, told BI that he believes safety is the one factor holding back Musk’s company from delivering a robotaxi service that will be available to the broader public. Tesla’s robotaxi debut in June will be invite-only and limited to 10-20 robotaxis.
Musk himself said in an X post that Tesla is being “super paranoid about safety” and that the June 22 launch date can shift.
Morningstar published a study this year that compared average miles driven between disengagements — miles driven before a human or remote operator had to intervene — for Waymo and Tesla’s Full Self-Driving v13 software. The study found that Tesla was still behind Waymo. Tesla was around 10,000 miles between disengagements, and Waymo was around 17,500 miles between disengagements.
Goldstein said that he believes Tesla will be able to operate robotaxis once it goes beyond the threshold of 10,000 miles between disengagements.
“For consumers, once a Tesla is safer than a human driver, I think that will be enough to drive consumer adoption,” he said.
A Tesla spokesperson did not respond to a request for comment.
