Optimism about robotaxis returns to automotive industry

But Cruise lives on, and has restarted operations in Phoenix and the Texas cities of Houston and Dallas after a new CEO, Marc Whitten, was hired in January.

The brush with authorities has chastened GM and forced it to improve. “Our target now, instead of being better than an average driver, is to be better than a role-model driver,” CEO Mary Barra said on the company’s second-quarter earnings call.

However, Cruise has walked back its plan to roll out its dedicated Origin vehicle, which moved away from the standard converted car design to a pod with two rows of seats facing each and no steering wheel.

The Origin was canned due to “regulatory uncertainty”, Barra said. “Because the Origin doesn't have steering wheels and some other motor-vehicle safety-standard components, it doesn't meet motor-vehicle safety standards,” she added.

Instead Cruise will now use the Chevrolet Bolt EV for its autonomous operations, which also helps to cut costs.

GM”s regulatory difficulties could explain why Tesla delayed revealing its own robotaxi from 8 August to 10 October. The American EV specialist offered no solid reason for the delay after flagging the original date with great fanfare.

Another reason might have been similarity with the Verne two-seater.

CEO Elon Musk told analysts on Tesla's second-quarter earnings call that he pushed back the launch “because I wanted to make some important changes that I think would improve the vehicle”.

Tesla promises to create a robotaxi fleet from its existing fleet of cars by means of software updates have lost much of their credibility after it missed a series of self-imposed deadlines.

Its cars still deliver only level-two autonomy, meaning the driver still has to pay attention. Robotaxis are at level four.

Musk did appear to rule out offering vehicles to a ride-hailing platform such as Uber’s in the event that Tesla does manage to solve autonomy and update its Full Self Driving package to actually deliver what the name currently falsely promises.

“This would be a Tesla network. And there's an important clause we've put in every Tesla purchase, which is that the Tesla vehicles can only be used in the Tesla fleet. They cannot be used by a third-party for autonomy,” Musk told analysts.

Uber’s counter argument is that the ride-hailing business isn't as easy as outsiders might think it is.

“Our technology obscures a huge amount of complexity,” Khosrowshahi said. “Uber can provide enormous demand without AV players needing to invest capital towards acquiring customers or building the marketplace tech that delivers reliability.”

Uber originally intended to supply its own AVs but killed that dream in 2020, when it sold its AV business to Aurora, and there's no plan to revive it, Khosrowshahi said. 

This focus on the technicalities of how an autonomous ride-hailing business will operate suggests that players are more confident that the driving piece will eventually be solved.

Rimac founder Mate Rimac said in an interview with Top Gear that Verne’s focus was “not so much on that” and instead of the details like the design of the car and the “mothership” centres where cars would come back to be charged and cleaned. 

Instead, the self-driving part of Verne would be handed over to Israeli AV tech firm Mobileye, whose Drive system will also be used in future autonomous driving programmes from Volkswagen.

However, the driver part of the equation isn't solved yet. “It's a very long road,” Nvidia automotive chief Danny Shapiro told Autocar in June.

Nvidia supplies high-powered chips designed to cope with the data demands of self-driving for future and current vehicle programmes at BYD, Hyundai, JLR, Mercedes-Benz, Volvo, Xiaomi and Xpeng.

The company has described the process of solving vehicle autonomy as a triathlon. “It's almost like three different races,” Shapiro said. "The key elements are AI training, simulation and autonomous driving – or put another way, train it, test it, deploy it."

Nvidia is at the heart of the AI boom right now, thanks to its powerful chips, and the idea is similar with AVs.

“It’s the concept of [AI tool] ChatGPT, but we're not talking about text in and text out, we're talking about video in and controls for steering and braking out,” Shapiro said.

AI is useful in the training in that it can create edge cases that a test car can drive millions of miles in the real world yet never see but still need to cater for.

Nvidia is supplying the power behind the new electronic architecture for Mercedes' MMA platform, which will make its debut next year in the new CLA electric saloon.

“From that point forward, every single vehicle that we will launch will have a supercomputer on board,” CEO Ola Källenius said on Mercedes' second-quarter earnings call at the end of July.

Mercedes and BMW are the only two companies in Europe with any claim to offer hands-off, eyes-off systems in their top-end cars, but these are still limited in terms of speed and road type.

This year, Mercedes will increase the top speed of its level-three system from 37mph to 56mph and is eyeing further advances, even if it’s still a long way from full autonomy.

“Once you have put a flag on the moon, you start building a colony on the moon,” Källenius said.

Other car makers including Ford, Stellantis and the Volkswagen Group are eyeing level-three system launches starting in 2025.

As ever, the road to vehicle autonomy is slow going, with ravines to bridge and only the occasional smooth patch. Regulation, insurance and the political backlash from eliminating the livelihood of millions of taxi drivers are roadblocks that need careful dismantling by a network of AV partners working together.

“The market will have humans and AV players as part of it during this pretty long hybrid period,” said Khosrowshahi. “The fact is this is not turning out to be a winner-take-all market."