The Fuse

This Week In AVs: Amazon In Talks To Buy Zoox; AV Companies Log Virtual Miles In Pandemic; And More

by Alex Adams | @alexjhadams | May 28, 2020

Amazon In Talks To Buy Zoox
Amazon is in advanced talks to buy self-driving startup Zoox, in a move that would expand the e-commerce giant’s reach into AV technology. First reported in the Wall Street Journal, it is anticipated that the acquisition of Zoox would be for less than the $3.2 billion it raised in a 2018 funding round—and the completion of the deal is still weeks away. If completed, the acquisition would represent Amazon’s latest—and most high-profile—foray into autonomous technology, driven in part by a desire to transport more of its goods itself. Last year, the company participated in AV startup Aurora Innovation’s $530 million funding round.

Analysts, however, are speculating over whether Amazon may have wider ambitions beyond goods delivery with its acquisition of Zoox. Backed by its high shipping volumes, background in innovation and the large amounts of capital it has to invest, Brian Nowak of Morgan Stanley believes Amazon is one of the few companies that could compete with Waymo, Uber and Lyft in robotaxis and autonomous driving. Nowak theorizes that Amazon could offer discount rides to Prime members—and could also become a competitor to the likes of Tesla and GM.

Pandemic Forces Companies To Turn To Simulated AV Miles
As the pandemic made on-road, in-person testing off-limits to AV developers, companies have instead turned to simulations and older data collected from previous tests in order to continue the development process and improve their systems. With AV fleets grounded, one such company that has turned to data is Aurora Innovation, with CEO Chris Urmson telling MIT Technology Review, “Our vehicle operators, who can’t all be on the road right now, have joined forces with our triage and labeling teams to mine our massive collection of manual and autonomous driving data for additional interesting on-road events that can be turned into virtual tests.”

Similarly, Waymo also recently announced that it is using the data from its on-road AV miles to generate camera images for simulation through AI. A Waymo spokesperson told Venture Beat the technique is a “new approach that allows us to generate realistic camera images for simulation directly using sensor data collected by a self-driving vehicle.”

Bicycles, Pedestrians Make Guangzhou A Better AV Test Site
Research recently released by Chinese self-driving developer WeRide.ai, comparing the company’s 2019 testing in Guangzhou and Silicon Valley, shows the mileage in the Chinese city to be 30 times more efficient than the miles logged in the United States. The greater return, WeRide theorizes, is down to the higher number of pedestrians, cyclists and traffic interactions in Guangzhou. Cyclists riding against traffic are 60 times more common in Guangzhou than Silicon Valley, WeRide notes, with traffic lane-changes and cut-ins five times more frequent. The findings are likely to be equally pertinent for other AV developers in the United States. As a recent Forbes article notes, self-driving companies testing in California and other parts of the country are doing so in geofenced locations with wide streets, temperate weather and better infrastructure.