Daimler, NVidia, and Bosch are teaming up to build robo-taxis for the 2020s

Mercedes maker Daimler, one of the world’s oldest automakers, has just forged a deal with Silicon Valley chipmaker NVidia. Today, the two companies—along with automotive components giant Bosch—provided more details on plans to build fleets of fully AI-driven taxis that will hit the road “early in the next decade.”

Of course, NVidia has its own long history–in exactly the tech that Daimler and Bosch need–graphics processing units and other advanced artificial intelligence chips required to train and operate fully autonomous cars. “There’s a lot of work that’s been going on behind the scenes, of course,” says Danny Shapiro, NVidia’s senior director of automotive. “There is also some testing that is going to start taking place next June.”

Tests will begin in California, but will “quickly evolve beyond that,” says Shapiro. Daimler and Bosch, who are leading the effort, will also introduce a pilot project of autonomous shuttle cars in a “city located in the San Francisco Bay [area] in Silicon Valley” in the second half of 2019, per a press release.

Daimler and Bosch haven’t yet said who will operate these fleets of taxis–whether they will establish their own services or whether they will sell or lease vehicles to established ride-sharing players like Didi, Lyft, and Uber. (Uber is still developing its own autonomous cars, after a fatal crash in March; but the creator of that program, Jeff Holden, recently left the company.)

The big advance, according to NVidia, is the development of a suite of hardware and software. It’s centered on the new Xavier AI accelerator, a 9-billion transistor chip that performs a mind-bending 30 trillion operations per second. The company has also developed a new machine learning system, called DGX-2, to train autonomous car algorithms, and a simulation system, Constellation, that allows testing of Xavier-based systems on billions of virtual miles—where no one can get hurt.