Tesla says that next year, it’ll leverage a fraction of the millions of Tesla cars on the road to launch an autonomous taxi service powered by its Autopilot software — one that might compete with Waymo’s eponymous Waymo One and other launched and forthcoming driverless ride-hailing programs.

The network — the evolution of an idea first articulated in a document published Tesla CEO Elon Musk three years ago — won’t go live in all regions simultaneously. Regulatory approval will dictate its rollout, and only the Model 3 and Model S will be eligible initially. But any owner will be able to add or remove their car from the Tesla network and earn revenue (of which Tesla will take a 25% to 30% cut), and in areas where there aren’t a large pool of vehicles available, Tesla will supply its own.

Robo-taxi customers will be able to summon participating cars “from [their] parking lots” using a bespoke mobile app, Musk said during a presentation to investors this afternoon, and “get in and go for a drive.” He expects the network will have as many as a million cars in the next year and a half.

Tesla robo-taxi

Above: Tesla’s robot-taxi app.

Image Credit: Tesla

Tesla expects that rides in its driverless taxi fleet will cost $0.18 cents a mile compared to the $2 to $3 cost of traditional ridesharing. And on the vehicle owner side of the equation, it’s projecting gross profit of $0.65 per mile for a total of $30,000 per car per year on average, assuming the cars in question last about 11 years.

VB Event

The AI Impact Tour

Connect with the enterprise AI community at VentureBeat’s AI Impact Tour coming to a city near you!


Learn More

Musk predicts that two years from now, Tesla will produce cars without steering wheels or pedals, and he anticipates that within three years, those cars will cost $25,000 or less. That’ll be thanks in part to an upcoming battery pack designed for a million miles of operation and motors with 4.5 miles per kilowatt energy efficiency.

For the uninitiated, Autopilot is effectively an advanced driver assistance system (ADAS) that taps machine learning algorithms and an array of cameras, ultrasonic sensors, and radars to perform self-parking, lane-centering, adaptive cruise control, highway lane-changing, and other feats. It’s regularly updated via over-the-air (OTA) updates; earlier this month, Telsa rolled out a feature that enables cars with Autopilot to change lanes without driver confirmation.

This afternoon, Tesla unveiled the latest generation of the chipset that runs Autopilot — Autopilot Hardware 3.0 — that’s substantially improved over the current generation, and shipping in all new Model 3, Model S, and Model X cars. It features over 144 tera operations per second (TOPS) of neural network performance, two AI accelerators, a graphics chip, and an image signal processor, all of which translates to a 2.5 times improved power consumption and 80% reduced cost overall. Moreover, Tesla claims it’s seven times more powerful than competitor Nvidia’s Drive Xavier chip.

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.