Supercomputer on wheels

The autonomous car that is no longer just a vehicle.

Tensor presented at CES 2026 with its Robocar, a car designed first as AI and only then as a car, unlike electric vehicles that receive autonomy as an extra, the Tensor Robocar was designed from the ground up for level four autonomous driving.

At the heart of the vehicle is a true mobile data center built with eight Nvidia Drive AGX Thor systems based on the Blackwell architecture, the result is brutal, with more than 8,000 tops of computing power capable of processing gigantic volumes of data from sensors, cameras, radars and lidars in real time.

Everything in Robocar revolves around artificial intelligence.

All mechanisms such as steering, braking, perception, decision making and even recharging were integrated into a single unified autonomy platform. The perception of the environment is carried out through a double lidar system, Robocar offers 360º vision with up to 25 million aces per second, while the Sentinel takes care of blind spots and the near field.

Instead of relying on a single sensor, Tensor relies on redundancy and total overlap, safety follows the same logic, Robocar eliminates single points of failure with fully redundant electronic steering and braking systems, including four-wheel steering and multiple independent braking systems.

And the car does not need anyone to drive, much less to recharge, it supports ultra-fast and fully autonomous charging, going from 20% to 80% battery and about 10 minutes with powers close to 400 kW. Externally, LED panels communicate with pedestrians and cyclists, while inside the vehicle has the first mass-produced folding steering wheel, symbolizing the definitive transition between human driving and total autonomy.

For companies this is not just a new car, it is a change in the economics of transportation, fewer human drivers, more efficiency, more predictability and a fleet that thinks for itself. The Robocar makes it clear that the future of autonomous vehicles will not be decided by sensors or batteries alone, but by AI-controlled computing power.