The "Agentic" Personal Car vs. The Fleet Robotaxi

Three significant events in the first quarter of 2026 relate to autonomous driving technology. In January, the startup Tensor showcased a self-driving car at CES, which can fold its steering wheel while operating. That same month, Mercedes-Benz halted its Drive Pilot system, the only fully autonomous driving option available to American consumers.

In March, Sony and Honda announced they would cancel the Afeela 1 project before any customer purchases. This sequence of events suggests a shift in priorities, moving from ambitious personal autonomous cars to more modest implementations. The most notable example of this appears to be Tensor's Robocar. Described as a Level 4 vehicle for private ownership, the Robocar offers dual functionality: it can be driven or operate autonomously.

The hardware is impressive, featuring eight NVIDIA DRIVE AGX Thor chips, over 8,000 TOPS of computational power, more than a hundred sensors, and a foldable steering wheel. However, the timeline for delivery remains unclear, and the system's capabilities are not yet publicly known. Tensor is not the only company pursuing this vision.

Lucid Motors aims to deliver a consumer-owned Level 4 car using two Thor chips in the next few years, a step beyond Tensor's single-chip approach. The claim of 8,000 TOPS of computational power is significant but does not necessarily equate to superior safety or intelligence. For instance, the Volvo ES90 has only half that capacity.

The real value of this computational power lies in redundancy and future upgrades via over-the-air software updates, which could prove crucial for a vehicle owned for a decade. The concept of the car earning money while idle is appealing. The idea is that the car would drop off passengers and then spend the day on a ride-hailing platform, earning income.

This model transforms the car owner into a one-vehicle fleet operator, with responsibilities like platform commissions, energy costs, cleaning, and insurance. It is not yet clear whether autonomous cars will be profitable under this model, as no personal autonomous cars are currently earning in this way. The shift from Level 3 highway cruising systems, which require the driver to pay close attention, to Level 2++ systems that enable hands-free, eyes-on driving marks a significant change in commuting. However, the transition from Level 3 to Level 2 is not a leap towards true autonomy.