The Private Route to Autonomous Driving: Tensor’s 2027 Robocar Era
The vision of a fully self-driving car, once relegated to science fiction, is rapidly materializing, particularly in the realm of autonomous ride-hailing. Today, consumers can summon Waymo services in numerous American cities, hail Tesla’s Robotaxi in Austin, Texas, and utilize various autonomous taxi fleets across China. Now, a prominent former Chinese robotaxi operator is poised to introduce an even more ambitious concept: a private, ground-up Level 4 autonomous vehicle for individual ownership. Enter the 2027 Tensor Robocar.
A Decade in the Making
Founded in Silicon Valley in 2016 as AutoX, the company initially dedicated its efforts to developing autonomous commercial vehicles and robotaxis. The following year marked the commencement of autonomous vehicle testing in both California and China. During the global COVID-19 pandemic, the company made the strategic decision to relocate its operations to China full-time, subsequently building a substantial fleet of over 1,000 autonomous taxis providing public rides in five cities.
However, in a significant strategic pivot over the past year, Tensor has completely divested from its Chinese operations. According to Amy Luca, the company’s head of marketing, this decision was primarily driven by mounting concerns over data privacy. Following this transition, the company rebranded as Tensor, returned its headquarters to San Jose, California, and redirected its focus toward engineering a truly autonomous vehicle designed for private consumers rather than large corporate fleets.
At its core, the Robocar presents as a sophisticated electric vehicle (EV), equipped with a substantial 112-kWh battery pack offering an estimated range of 250 miles. The current configuration features a single rear motor with an unspecified power output. While the vehicle’s curb weight remains undisclosed, preventing a precise performance calculation, Tensor has highlighted the battery’s advanced 845-volt architecture, enabling rapid charging from 10 to 80 percent in a mere 20 minutes. Furthermore, the company is actively developing an automated charging system, employing a robotic arm designed to autonomously connect the vehicle for charging.
Pending the full realization of this automated charging solution, drivers will still need to interact with the vehicle’s doors. However, Tensor has addressed this with elegantly designed coach-style doors that open from the center. These doors are fully powered and equipped with integrated sensors to prevent accidental contact with other vehicles or surrounding obstacles.
A Genuine Autonomous Vehicle
Tensor officially classifies the Robocar as an SAE Level 4 autonomous vehicle. This designation signifies that the vehicle is capable of operating entirely without human intervention, although it retains a conventional steering wheel and pedals to facilitate manual operation when desired. In many respects, the Robocar shares significant similarities with Waymo’s autonomous taxi service; however, the crucial distinction lies in the ability for private individuals to own and park the vehicle in their own driveways, or alternatively, to have it autonomously park itself.
This advanced level of capability represents a significant leap beyond the technology currently offered by Tesla with its Full Self-Driving (Supervised) system. While Tesla’s system remains the most advanced option currently available to private purchasers in the United States, it still mandates constant human supervision, requiring the driver to remain alert and prepared to assume control at a moment’s notice.
To achieve this high degree of autonomy, Tensor opted to design the entire vehicle from the ground up specifically as an autonomous platform, rather than attempting to retrofit an existing production model. The development phase for this specialized vehicle commenced in 2020, shortly after the company launched its autonomous taxi service in China.
The path to true autonomy necessitates the integration of an extensive array of sensors and a formidable onboard computing infrastructure. The Tensor Robocar is outfitted with over 100 individual sensors, including five separate lidar arrays—one positioned centrally on the roof and four others strategically located around the front, sides, and rear of the vehicle. Complementing these are 37 cameras, 11 radar units, and 10 ultrasonic sensors. The rooftop lidar system possesses the remarkable capability to detect objects up to nearly 1,000 feet away, providing comprehensive 360-degree environmental awareness.
Ensuring the optimal performance of these numerous sensors—keeping them clean and unimpeded by adverse conditions—requires a sophisticated support system. Tensor has equipped the vehicle with 30 washer nozzles and 13 miniature wipers. Additionally, integrated heating elements are incorporated to prevent sensor fogging and the accumulation of snow or ice. Tensor further enhances this protective measure by incorporating physical covers that automatically deploy over the sensors when the vehicle is powered down, effectively safeguarding them from potential damage and external contaminants—a feature that surpasses the solutions employed in Waymo’s Zeekr-based vehicles.
The immense computational demands of processing this vast sensor data are met by a powerful onboard computer featuring eight Nvidia Drive Thor-X chips. This system boasts an aggregate processing capacity of 8,000 TOPS (trillion operations per second). While the vehicle maintains a persistent connection to the cloud, the majority of the intensive processing tasks are executed locally within the vehicle itself. This localized approach ensures consistent operational capability even in environments where a stable 5G signal may be unavailable. To mitigate connectivity interruptions, the Tensor is equipped with three redundant communication channels, maximizing its ability to maintain contact with external networks.
The Robocar’s software infrastructure is built upon the Tensor Foundation Model, an advanced AI system that operates two distinct processing pathways in parallel. The first pathway was trained using data from professional human drivers, while the second was developed using a Visual Language Model (VLM). This dual-pathway approach enables the system to effectively resolve unusual and unforeseen edge cases that might challenge a single-source system. Tensor asserts that the vehicle is capable of operating reliably in adverse weather conditions, including rain and snow, thereby expanding its usability beyond regions with consistently favorable climates.
Visual indicators are also employed to communicate with the external environment. Displays integrated into the lower exterior corners of the vehicle are designed to broadcast simple messages and pictograms to pedestrians, clearly indicating that the vehicle is operating autonomously and is actively detecting their presence.
A New Paradigm of Data Ownership: Your Car, Your Data
A significant benefit of the Tensor Robocar’s architecture is its onboard data processing capability. Because the vast majority of computational tasks are performed locally, Tensor does not inherently require access to the data generated by the vehicle. While the Robocar is capable of sharing information with the cloud, this data transfer is strictly opt-in; owners must actively consent to such sharing. Consequently, all data collected by the vehicle remains under the owner’s control.
Owners have the ability to access all data collected on their vehicle and its users through the vehicle’s interface or via the companion mobile application. Furthermore, owners possess the complete authority to delete any or all of this data. This comprehensive data management capability extends to all biometric information, including facial and palm recognition data, which are utilized for essential functions such as vehicle access and theft prevention.
Although the vehicle is equipped with interior cameras and microphones to facilitate driver monitoring during manual operation and to enable voice interactions with the onboard AI assistant, each of these components features a physical cover and dedicated off switches. Owners can easily disable these features for enhanced privacy when desired.
Conversational Intelligence: Talk to Your Car
The full potential of the Robocar’s communication system is realized when drivers choose to utilize the onboard microphones. The vehicle is equipped with an Agentic AI system, underpinned by a Large Language Model (LLM), designed to interact with occupants in a manner that closely simulates human conversation. Rather than issuing terse commands, users can engage in a natural dialogue with the car to specify their desired destinations.
Similarly, the vehicle can be summoned from any location through a simple voice call or text message requesting it to come and pick up the user. The system is also designed to learn the owner’s routines. When integrated with a personal calendar, the car can anticipate upcoming trips and proactively calculate the necessary battery range and charging requirements to support those journeys.
Manual Control Option: Drive Yourself, If You Want To
While the Tensor Robocar is engineered primarily for autonomous operation, it retains a conventional steering wheel and pedals, allowing the owner to drive the vehicle manually whenever they choose. When under human control, the vehicle can provide a spectrum of driver-assistance features, ranging from fundamental aids like automatic emergency braking to advanced Level 3 semi-automated driving capabilities that allow for hands-free, eyes-off-the-road operation.
When the driver opts not to drive, the system can be switched to autonomous mode. In this mode, the steering wheel retracts smoothly into the dashboard, and the central infotainment screen slides horizontally to conceal the controls. (The identical passenger-side screen remains stationary.) The accelerator and brake pedals also retract out of the way. Should the driver decide to resume manual control, these components will seamlessly emerge from their stowed positions.
The vehicle’s steering, braking, and acceleration functions are all managed through electronic controls (by-wire systems), eliminating the need for physical mechanical linkages. Critical systems, including the entire sensor suite and the drive-by-wire components, are equipped with multiple redundancy layers. This ensures that the vehicle can continue to operate safely even if a failure occurs in any of its sensor or drive systems. The integration of rear-wheel steering allows the back tires to pivot up to 7 degrees in either direction. This capability grants the vehicle a claimed turning circle of just 37 feet, which is comparable to that of a Tesla Model Y despite the Robocar being substantially larger.
In the unfortunate event of a collision, Tensor is targeting top safety ratings from the National Highway Traffic Safety Administration (NHTSA), the Insurance Institute for Highway Safety (IIHS), and the European New Car Assessment Programme (Euro NCAP).
Maximizing Utility: Make Your Car Work for You
