Volvo makes a bigger bet on Nvidia for hands-free automated driving

Volvo is developing an automated-driving system with Nvidia for future vehicles. The system is designed to navigate both urban streets and highways. It matters because Volvo wants to bring advanced automated driving across its vehicle range, rather than limiting it to electric models.

The system combines Volvo’s vehicle safety systems and engineering with Nvidia’s Hyperion sensor and compute architecture. It also uses the Alpamayo driving software stack and the Halos safety evaluation layer. Nvidia says Alpamayo runs two complementary software approaches together: AI models and conventional software using machine learning, cameras, and radar.

Volvo has not named the first vehicle or announced the partnership’s value. Its head of software development said the first vehicle will be a brand-new model arriving “in the not-too-near future.” The first vehicle using the new Nvidia stack is expected to debut within the decade, with the technology potentially extending to later models.

Nvidia’s three technologies handle different parts of Volvo’s planned automated-driving system. Hyperion is the sensor and compute architecture. Alpamayo is the driving software stack. Halos is the safety evaluation layer. Volvo will combine all three with its own vehicle safety systems and engineering.

Alpamayo includes AI models, datasets, and simulation and development tools for building autonomous-driving software. Nvidia also pairs it with classical software that uses machine learning and camera and radar perception. Hyperion provides the architecture for the sensors and computing that support this software. Halos is intended to evaluate the system’s safety.

The technologies are part of an expanded partnership, not a completely new relationship. Volvo already uses Nvidia’s Orin computer chip in all its battery-electric vehicles. The automaker says it plans to standardize more on Hyperion, including the sensor set, to open the technology stack to a wider partnership with Nvidia.

The partnership’s potential scale is broader than Volvo’s battery-electric vehicles. Volvo says the new Nvidia technology stack is intended for its electric vehicles, plug-in hybrids, and gasoline-powered vehicles. That would make the platform relevant across the automaker’s major powertrain categories.

The expansion builds on an existing relationship that is already important for Volvo’s EVs. All of Volvo’s battery-electric vehicles use Nvidia’s Orin computer chip. The chip processes information from cameras, radar, and other sensors, helping the vehicle understand its surroundings and make driving decisions.

Volvo has not disclosed the partnership’s value or named the first vehicle. It also has not provided a precise count of vehicles that could eventually use the new system. The stated direction is to standardize more on Nvidia’s Hyperion architecture and sensor set, allowing the technology to extend beyond EVs to plug-in hybrids and gas-powered vehicles.

When the new Nvidia technology stack arrives, Volvo vehicles will use a more standardized combination of sensors, computing, driving software, and safety evaluation tools. The system is intended to support automated driving on urban streets and highways. It will combine Volvo’s safety systems with Nvidia’s Hyperion, Alpamayo, and Halos technologies.

The first application will be a brand-new vehicle. Volvo’s head of software development said it will arrive “in the not-too-near future,” and later clarified that the first vehicle with the new Nvidia stack will debut within the decade. Volvo has not identified the model.

The change also reflects Volvo’s broader move toward becoming a full-tech-stack company. The automaker wants the technology to serve electric vehicles as well as plug-in hybrids and gasoline-powered vehicles. After the first launch, Volvo indicated that the technology could continue into later vehicles, although it gave no detailed rollout schedule.

Nvidia combines AI models with conventional software to create two software approaches running at the same time. Ali Kani described the Alpamayo stack as redundant and diverse. The goal, as presented in the article, is to avoid relying on only one kind of software or one way of interpreting the vehicle’s surroundings.

Alpamayo includes a family of AI models, datasets, and simulation and development tools. Alongside it, classical software uses machine learning plus camera and radar perception. The vehicle therefore processes information through both the AI-based and conventional approaches. Nvidia calls its use of multiple sensor types a multimodal-sensor philosophy.

Kani said this combination is safer because the approaches and sensor inputs complement one another. The open-source Alpamayo software has been downloaded by more than 650,000 developers. Volvo plans to use the stack with its own safety engineering and Nvidia’s Hyperion sensor and compute architecture.

A standardized technology platform gives an automaker a common foundation for sensors, computing, software, and safety systems. Instead of developing each car line separately with different suppliers and technical approaches, Volvo can organize more vehicles around the same stack. The article connects this change to Volvo’s move toward being a full-tech-stack company.

In practice, Volvo plans to standardize more on Nvidia’s Hyperion architecture, including its sensor set. The same broader partnership is intended to support electric vehicles, plug-in hybrids, and gasoline-powered vehicles. Nvidia’s Orin chip already processes cameras, radar, and other sensor information in all of Volvo’s battery-electric vehicles.

This approach could help Volvo extend software and sensor work across more vehicles. It also gives the automaker a path to introduce the new automated-driving system beyond EVs. The article does not provide specific cost, timing, or production benefits, but Volvo says the technology can continue beyond the first vehicle launched within the decade.

A vehicle first needs information about its surroundings. Cameras, radar, and other sensors collect that information. A computer then processes the sensor data, while driving software interprets it to help the vehicle understand its environment and make driving decisions. Volvo’s battery-electric vehicles already use Nvidia’s Orin chip for this processing.

Machine learning supports perception and decision-making software. Nvidia’s Alpamayo stack combines AI models, datasets, and simulation tools with classical software that uses machine learning, cameras, and radar. These two software stacks run at the same time. Hyperion supplies the sensor and compute architecture that connects the system’s inputs and processing.

The approach also includes safety redundancy and diverse inputs. Nvidia says using multiple software approaches and multimodal sensors makes the product safer. Halos adds a safety evaluation layer, while Volvo contributes its vehicle safety systems and engineering. Together, these parts form the planned automated-driving stack for city streets and highways.