At CES 2025 in Las Vegas, NVIDIA CEO Jensen Huang announced a significant partnership with Toyota to develop next-generation autonomous vehicles (AV) using the advanced “Thor” AGX processor, marking a new chapter in autonomous driving innovation.
This collaboration, unveiled during Huang’s keynote on January 6, represents a crucial step toward making fully autonomous vehicles available to the public. The global AV market, estimated to be the first multi-billion-dollar robotics industry, is gaining momentum thanks to the partnership between NVIDIA, Toyota, and other key players.
At the heart of this announcement is the Thor processor, a universal robotic computer that, according to Huang, offers “20 times the processing power” compared to its predecessor, the Orin chip. Thor integrates data from various sensors, including high-resolution cameras, radar, and LiDAR, enabling real-time autonomous navigation.
“This is a universal robotic computer,” Huang said during the presentation. “Not just for cars – Thor is the brain for robots, whether they’re autonomous vehicles, humanoid robots, or industrial manipulators.”
Already in full production, the Thor processor marks a significant leap forward for NVIDIA and its automotive partners. Its integration into Toyota’s next-generation vehicles solidifies NVIDIA’s position in the automotive industry, which, according to Huang, already represents a $4 billion business for the company, with projections to grow to $5 billion by the end of the year.
At CES, NVIDIA also introduced Drive OS, a software-defined AI computer that has achieved ASIL D certification – the highest functional safety standard for vehicles under ISO 26262. This certification, the result of 15,000 years of engineering work, ensures that NVIDIA’s CUDA platform is now functionally safe for robotics and autonomous vehicles.
To address the challenges of training AVs in real-world conditions, NVIDIA launched the Omniverse and Cosmos platforms for generating synthetic data and simulating driving environments.
“With these tools, we turn thousands of drives into billions of miles,” Huang explained. “This way, we create vast amounts of data for training autonomous vehicles.”
The combination of Omniverse and Cosmos enables the creation of digital twins of real-world environments, generating photorealistic driving scenarios and enriching data with almost endless variations. This approach reduces the gap between simulation and reality, ensuring that AI models are trained on physically accurate data.
Huang also announced plans to deploy the Thor processor in vehicles from companies like Volvo, Rivian, and Mercedes-Benz, as well as in autonomous trucks from companies like Aurora and Waabi. Aurora plans to use Thor chips in its driverless trucks by 2027, with the first samples in production as early as 2025.
“Delivering a single driverless truck will be monumental. Delivering thousands of them will change the way we live,” said Chris Urmson, CEO of Aurora.
Despite decades of efforts and challenges such as regulations, accidents, and the complexity of AI training, NVIDIA’s advanced simulation technologies and hardware are accelerating the arrival of the AV revolution. The Thor processor, along with its accompanying platforms, represents a key step toward achieving full vehicle autonomy, known as Level 5, where vehicles can operate without human intervention in all conditions.
As Huang concluded, “The autonomous vehicle industry is here. This is an incredibly exciting time.”
With NVIDIA at the helm, the vision of fully autonomous vehicles is becoming an increasingly tangible reality, promising to reshape transportation, logistics, and robotics on a global scale.