NVIDIA’s GTC conference kicked off with a series of announcements highlighting its role in advancing autonomous driving technology.
The chipmaker provides automakers and autonomous vehicle companies with a handful of Nvidia-branded tools to help power self-driving cars and create digital twins of factories. On Tuesday, companies including Torc, Gatik, and even General Motors announced plans to use Nvidia products for vehicle production, robotics to automated driving.
To avoid any confusion among the slew of brand names Nvidia has bestowed upon its hardware and software tools, here’s a quick glossary of terms:
- Drive AGX: Nvidia’s in-vehicle supercomputer that processes real-time sensor data.
- Drive Orin SoC: A system-on-a-chip that is a more advanced central computer. Orin processes data from various sensors and powers new AVs at Level 4 autonomy.
- Drive Thor SoC: Thor is the next step up. It’s optimized for Transformer architecture, so it can accommodate generative AI, and is designed to handle everything from self-driving operations to cockpit functions to infotainment.
- DriveOS: This is Nvidia’s safety-focused operating system for Nvidia’s AV platform that promises safe, real-time AI processing and integration of advanced driving and cockpit features.
- Omniverse: Nvidia’s Omniverse is a simulation platform that allows automakers to develop and operate complex, AI-enabled virtual environments to generate synthetic data, test AV software, build digital twins of factories, and more.
- Cosmos: Cosmos is built to power world model training for physical AI development of AVs and robots.
On top of all that, Nvidia today unveiled Halos, which it defines as an AI-powered safety system for AVs and future physical AI, such as humanoid robots. Halos brings together quite a few of Nvidia’s lineup of automotive hardware and software safety solutions, so think of it as more of an umbrella.
Here’s a quick roundup of Nvidia’s automotive announcements from Day 2 of GTC.
General Motors
GM announced it has expanded its partnership with Nvidia in a collaboration that touches every aspect of the automakers business, including its factories, robots and self-driving cars. Let’s start with the factories.
GM said it will use Omniverse with Cosmos to train AI manufacturing models and help it build next-generation factories. Omniverse will allow GM to build a digital twin of its factories to virtually test new production processes without disrupting existing vehicle production, for example. It will also use Omniverse to train robotics platforms for operations like material handling and transport.
When it comes to self-driving cars, GM said it will use Nvidia’s Drive AGX for its in-vehicle hardware for future advanced driver-assistance systems and in-cabin safety experiences.
Gatik
Self-driving truck company Gatik, which is backed by Isuzu and Goodyear Ventures, has also joined Nvidia’s automotive ecosystem. The Silicon Valley and Toronto-based company, which specializes in autonomous middle-mile logistics via self-driving box trucks, says it will develop and deploy Drive AGX, accelerated by Drive Thor, to serve as the AI brain across its fleet of trucks. Gatik says it’s also running its AI models on the DriveOS system for safety.
The startup noted that the collaboration will help accelerate the deployment of Level 4 autonomous trucks at scale for the company’s customers, which include Walmart, Kroger, and Tyson Foods.
Plus
Plus, an autonomous trucking software startup, said Tuesday it will use Cosmos world foundation models to accelerate the testing and development of SuperDrive, its autonomous driver.
Plus’s SuperDrive system is built on Nvidia’s Drive AGX platform, according to the company. In a statement, Plus also said that it is pioneering “AV 2.0 technologies, which comprise generative AI, visual language models and other foundational models.” As we can see from the glossary above, Nvidia’s AGX platform is more suited to ADAS and low-level autonomy. To get that more advanced sensor fusion and on-board compute that’s necessary for higher levels of autonomy, usually companies rely on Nvidia’s Orin or Thor SoCs.
TechCrunch has reached out to Plus to ask for clarification.
The startup recently made deals with commercial vehicle manufacturers including Traton Group, IVECO, and Hyundai to integrate SuperDrive into their trucks. Plus, which has testing its technology on public roads in Texas and Sweden, has targeted a 2027 commercial launch.
Torc
Yet another self-driving truck company, Torc, announced it is working with Nvidia to develop a scalable physical AI compute system for its AVs. Virginia-based Torc, a subsidiary of Daimler Truck AG, will also work with Flex, which builds automotive-grade compute platforms.
Torc says it is using a cocktail of Nvidia chip architecture, including Drive AGX, Drive Orin, and DriveOS to support the future deployment of autonomous driving capabilities as it works towards a 2027 commercial launch.
In October 2024, the company achieved its first driver-out test on a closed course in Texas.
Volvo
While Volvo isn’t collaborating with Nvidia to accelerate its automated driving technology, the automaker is relying on Nvidia’s Blackwell GPUs to power aerodynamics simulations.
Rather than use Nvidia’s Omniverse simulator, Volvo is working with Ansys, a software simulation company. Ansys’s so-called “Fluent” simulation software, powered by eight Blackwell GPUs, has helped Volvo design its new EX90 electric vehicle in a way that reduces aerodynamic drag and, as a result, improves battery performance.
Ansys says its Fluent simulator helped Volvo reduce total simulation run time from 24 hours to 6.5 hours, allowing for multiple design iterations per day, optimized vehicle design, and accelerated time to market.
