• Autonomous Vehicles (AV) was the main focus at CES 2025.
• NVIDIA’s Cosmos will speed up the AV development cycle through its “World Foundation Model.”
• Mobileye discussed pathways to full autonomy and how it is pursuing its technology development.
• Waymo will continue to expand its operations in more US cities and AV testing globally in 2025.
• John Deere is bringing driverless technology to niche verticals like farming and construction.

Over the past few years, CES has evolved into a technology-focused event as the world sees a rapid diffusion of emerging technology across various industries in the post-COVID-19 era. Automotive is an industry that has grown its presence and made headlines at CES in past years.

2024 was a successful year for self-driving technology, in particular the robotaxi industry, paving the way for autonomous vehicles (AV) to become one of the hottest topics at CES 2025. Players like Zoox, May Mobility, and Waymo were present at the event. Waymo made headlines from the keynote and also showed off its largest-ever booth. Here’s a quick overview of the main announcements on driverless technology made at CES 2025.

NVIDIA

At CES 2025, NVIDIA made several announcements. However, the one that stole the spotlight was the unveiling of NVIDIA Cosmos, a World Foundation Model (WFM) platform to accelerate physical AI development. The Cosmos can be seen as an extension of NVIDIA’s Omniverse platform, a tool for autonomous driving simulation. This new platform allows developers to generate videos of real-world environments to train AVs for specific scenarios. Furthermore, 3D scenarios developed in the NVIDIA Omniverse can be used to generate photoreal videos using Cosmos models. This offering is an excellent solution for companies with very little or no autonomous driving data. Cosmos WFMs will be available under open model license to accelerate work on AV and robotics. The platform already has several early adopters, including Uber, Wayve and Xpeng.

Source: NVIDIA

NVIDIA also announced its partnership with Toyota, one of the world’s largest global automakers. Toyota will use NVIDIA Drive AGX Orin running NVIDIA Drive OS. Autonomous Trucking company Aurora, in partnership with Continental, will use NVIDIA Drive Thor and Drive OS for its Aurora Driver, a level 4 autonomous driving system. The semiconductor company has proposed its three-computer solution, NVIDIA DGX, to train its AI models in the data center, NVIDIA Omniverse for simulation, and NVIDIA AGX in-vehicle computer to process real-time sensor data at the edge. These three computer solutions create a development loop from car to cloud to simulation. Adding Cosmos to this mix will make a flywheel that can generate billions of virtual driving data from thousands of real-world-driven data.

Source: NVIDIA

Analyst Take

• NVIDIA’s open platform and its chip to cloud approach for AV development offers hardware and software tools for automakers to either opt for the full stack solution or adopt any of the NVIDIA AV tools. This comprehensive 360-degree approach has allowed different players to choose NVIDIA depending on their state of AV technology.
• Partnering with Toyota will open the doors for NVIDIA to enter the ADAS/AD mass-market which it has been struggling with over the years. With strong competition from Chinese players and Tesla, many legacy OEMs are looking for partners such as NVIDIA that can help in faster development of AD technology.
• NVIDIA’s Cosmos would serve as the catalyst in the AV development cycle as it can generate a plethora of real-environment scenarios, allowing various players in the AV value chain to use it differently as per their requirement and AV development stage.
  

Mobileye

At its press conference, Mobileye, Intel’s autonomous driving subsidiary with a significant share of the ADAS market, did not focus on the commercial progress of its own products, but rather explained and shared its development pathway from Advanced Driver Assist System to fully autonomous solutions.

Mobileye referred to two critical dimensions needed to achieve scalable deployment of autonomous vehicles – precision (minimizing errors) vs recall (handling diverse scenarios, geographies, environment). Referring to its safety architecture, Mobileye explained how it addresses unreasonable risks and mitigates errors with a Primary Guardian Fallback (PGF) decision-making model that generalizes the majority rule to non-binary decisions.

Source: Mobileye

Its large data set with ~30 billion miles of data harnessed in 2024 alone provides data essential for memory creation in addition to detailed map data for hands off driving. Mobileye also discussed how it was using sensor technology to build layered visuals with 3D perceptions providing redundancy and reliability. Mobileye explained how it aimed to climb the precision axis with its imaging radar technology, which delivers high resolution and addresses the cameras’ weak spots. All of this is powered by Mobileye’s SoC EyeQ6.

Analyst Take:

• It was interesting to understand the underlying technology involved in the journey towards full autonomy at Mobileye’s press conference. However, it would have been better if Mobileye would have provided a better understanding of the new products – Mobileye Chauffer™ and Mobileye Drive™ – along with the sensors and commercial milestones unlike previous years where the company featured design wins at CES.
• Mobileye did not publish the TOPS performance of its EyeQ6 SoC. It highlighted how TOPS was not the ideal measure and instead a full system performance should be judged. Although this is true, it side-steps the ultimate aim of software defined vehicles, where centralization of compute power is as important as the function performance of ADAS & AD.

Waymo

US-based robotaxi company Waymo had a pivotal 2024 in terms of momentum as the company surpassed 4 million driverless pay rides and announced the expansion of its services in three new US cities in 2025. Waymo will start testing its robotaxi vehicles in new cities within the US and globally in 2025. Waymo also showcased its sixth generation Waymo Driver sensor suite on the new robotaxi electric vehicles from the Hyundai Ioniq 5 and Zeekr’s purpose-built van, Zeekr RT. Waymo has optimized the cost for the sixth-generation self-driving system by reducing its camera from 29 to 13 and LiDAR from five to four. In the future, Waymo may extend its business to local delivery, trucking, and licensing its tech to automotive OEMs.

Source: Counterpoint Research

Analyst Take:

• Waymo has made significant progress in 2024 and will continue to expand its operation and add new testing sites internationally in 2025. However, starting international operations will be a challenge as the company lacks autonomous driving data.
• Waymo’s parent company, Google, is at the forefront of AI advancements. Building upon and leveraging Google’s Gemini, Waymo is working on an End-to-End Multimodal Model for Autonomous Driving (EMMA) that works only on camera systems similar to Tesla and Wayve. With this new approach, Waymo can address scalability concerns.

John Deere

John Deere announced its first autonomous tractor at CES 2022, three years after the company announced four new machines – a large tractor, an articulated dump truck, a tractor for orchards, and an electric battery mower. With its new autonomous offerings, the company is set to revolutionize farming, landscaping, and construction and aims to address the labor shortage across the three industries. Its second-generation technology stack consists of 16 stereo cameras for the 360-degree view of the field and two NVIDIA Drive Orin to process real-time data at the edge.

Source: John Deere

Analyst Take:

• John Deere revolutionized the agricultural, construction and commercial landscaping industry by leveraging autonomous vehicle technology to address the growing labour shortages among these sectors.
• Due to its less complex environment in comparison with driving in the real world, the technology has a faster development time with less edge cases to address.