Mobileye And Innoviz Move Up On The Autonomy Scale

The Consumer Electronic Show (CES), held in Las Vegas from January 7-11, 2025, devoted significant floor space and exhibits in the area of autonomy - from the likes of Waymo and Mobileye for passenger cars and ride-hailing, autonomous racecar competitions organized by Indy Autonomous Challenge, and big machine companies like Caterpillar, John Deere, Komatsu and Oshkosh presenting autonomous solutions in mining, construction, agriculture, airport operations, lunar construction (yes!) and garbage hauling.

Mobileye, previously owned by Intel and spun out as a separate NASDAQ listed company in 2022 has a current market capitalization of ~$14B and revenues of $1.8B/year (easily the leading player in the pure play autonomy landscape). Its customers include 50 automotive OEMs who use their ADAS based solutions (cameras, chips, software) across 1200 car models and 190M cars. Mobileye’s strategy has always been to move up in the autonomy game to support L4 (complete autonomy in prescribed geographies, traffic and weather conditions) and L5 (complete autonomy, anywhere, anytime). At CES 2025, it announced collaborations for L3 (highway autonomy at prescribed speeds with human driver able to take control within 10 seconds) and L4 (no human driver required in specific Operational Design Domain or ODD) autonomy solutions. Volkswagen is one of the customers across multiple models (VW ID Buzz, Audi, etc.) that advertised the L4 capable ID Buzz vehicle.

CEO, Amnon Shashua laid out Mobileye’s vision and roadmap for revolutionizing autonomy from its current L2 offerings (hands on, eyes on, minds on) to L5 autonomy (no human driver required, autonomy everywhere). At a press briefing at CES 2025, he described the trajectory of Mobileye’s roadmap from L2 to L5 capabilities (Figure 1).

Dr. Shashua describes 2 key concepts in the autonomy revolution:

1) Precision - which is in essence a metric that captures the quality of the autonomy stack in terms of disengagements and safety at different autonomy levels).

2) Recall - which is a set of constraints in which the autonomy stack can work, for example, traffic and weather conditions, geography and cost effectiveness for deployment in affordable cars.

L5 is the nirvana state with 100% Precision and Recall (its great to dream, but is it really achievable???)

Waymo’s L4 robotaxi service is high on the Precision axis (Y axis, green dot)) but very expensive and constrained in terms of where it can operate. Tesla’s L2 capability on the other hand is very high in Recall (X Axis, red dot) but low on precision which means it is not really autonomous and needs constant human control. Mobileye to date has been following the Tesla path with its SV52 (SuperVision™ based on its proprietary EyeQ5 chip) launched in 2024. In 2026 the SV62 (based on the EyeQ6 chip) with higher precision is to be deployed in the Porsche and Audi product lines. The SuperVision™ capabilities are L2 systems which need constant human driver attention and control.

After this, the Mobileye roadmap zig-zags. With the Chauffeur™ (expected commercial launch in 2027), Recall will be sacrificed (only highways, up to 130 kph speeds, higher cost) but precision greatly improved to deliver affordable L3 capability. This is enabled by fusing camera, radar and a front-facing LiDAR (hence the higher cost). The final step in what Dr. Shashua describes as a revolution is the Mobileye Drive™ which is driverless, and used by robotaxi and trucking customers and integrated with a total of 13 cameras, 5 imaging radars and 3 LiDARs. Beyond 2027 as the Chauffeur™ increases Recall (more road conditions and geographies, lower cost), the addition of sensors and compute capabilities are critical for delivering L4 capabilities with higher degrees of Recall.

The keys in enabling Mobileye’s autonomy revolution are:

1) Diverse driving data collected to date across global geographies, traffic patterns and weather with its SuperVision™ camera based systems (300 PB to date covering millions of miles. This is critical for training its AI based perception and driving stacks.

2) Semiconductor capabilities, such as the EYEQ generation of chips

3) Redundancies in sensors, compute stacks and algorithms to ensure successful edge case negotiations through intelligent fusion

4) Sensor Capabilities, most notably its internally developed imaging radar chips, which draws on IP during its tenure within Intel. Mobileye believes that its imaging radar is a perfect complement to cameras, especially in bad weather and will play a big role in increasing Recall (by increasing availability when cameras and LiDAR are compromised). The resolution is twice that of other imaging radars on the market enabling small object hazard detection, segmentation and identification (at speeds of 130 km/hr, OEM requirements are 130 m, Mobileye claims it’s radar meets 136 m). Costs are expected to be in the $100 range. Adding imaging radar to achieve 360° Field of View (FoV) is a critical enabler to achieving higher Recall and delivering L4 capability through Mobileye Drive™.

The Role of LiDAR and Innoviz

Mobileye’s evolution (or revolution) to L3 and L4 capabilities through Chauffeur™ and Mobileye Drive™ include use of front and side facing LiDARs, At CES 2023, Mobileye promoted its internal development of a 1320 nm FMCW (Frequency Modulated Continuous Wave) LiDAR, leveraging Intel’s photonics capabilities. Roughly 18 months later, it wound down its internal LiDAR efforts, claiming that its internal development was less essential to its future efforts, and “that the availability of next-generation FMCW lidar is less essential to our roadmap for eyes-off systems”. It cited as reasons the progress on its EyeQ6-based computer vision perception, performance of its imaging radar, and better-than-expected cost reductions in third-party time-of-flight (ToF) LiDAR units.

Innoviz, an Israel based, ToF LiDAR company announced at CES 2025 that it had been selected as the LiDAR supplier by Mobileye. Founded in 2016, the company has been developing its MEMs (Micro-Electro-Mechanical Systems) based LiDAR for the past 8 years. It went public in 2021 through a SPAC (Special Purpose Acquisition Company) merger.

The InnovizOne has been designed into BMW’s BMW Personal Pilot and manufactured in partnership with Magna International in an automotive-grade production facility. The next generation InnovizTwo is more compact, with a higher range and resolution, designed to work through obscurants (mud, insect splatter, rain).

A few events and their timelines are important to understand Innoviz’ selection as a LiDAR partner for Mobileye and Volkswagen:

1) Early 2023: Volkswagen announced Mobileye as its autonomy stack partner for a mid-decade L4 autonomous Light Commercial Vehicles (LCV) launch in 2023. This included L4 trials with Volkswagen ID Buzz electric vans.

2) Mid-2023: Innoviz announced in a SEC filing that “a major existing global automotive customer aims to expand its use of the InnovizTwo LiDAR to a new light commercial vehicle program”. This likely means that Innoviz has been working with Mobileye on its Innoviz LiDAR integration for at least 2 years.

3) Mid-2023: The next generation InnovizTwo was selected in 2023 by Audi which designated the company as its Tier 1 supplier, following an extensive round of audits on multiple fronts - manufacturing, costs, quality and production strategy. It positions Innoviz as the primary choice for LiDAR for other Volkswagen models for supporting L3 and L4 autonomy.

4) Late 2024: Mobileye announces shutdown of its internal LiDAR effort.

5) Late 2024: Innoviz announces $80M in NRE (Non Recurring Engineering) funding for designing its InnovizTwo LiDAR into different OEM platforms.

6) Early 2025: Innoviz announced at CES 2025 that it had been selected as the LiDAR supplier by Mobileye, for the Chauffeur™ and Mobileye Drive™ programs.

According to Omer Kielaf, CEO of Innoviz, its success in getting designed into major OEM platforms “are our history of deep working relationships, disciplined project management, investing in becoming a Tier 1 designated supplier, the cost-performance-durability-size of the InnovizTwo LiDAR, and the ability for customers to source short and long range lidars from the same supplier”.

As part of its drive to move into higher levels of autonomy (L3 and L4), Innoviz is also making investments in hardware and perception software developments.

On the hardware front, making the InnovizTwo resilient to optical path obscurations (mud, rain, insect splatter) is a critical requirement for autonomous operation. Per Mr. Kielaf, having such obscurations in an L2 system where a human operator is always in control is not a significant problem. With an L3 system, it becomes an issue because it now requires the human driver to disengage L3 capability and takeover, reducing the convenience of the feature. With L4 (no human driver), this is an absolute requirement since loss of perception at high speeds can be lethal. The optical design of the InnovizTwo LiDAR ensures redundancy in the optical path for the transmit and receive path for each pixel (see Figure 3). As an example if the window is 25% covered, no single pixel will be “dead” and some pixels will lose about 7% of the range performance.

On the software front, Innoviz announced a partnership with NVIDIA which will enable delivery of high-performance, AI-enabled perception systems to its customers. The software will be “optimized to run with the NVIDIA DRIVE Orin platform, allow for real-time processing and advanced understanding of the vehicle's environment, enabling exceptional object detection, classification, and tracking capabilities”. Per Omer Kielaf, "By leveraging the NVIDIA DRIVE Orin platform, we will be able to offer a powerful combination of cutting-edge LiDAR sensors and perception software, enabling automakers to scale from L2+ ADAS all the way to fully autonomous vehicles".