Innoviz Technologies (NASDAQ:INVZ) used a “Physical AI Webinar” to discuss a newly released white paper and explain how the company views the role of LiDAR in “Physical AI,” “World Models,” and the evolution of the automotive LiDAR market. CEO and founder Omer Keilaf and David, Innoviz’s Director of Industry Solutions, also outlined customer programs the company says are underway in automotive autonomy and described why Innoviz believes automotive-grade sensing requirements will influence adoption in other industries.
Physical AI and “World Models” as a framework for LiDAR’s next markets
Keilaf contrasted “Digital AI”—models trained on text or images to generate content—with “Physical AI,” which he described as AI applied to unstructured real-world environments where systems must interpret surroundings and make decisions amid uncertainty. He said autonomous driving was the first “designed to scale” application of Physical AI.
Keilaf said Innoviz chose to focus on the topic now because interest in applying AI beyond autonomous driving is increasing, and he argued that many industries require models trained on real data to handle unstructured, uncertain environments. He added that the “Physical AI” terminology provides an “umbrella” for explaining Innoviz’s addressable markets beyond automotive, where LiDAR is often assumed to be relevant only to self-driving cars.
Why Innoviz argues 3D sensing is central to Physical AI
Keilaf called perception a bottleneck for Physical AI and said LiDAR is often referred to as “ground truth,” even when training vision-based systems. He argued that models built on inferred data or simulations risk bias compared with real-world truth, and he said high-capability 3D sensors can provide highly accurate data for building better 3D models.
He said the white paper was released in two parts: the first covering the concepts of Physical AI and World Models and discussing market size, and the second offering his view of how the LiDAR market is likely to evolve, including continued consolidation among suppliers.
Automotive LiDAR: consolidation, shifting requirements, and “availability” for Level 4
Keilaf described the automotive LiDAR market as a “long game” that has moved through hype-driven disruption and consolidation. He said there were “probably 200 or maybe more LiDAR companies” over the last decade, with billions of dollars invested, but that the automotive supply base typically consolidates to a small number of suppliers over time. He argued that frequent changes in OEM requirements—from 2013 through 2016, 2019, and 2026—have made it increasingly difficult for many companies to “make the cut.”
He also walked back a prior view that OEM “awards” were the key measure of a LiDAR company’s success, saying that in retrospect some companies received awards but still failed. In his view, the harder challenge is moving from award to deployment, and he said the second part of the white paper addresses what is needed to reach “serious production.”
A central point in the discussion was what Keilaf called the “dirty secret” of the LiDAR space in the context of Functional Safety and redundancy. He argued that redundancy requires non-correlated failure modes between sensors, and he gave examples where cameras fail (direct sun, weather, rain droplets) while claiming many LiDAR architectures also struggle in those conditions. He highlighted occlusion from mud, bugs, and other debris as a critical challenge—particularly for Level 4 systems with no driver fallback.
Keilaf said Innoviz designed its second-generation system based on real-world lessons from its first-generation product, and claimed the company redesigned optics so that even if mud is thrown on the sensor, it can still “see” well. He emphasized that, in his view, the key distinction between LiDAR for Level 2, Level 3, and Level 4 is sensor availability rather than range or resolution.
- Level 2: driver acts as redundancy; LiDAR availability is less critical.
- Level 3: higher availability required; driver may be asked to reengage if sensors degrade (for example, from mud).
- Level 4: no driver fallback; Keilaf described the availability requirement as a “quantum leap.”
He said automotive requirements and volumes set a high bar that can benefit other Physical AI markets by pushing performance, safety compliance, and cost down through scale.
Technology cycles and Innoviz’s view of the “right” automotive approach
Keilaf discussed what he characterized as repeated “hype” around emerging LiDAR approaches over the past decade. He cited a January 2016 press release from Quanergy promoting Optical Phased Array (OPA) as an example, saying he concluded early that the approach “doesn’t work,” despite investor interest at the time. He also referenced past debate around 905 nm versus 1550 nm approaches, and said new “hypes” continue to emerge.
Looking forward, he said OEMs have long preferred behind-the-windshield installation for aesthetics but that it requires a much smaller device, significantly lower power consumption due to solar load, and higher performance to overcome windshield attenuation. Keilaf said Innoviz’s third-generation product, InnovizThree, is designed for this shift, while InnovizTwo is expected to enter series production “end of the year” with Volkswagen.
He argued that these behind-the-windshield demands may drive further supplier reduction, and said he believes some technologies—he listed OPA, 1550, and FMCW—are not a good fit for automotive. He stated that, in his view, time-of-flight (ToF) at 905 nm is the only technology that has achieved the needed combination of volume, price, performance, and power consumption for automotive applications.
Competition, national security concerns, and customer programs
On the competitive landscape for automotive ToF 905 nm LiDAR, Keilaf said that “in the West” it is “primarily us and Valeo,” alongside Chinese vendors. He said Chinese vendors benefited from government support and from strong demand tied to Level 2 systems in China, which he said helped the ecosystem develop.
Keilaf also pointed to U.S. policy activity tied to national security concerns, saying there are “several acts now in the U.S.” regarding the use of LiDAR from China, including the “SAFE LiDAR Act,” and he referenced a “new act” he said had recently come from the Department of Transportation. He framed LiDAR as a mapping tool that could be deployed across cars, infrastructure, borders, robotics, and drones, generating sensitive data via “digital twin” capabilities. He said he believes the market may remain split between Western suppliers outside China and Chinese suppliers operating in China.
On commercial traction, Keilaf cited multiple programs:
- Work toward a Level 4 launch with Volkswagen for the ID. Buzz, which he described as a “serious production robotaxi in Europe by an OEM.”
- Collaboration and a commercial agreement with Mobileye, which he said is using Innoviz sensors on other programs.
- Use of nine LiDAR per vehicle in Level 4, and reference to another platform, HOLON, which he said is expected to reach start of production after the ID. Buzz.
- Daimler Truck selecting Innoviz sensors, described as using multiple sensors around the vehicle.
- An Audi Level 3 program.
- Ongoing work on RFIs and RFQs, with Keilaf saying he expects additional awards this year tied to either InnovizTwo or InnovizThree depending on customer timelines and windshield integration needs.
Keilaf positioned automotive as Innoviz’s “base camp” for expansion into non-automotive markets, highlighting security as an early focus area. He argued that LiDAR can address limitations of radar in complex visual scenes and described scenarios such as security perimeters where, in his view, radar resolution is insufficient. He said Innoviz is pursuing premium performance-driven opportunities in security and expects the company’s automotive-grade capabilities—especially durability and long-range performance—to translate to other Physical AI applications.
About Innoviz Technologies (NASDAQ:INVZ)
Innoviz Technologies Ltd. (NASDAQ: INVZ) is a developer of high-performance solid-state LiDAR sensors and perception software designed to support advanced driver assistance systems (ADAS) and autonomous driving applications. The company’s core business focuses on providing automotive-grade LiDAR hardware, along with software algorithms that enable accurate 3D mapping, object detection and environmental perception in real time. Innoviz’s technology is tailored for integration into passenger vehicles, commercial fleets and other mobility platforms seeking improved safety and autonomy.
Founded in 2016 and headquartered in Rosh Ha’ayin, Israel, Innoviz has emerged as a key supplier to leading global automakers and Tier 1 suppliers.
