Mobileye Unveils 'Version 3.0' at CES 2026: Massive ADAS Pipeline, VW Robotaxi Launch, and the Leap into Humanoid Robotics

Mobileye CEO Professor Amnon Shashua took the stage at the 2026 Consumer Electronics Show (CES) to announce a transformative era for the company, officially dubbing it "Mobileye 3.0". Highlighting a massive $24.5 billion future pipeline and the strategic acquisition of Mentee Robotics, the presentation underscored Mobileye’s transition from an automotive technology supplier into a dominant force in the broader realm of physical artificial intelligence.

Joined by Christian Senger, CEO of Volkswagen Autonomous Mobility, Shashua detailed aggressive timelines for driverless commercial rollouts, next-generation silicon benchmarks, and an industry-first approach to scalable humanoid robotics.

Unprecedented Automotive Pipeline and ADAS Evolution

Reflecting on an exceptional 2025, Mobileye revealed that it successfully secured 95% of all Requests for Quotations (RFQs) among its top 10 customers and onboarded two new major Original Equipment Manufacturers (OEMs): Volvo and Subaru.

Key automotive business metrics and milestones include:

• Record Future Revenue: Mobileye's eight-year pipeline—based strictly on signed contracts with OEMs—now stands at $24.5 billion, a figure whose annualized average is already 50% larger than the company's 2025 total revenue of nearly $2 billion.

• Massive Deployments: The company has deployed over 230 million chips since its inception, effectively equipping approximately one-eighth of all cars currently on the planet with its advanced driver-assistance systems (ADAS) technology.

• Mapping at Scale: Mobileye’s Road Experience Management (REM) network is continuously updated by a fleet of eight million vehicles, which harvested a staggering 32 billion miles of data in 2025 alone.

• The Shift to Surround ADAS: The industry is rapidly moving from standard front-facing cameras to comprehensive "Surround ADAS," powered by the EyeQ6 High chip using multiple cameras and radars. Mobileye recently secured a design win with a major US OEM, adding to a previous Volkswagen agreement, yielding a combined total of 19 million units slated for start of production (SOP) by mid-2028.

Scaling Robotaxis: The Volkswagen and MOIA Ecosystem

A core component of the presentation focused on Mobileye’s Level-4 autonomous driving partnership with Volkswagen and its mobility service brand, MOIA. The collaboration integrates Mobileye's self-driving system into the purpose-built ID. Buzz AD, creating a turnkey autonomous Mobility-as-a-Service (MaaS) product.

The timeline and scale targets for the robotaxi initiative are highly aggressive:

• Near-Term Milestones: The Level-4 ready vehicle will be completed in February 2026, paving the way for driverless operations to begin in the United States in the third or fourth quarter of 2026.

• European Expansion: EU homologation is anticipated in 2027, launching a rapid market expansion alongside partners like Ruter, Holo, and Uber.

• Long-Term Targets: The MOIA ecosystem aims to launch in six cities by the end of 2027 and scale to operate more than 100,000 active self-driving vehicles on the road by the end of 2033.

Christian Senger emphasized that autonomous mobility is no longer purely a technology hurdle, but rather a "scaling and business model challenge," which the specialized Volkswagen-Mobileye ecosystem is uniquely positioned to solve.

Next-Generation Silicon and the "Fast and Slow" AI Stack

To support unprecedented scaling, Shashua outlined Mobileye’s sophisticated, dual-layered technology architecture, actively critiquing simplistic, direct-trajectory approaches to Generative AI in autonomous driving. Instead, the company employs a "fast and slow" model.

• Fast System (10 frames per second): Manages all immediate safety protocols and collision avoidance algorithms.

• Slow System (1-2 frames per second): Utilizes a Vision-Language-Semantic-Action (VLSA) model to process complex, non-standard traffic scenarios (such as safely yielding to a police officer while ignoring a red traffic light). Acting as a virtual "accompanying adult," the VLSA comprehensively interprets scenes and outputs semantic instructions to the planning algorithms, which will asymptotically eliminate the need for remote human tele-operators.

• Artificial Community Intelligence (ACI): To train complex driving policies, Mobileye uses self-play reinforcement learning on non-photorealistic simulators generated from its global REM maps. By populating these maps with agents exhibiting thousands of diverse and sometimes erratic behaviors, Mobileye can simulate and train over billions of hours overnight without being bottlenecked by heavy computational requirements.

Powering these complex software architectures is Mobileye's custom silicon. The new EyeQ7 chip, ready for production in Q3 2027, features 22 accelerator engines and 12 CPU cores. Mobileye asserts that the EyeQ7 significantly outperforms competitors; for instance, it processes nine-million parameter vision transformers at a latency of just 0.5 milliseconds, compared to the Nvidia OrinX, which registers at 1.5 milliseconds. Crucially, the EyeQ7 is capable of running a massive 15.6 billion parameter language model directly on-board.

The Dawn of Mobileye 3.0: Entering Physical AI via Mentee Robotics

Concluding the keynote, Mobileye announced its strategic acquisition of Mentee Robotics, marking its official entry into the broader field of physical artificial intelligence. Shashua noted that autonomous cars and humanoid robots share profound technological synergies, relying heavily on fast and slow AI architectures, massive simulation, and Sim2Real (simulation to reality) transfers.

Mentee Robotics' third-generation model showcases remarkable, vertically integrated engineering designed to operate in open-ended, unstructured environments.

• Hardware Specifications: Standing 175 centimeters tall and weighing 72 kilograms, the robot features a heavy-duty 25-kilogram payload capacity and swappable batteries for uninterrupted 24/7 operation.

• Innovative Hands: Instead of relying on costly tendon-based designs equipped with fragile fingertip sensors, the robot utilizes rigid links connected directly to its motors. This dramatically reduces manufacturing complexities while maintaining high-fidelity human-like dexterity.

• AI Autonomy: Operating entirely via full-stack AI with zero tele-operation, the system is driven by two Nvidia OrinX chips, demonstrating powerful out-of-the-box capabilities for natural language instruction following and internal cognitive mapping.

A core differentiator for Mentee Robotics is its proprietary Real2Sim2Real learning architecture. To allow the robot to adapt dynamically to varied, unpredictable environments like private residences, a human user can "mentor" the machine by physically demonstrating a novel task. The robot records this observation, sends the video to the cloud to rapidly construct a simulation, trains on the new task via reinforcement learning, and seamlessly transfers the capability back to physical execution in reality.

Commercial deployment with signed production partner Aumovio (formerly Conti) will execute in two distinct stages. Proof-of-concept (POC) units will begin with multiple customers in 2026, leading to a commercial rollout in 2028 for highly structured industrial environments, such as fulfillment centers that currently suffer from severe labor shortages. Consumer deployments for unstructured environments, specifically targeting spaces like elderly home care, are slated for 2030.

With its absolute dominance in ADAS secured, an imminent commercial launch for its robotaxi ecosystem, and a disruptive new growth engine in humanoid robotics, Mobileye is aggressively positioning itself to lead the next frontier of the physical AI revolution.