As an audience member at yesterday's AI Days organized by $RIVN (+0,73%) yesterday, I was impressed by the clarity with which Rivian presented its technological turnaround, which will take the company from a purely "software-defined vehicle" to an "AI-defined vehicle".
The rate of change that awaits us through artificial intelligence will be an order of magnitude higher in the next three to four years than in the recent past. The key to this development lies in Palo Alto, the center for autonomy, where Rivian has decided to directly control the entire architecture and software of its vehicles. This is the basis for enabling AI throughout the business and in the vehicles.
The autonomy strategy has fundamentally shifted: Away from an old, rule-based approach (where objects were recognized and then handed over to strict human-defined rules) to an AI-centric approach. Today, it's about building a neural network-like understanding of driving, supported by innovations such as transformer-based coding. This led to a complete reboot of the platform in 2022.
The Gen 2 R1 vehicles, launched in mid-2024, epitomized this first step, equipped with 55 megapixel cameras and five radars running on an inference platform ten times more powerful than Gen 1. This AI-centric system is continuously improved by the so-called Data Flywheel: each vehicle driven collects relevant data that is used to train the large driving model, which in turn is fed back to the vehicles via updates. Better sensors and computing power constantly improve what has already been learned and the ability to master nuances and complex situations, just like when a driver puts on glasses and gets a better "brain".
The next big technological leap will come with the Gen 3 autonomy platform, which will be launched at the end of 2026 with the R2. Here, Rivian is focusing on consistent vertical integration, meaning that key technology components will be developed in-house. The Gen 3 will include 65 megapixel cameras, an improved radar arrangement and, for the first time, a front LiDAR (light radar). LiDAR is important because it provides optical data, but unlike cameras (which are weak in poor light or fog), it uses an active light source and therefore provides very precise 3D models of the driving space.
The most important innovation, however, is the in-house Rivian Autonomy Processor (RAP1). This is the first in a series of special chips for physical AI. RAP1 achieves 800 TOPS (which stands for **trillions of operations per second** and is a measure of AI performance) and enables a total performance of 1,600 TOPS in the Gen 3 compute platform. The reason for developing this silicon in-house is clear: speed, performance and cost reduction, as there is no margin from suppliers and the design can be perfectly optimized for the use case in the car. Rivian is planning an aggressive expansion of the autonomy functions: Hands-free driving will be expanded to over 3.5 million miles in North America as early as this month (previously under 150,000 miles).
Point-to-point capability will follow from 2026, with the vehicle driving completely autonomously from start address to destination address. Eyes Off (driving without having to watch the road) and finally Personal Level 4 (the vehicle operates fully autonomously, e.g. for pick-up or delivery services) are then planned. These advances are backed up by a large driving model developed in-house and a cloud-based simulator that runs through millions of scenarios before new functions are released.
Beyond autonomy, Rivian Unified Intelligence (RUI) is used as a common AI foundation throughout the company. RUI is already helping with quality control in manufacturing and reducing repair time in service by hours by supporting technicians with AI systems trained on real-time data and repair histories. For customers, this intelligence will be visible from early 2026 in the Rivian Assistant, which will be available to all Gen 1 and Gen 2 customers. The assistant uses natural language to execute complex vehicle commands (e.g. only warming certain seats), remembers the context and history of the conversation and massively simplifies interaction with the vehicle.
In summary, years of laying the foundations of the software architecture and electronics are now bearing fruit and the company is entering a phase of exponential improvement, comparable to an upward helix where each cycle of the data flywheel leads to a better result. Rivian is thus positioning itself as one of the leading players fully exploiting the potential of AI in the automotive industry.
I look forward to further developments and will keep you up to date.
