According to Sawyer Merritt, Rivian has announced its next-generation Autonomy hardware and software platform, featuring a self-improving, end-to-end AI system designed for scalable autonomous driving (source: Sawyer Merritt, Twitter). The system integrates 11 cameras (65 megapixels), 5 radars, and a new front-facing LiDAR, all powered by an in-house RAP1 chip manufactured with TSMC's 5nm process. Rivian's software-first approach leverages the Rivian Autonomy Platform and an end-to-end data loop for continuous training. Notably, the company introduced its Large Driving Model (LDM), a foundational AI model for autonomy trained similarly to large language models, using Group-Relative Policy Optimization to distill advanced driving strategies from vast datasets. This integrated approach positions Rivian as a serious competitor in the AI-powered self-driving vehicle market and highlights opportunities for AI-driven innovation in automotive autonomy.

According to Sawyer Merritt, Rivian has introduced its RAP1 Autonomy chip and third-generation Autonomy Computer (ACM3), targeting advanced autonomous driving applications. The RAP1 chip, manufactured using TSMC's 5nm process, features a custom neural engine, 800+ TOPS compute, and an in-house software stack. ACM3 achieves 1600 sparse INT8 TOPS and processes 5 billion pixels per second. Rivian's proprietary RivLink technology allows low-latency interconnect for scalable processing. The AI platform is supported by Rivian's own compiler and software. Rivian plans to integrate LiDAR with ACM3 in future R2 models, enhancing sensor redundancy and real-time detection for complex driving scenarios. Validation is underway, with commercial deployment expected by late 2026 (source: Sawyer Merritt).