RJ Scaringe, Rivian’s founder and chief executive, cast the new technology platform as a pivotal turn for the fast-growing EV brand. He pointed to the company’s homegrown 1,600 sparse-TOPS inference chip as a catalyst for dramatic progress in autonomy, ultimately aimed at reaching Level 4 capability — driving with little to no human oversight in most conditions. “This represents an inflection point for the ownership experience,” Scaringe said, emphasizing Rivian’s goal of giving drivers “their time back” as cars take on more responsibility.
Much of Rivian’s confidence hinges on the Rivian Autonomy Processor, or RAP1, a 5-nanometer chip designed in-house to support the company’s vision-centric approach to physical AI. By integrating processing and memory in a single multi-chip module, RAP1 aims to boost efficiency, sharpen performance and satisfy rigorous safety standards. Its RivLink interconnect technology allows multiple chips to be linked at low latency, giving the system the ability to scale as compute demands rise. RAP1 also runs on Rivian’s own compiler and platform software, reinforcing the company’s push toward vertical integration.
The processor anchors Rivian’s third-generation Autonomy Compute Module, ACM3, capable of delivering 1,600 sparse INT8 TOPS and processing more than 5 billion pixels every second — numbers intended to deliver the split-second perception and decision-making required for advanced autonomy. Rivian also confirmed plans to add LiDAR to its coming R2 lineup, bringing dense 3-D mapping, redundant sensing and sharper detection for rare but critical driving scenarios. The new hardware, including LiDAR and ACM3, is now in validation and expected to reach R2 vehicles by late 2026.
Software was an equal star of the day. Rivian outlined its autonomy platform and described a comprehensive data-training loop built around what it calls the Large Driving Model — an autonomous-driving analogue to the large language models now common in consumer AI. Trained with Group-Relative Policy Optimization, the model is intended to extract sophisticated driving behaviors from immense datasets and translate them into safer, smoother decision-making behind the wheel.
In the nearer term, Rivian said owners of its second-generation R1 vehicles will soon see major upgrades through Universal Hands-Free, or UHF, an extended-duration assisted-driving feature covering more than 3.5 million miles of U.S. and Canadian roads. The system will operate even on non-highway stretches, provided lane markings are clear.
The company also announced Autonomy+, a subscription package launching in early 2026 priced at $2,500 upfront or $49.99 per month. Rivian said the bundle will steadily expand and could become a significant business driver as new safety and convenience features roll out. Longer-term updates planned for Gen 2 R1s and future R2 vehicles include point-to-point routing, eyes-off capability and eventually personal Level 4 autonomy.
Rivian’s AI ambitions stretch beyond driving. Through its Rivian Unified Intelligence platform, the automaker is weaving multimodal data and multiple LLMs into service operations, predictive maintenance and new user features. The highlight is the Rivian Assistant — a next-generation voice interface arriving in early 2026 for both Gen 1 and Gen 2 R1 models. Built on edge-AI models tuned to understand the vehicle and the driver’s digital life, the assistant will link car systems with third-party services such as Google Calendar and will tap large language models for more natural conversation and reasoning.
Rivian says this integrated approach — from custom chips to cloud frameworks to in-car interfaces — will accelerate how quickly it can update and refine the entire vehicle experience. As the company moves into its next phase, it believes the combination of deep vertical integration and rapid innovation will help it stand out in the fiercely competitive EV landscape.