Rivian Goes All-In on Autonomy: In-House Chips, Lidar and an AI Helper That Handles Diagnostics

• Rivian designed its own silicon chips to power an improved autonomy system. 
• The R2 will be available with lidar sensors for even better hands-free performance.
• Drivers will soon be able to chat with the Rivian Assistant and get AI help in detecting vehicle issues.

At its first Autonomy & AI Day celebration, Rivian announced some big updates to its onboard processing and sensing hardware to improve autonomy capabilities, as well as the addition of an AI-based voice assistant and AI-backed diagnostics. This jam-packed event is certainly true to its name.

In-house chips

Key ingredient No. 1 is custom silicon. The first-generation Rivian Autonomy Processor (RAP1) chip powers the third iteration of the company's Autonomy Compute Module (ACM3). The architectural updates are aimed at eventually unlocking what the company calls Personal Level 4 autonomy, which goes beyond hands-off and eyes-off operation and would theoretically allow an owner to summon an empty vehicle or send the kids off with no one behind the wheel.

Nerdy RAP1 details: automotive-grade 5-nanometer architecture produced by TSMC, 14 processing cores, 8-core safety island for ISO26262 compliance, 800 trillion operations per second, integrates processing and memory into one module, uses an in-house developed AI compiler and platform software. The ACM3 uses a pair of liquid-cooled RAP1s and can handle 5 billion pixels per second and features RivLink, a proprietary low-latency interconnect that allows multiple chips to combine forces for increased processing power.

Lidar coming to R2

The extra computing power in turn allowed Rivian to add a lidar sensor to its upcoming R2 midsize SUV. Lidar creates a three-dimensional point cloud view that tells the vehicle the shape of its surroundings, from which it can infer the speeds of other vehicles and pedestrians. This additional modality will join the cameras and radar units that are already part of Rivian's sensing suite. For the R2, however, the ultrasonic sensors have been dropped and their functionality replaced by corner radar units that can handle short-range detection. 

Yes, the demonstrator vehicle is stickered up to look like R2-D2. Cute. The new sensor doesn't pop up on a telescoping pole (wouldn't that have been cool?) but is located at the forward edge of the roof just above the windshield, providing an optimal forward view.

The new hardware and lidar sensor won't be available on the R2 from launch but should roll out toward the end of 2026. The second-generation R1 platform, meanwhile, sticks with a pair of Nvidia Drive Orin processors and its current sensor suite.

Upgraded hands-free driving for R1

Rivian also shared a little bit more about its autonomy software. It's trained on what the company calls its Large Driving Model, which is similar to the large language models used in AI apps like ChatGPT. Rivian says this uses Group-Relative Policy Optimization applied to huge datasets to teach its vehicles how to drive. This end-to-end system is designed to improve itself over time using data from the deployed fleet.

Second-gen R1T trucks and R1S SUVs will get a Universal Hands-Free driving mode available on 3.5 million miles of roads in the U.S. and Canada that include highways as well as secondary roads with clearly painted lines. Access will be through an Autonomy+ package that can be purchased for $2,500 outright or subscribed to for $49.99 per month. The company has promised stepped improvements to Gen 2 R1 as well as R2 hands-free software, first with point-point operation in early 2026, then with eyes-off functionality. Personal L4 will follow at some point but may require additional compute hardware, likely more than the pair of processors that will be offered on the R2.

AI for everyone

As for artificial intelligence, Rivian plans to embed it into all aspects of its business. Starting early next year, drivers of both generations of R1 will meet the Rivian Assistant. It's a voice interface that can leverage multiple models for natural-language control and will integrate with third-party apps, the first being Google Calendar. From launch, the conversational voice interface will handle tasks like providing an idea of the user's schedule and navigating to the person's next appointment; deep integration lets it control various vehicle systems like climate and audio. And by popular demand, Rivian has added speech-to-text SMS messaging that works over Bluetooth. It's not as slick as the Android Auto and Apple CarPlay implementations, but it's something.

There are two wake phrases — "hey, Rivian" and "OK, Rivian" — that can be disabled for privacy or annoyance reasons, or you can just press the left steering wheel button to activate the assistant. The R2's infotainment system will have even more compute power than what's in the R1 now, so it will handle more AI tasks on-vehicle without having to rely on the cloud as much. 

Service technicians will also benefit from AI as a helper in finding and diagnosing issues. In the future, the same tech will show up on the vehicle and in the Rivian app to allow owners to diagnose issues directly.

There's a lot to understand here, and we'll be getting some hands-on time later today to better suss out the details, so stay tuned. And don't forget that Rivian has a deal to supply Volkswagen with software, so much if not all of this may translate to the German automaker's EVs in the near future.