Rivian’s AI Shift: Aiming Beyond Just Competing with Tesla

RJ Scaringe is in Rivian’s Palo Alto office, discussing the company’s bold move into self-driving technology when a Waymo robotaxi glides by outside. An analyst from Goldman Sachs hops in, snapping a quick selfie, prompting a chuckle from Scaringe. “That’s amazing,” he laughs. “So perfect.”

The appearance of Waymo underscores the challenges Rivian faces. Earlier, Scaringe addressed a crowd of investors and reporters, unveiling Rivian’s ambitious plan to develop its own AI chips aimed at achieving higher levels of autonomy, ultimately targeting Level 4, where no human supervision is needed in specific scenarios. The question looms: can Rivian accomplish in a shorter timeframe what has taken Waymo years to develop?

More critically, can it outpace Tesla, another EV company navigating its own complex transition into AI and robotics? Founded as the rugged alternative to Elon Musk’s brand, Rivian now seems to be racing down the same AI path.

Scaringe emphasizes that this shift isn’t merely a reaction to Tesla but a strategic response to the auto industry’s evolving landscape. Advances in transformer-based encoding and large-parameter models have reshaped Rivian’s approach to the “physical AI” of autonomous driving. Consequently, the company initiated a complete redesign of its autonomy platform in early 2022. With the launch of its Gen 2 platform, Rivian is building a “data flywheel” to train a comprehensive driving model using real-world data from its fleet. This end-to-end training means that enhancements in sensors or computing power directly boost the system’s capabilities, allowing for continuous improvement as technology advances.

Rivian must either forge its own path in autonomy or risk being overshadowed by Waymo, Tesla, and others. Throughout his presentation, Scaringe insists this isn’t a trend-chasing move but a result of years of dedicated work and thoughtful design.

“I spent some time last night with the team talking about this right before we’re about to show it today,” Scaringe shares onstage. “One of the lead engineers looked at me and said, ‘Boy, we’ve been working on this for years, and I haven’t been able to talk about it. It’s so cool. Tomorrow I can start to talk about what I do every day, all day long.’”

Rivian is eager to share its vision. The company’s “AI and Autonomy Day” resembles a high school science fair, showcasing various projects: a new chip for its AI goals, an AI-powered voice assistant that can recommend wineries or songs, and a lidar sensor for creating a 3D map of the environment. There’s even an R2 styled like R2-D2 from Star Wars.

The highlight is a test drive featuring Rivian’s new hands-free point-to-point capability. A software update allows Rivian owners to drive hands-free on 3.5 million miles of roads in the US and Canada. The full point-to-point feature will be available later, enhancing partial autonomy akin to Tesla’s Full Self-Driving.

Starting next year, if roads have clear lane markings, drivers can engage hands-free driving as part of a paid package called Autonomy Plus, available for a one-time fee of $2,900 or a monthly subscription of $49.99. In contrast, Tesla’s premium FSD option costs $8,000 upfront or $99 monthly. Rivian’s Autonomy Plus will be free for all customers until March 2026.

During my 45-minute experience in an R1S, an engineer drives while ready to take control if needed, but mostly keeps his hands off the wheel as the vehicle navigates intersections, traffic signals, and various road users. The car performs well, with the engineer only intervening once to avoid a red light. At one point, we maneuver between a passing truck and a cyclist, and while it feels close, the car maintains its lane without hesitation.

The system avoids the jerkiness often seen in first-gen driver-assist tech, likely due to its combination of 11 cameras and five radar sensors, providing a comprehensive view of the surroundings. In contrast, Tesla’s FSD relies solely on cameras.

As more vehicles with partial automation hit the streets, safety experts are increasingly concerned about driver attention and the potential for accidents. Rivian believes in a collaborative approach, allowing drivers to adjust steering or acceleration without disengaging the system. Nick Nguyen, director of product and programs for autonomy, explains, “We want you to interact with it as much or as little as you want.”

The real test lies ahead. Rivian’s Gen 3 system will incorporate its proprietary silicon chip and the same sensor suite as Gen 2, with the addition of lidar, a technology not commonly found in personal vehicles but prevalent in robotaxis like Waymo. Rivian plans to include lidar in its upcoming R2 model, taking a significant risk.

While lidar is expensive, costs are decreasing. Rivian hasn’t revealed its lidar supplier, but the difference between a camera-only system and one with multiple sensors is significant. During the presentation, James Philbin, Rivian’s VP for autonomy and AI, demonstrated how the software identifies objects using only cameras, then with cameras and radar, and finally with the full suite of cameras, radar, and lidar. The latter setup excelled, detecting hidden objects and pedestrians that the other configurations missed.

Scaringe clarified that the R2 will begin production next year without lidar or the Gen 3 autonomy computer, which will be added by the end of 2026. Some Rivian fans expressed frustration over the timing, but Scaringe believes demand for the R2 is so strong that it won’t significantly impact sales.

What could affect sales is an unreliable system. However, Scaringe assures that Gen 3 is robust. The AI computer will feature a dual-chip setup capable of 1,600 trillion operations per second, a figure that would have seemed impossible a few years ago.

What truly matters is the speed at which the computer processes information, such as pixels for camera-based systems. Rivian claims its Gen 3 will process 5 billion pixels per second, a specification not commonly discussed in the AI realm, but one Scaringe believes will impress. “Tesla recently talked about theirs,” he notes. “I’ll let you Google it, but it’s a lot less.”

The next major milestone is “eyes-off” driving, where users won’t need to monitor the road and can fully reclaim their time for other activities. Beyond that lies personal Level 4 capability, where the vehicle operates entirely on its own in certain conditions, without human oversight.

Whether Rivian reaches this goal will depend on building trust in its products, which requires a transparent approach to legal liability that many automakers, including Rivian, have yet to fully address. Scaringe points out that as autonomy improves, human driving will decrease from 80–90 percent of miles today to 10–20 percent in a few years, eventually reaching zero. At that point, a driver’s skill level becomes irrelevant to risk assessment, necessitating a shift in insurance models. Rivian has partnered with Nationwide for insurance claims but has yet to clarify how it will handle liability for accidents involving its autonomous vehicles.

“This is like the ticky-tacky of the real work streams to make this all real,” Scaringe explains. “Beyond the technology, but actually, like, the business systems that need to be designed.”

In Palo Alto, reactions to Rivian’s announcements are positive. Influencers and analysts express their excitement about the developments. However, Rivian’s stock dropped 6 percent on the day of the event, closing at $16.43 per share, while Tesla continues to trade above $460. Although Rivian’s stock has slightly recovered, the company faces significant challenges, including the loss of the EV tax credit and the substantial capital required to realize its AI ambitions.

The question remains: why is Rivian pursuing this path? While some fans may be puzzled by the company’s shift toward AI, it’s a necessary step to remain relevant in a changing landscape, according to mobility investor Reilly Brennan.

“Having an AI strategy is becoming essential for public auto companies in this era — especially for those aiming for high market valuations, like Tesla,” Brennan explains. “It’s akin to overachieving high school students trying to get into Harvard: it’s not mandatory to start a charity or play the contrabassoon, but it seems like everyone who gets in is doing it.”

There’s a mix of peer pressure and the need to discover new revenue streams. A $50-a-month autonomy subscription could provide a lucrative income source for Rivian, which must achieve profitability to thrive.

Scaringe is a firm believer in the potential of AI. He envisions a future where AI becomes as commonplace as “running water and electricity.” However, he doesn’t address the environmental impact of AI, which could pose a challenge for a company that claims to prioritize sustainability.

Rivian isn’t fully committed to following Musk’s relentless pursuit of robotaxis and humanoid robots, though Scaringe acknowledges that both remain possibilities. The company established its own robotics division, Mind Robotics, earlier this year.

“The part of the problem that’s not solved is the Level 4,” he concludes, standing up as my time runs out. “And so that’s our focus, on the tech.”