Rivian’s CEO Sees an AI Reckoning Coming for Automakers Stuck in the Past

The software struggles of today could become the AI problems of tomorrow, according to Rivian CEO RJ Scaringe.

Over lunch at a campground in Utah’s Wasatch Mountains during the launch event for the new 2027 Rivian R2, Rivian CEO RJ Scaringe dropped by to field a few questions from the media. The assembled journalists peppered Scaringe with queries ranging from the R2’s production numbers and pricing to shrinking EV sales in the U.S. and the growing global threat of Chinese vehicles. But the longest and most animated answer came when MotorTrend asked Scaringe to elaborate on an earlier comment predicting trouble ahead for legacy automakers that have struggled with the transition to software-defined vehicles (SDVs)—we’re looking at you, General Motors and Volkswagen—all as a new wave of AI-defined vehicles (AIDVs) start to emerge.

A Very Significant Gap

Scaringe believes AIDVs and their benefits will quickly become table stakes for automakers and demanded by owners, thanks to the ability to continuously improve upon a vehicle during its life cycle and enhance the overall ownership experience by stitching together digital services in and out of the car. “When you start to think beyond software-defined vehicles and to AI-defined vehicles, there’s a very significant gap that I think the vast majority of Western manufacturers have a real challenge with,” Scaringe said. “And I definitely think the gap is going to widen.”

An SDV will leverage AI capability to aspects of a vehicle ranging from autonomous driving systems to in-cabin digital assistants, said Greg Basich, associate director at the analyst firm Counterpoint Research. In general, legacy automakers are behind, but they’re catching up, he said. “Ford, GM, and Volvo are all upgrading to Google Assistant that uses Gemini,” Google’s AI assistant,” Basich added.

Although legacy brands tend to focus first on infotainment when it comes to AI, Scaringe said Rivian is also using the technology to help it make more informed decisions around vehicle health and diagnostics in an effort to help improve the ownership experience as well as change consumer expectations. “It’s important to think about leveraging AI to have informed prognostics around the health of your suspension or battery and to be able to integrate that, say, with an owner’s calendar to schedule service,” he said. “You’ll be frustrated if a service technician calls and says, ‘I’d like to have your car in for service this week,’ and you’re, like, ‘I’m in Europe. How do you not know that?’”

He added that the ability to easily add new features to a car via over-the-air (OTA) software updates long after it rolls off the assembly line has become commonplace for EVs especially, and a benchmark for automakers. And as with previous tectonic technological shifts such as high-speed internet and ubiquitous smartphones, Scaringe believes that through the use of AI and the ever-expanding capabilities of AIDVs, it will transform the car ownership experience. “All of these other services are going to start to know all these rich contextual things about your life because of AI,” Scaringe said. “What I think is going to happen is over the next five years consumers are going to increasingly just expect certain things [from AIDVs]. From a societal point of view, it’s just going to become more the way things work.”

A Major Obstacle for Legacy Automakers

Scaringe said a major obstacle for legacy automakers when it comes to AIDVs—or even SDVs—has been their inability to quickly pivot away from a deep-rooted reliance on traditional distributed electrical and electronics (E/E) architectures that use hundreds of electronic control units, or ECUs to a zonal architecture pioneered by Tesla and leveraged by Rivian and other EV startups. The zonal approach uses only a few cross-domain ECUs, much less wiring, a powerful central computer, and allows for easier OTA software updates.

Scaringe pointed out the dependency on distributed E/E architectures and their accompanying unwieldy software can be traced back to the advent of electronic fuel injection. “Prior to the 1960s, cars were analog, and they had no semiconductors,” he added. Bosch’s D-Jetronic electronic fuel injection system introduced in 1968 included the first semiconductor and software in production vehicles and started a larger reliance on suppliers and software.

This became part of a growing pattern that, as automakers added new features and systems, each required an ECU. “An air conditioning system that had some level of software and microcontroller built into it would have its own ECU,” Scaringe said. “Power windows that had intelligence would have their own ECU.” Before long, automakers added hundreds of ECUs throughout the vehicle, all developed by different suppliers, all with different software. “I think of it like a field of weeds,” Scaringe said.

Making this even more difficult is that tier-one suppliers don’t always create software for a component or system and outsource it to third-party developers. “You have the complexity of all these different suppliers with their sub-suppliers doing the software that all has to come together,” Scaringe said. “And if you want to do an over-the-air update, you have to coordinate amongst lots of different companies, many of which are a layer or two removed from the software itself. The only way to really do it right is to scrape away those roughly 100 ECUs, replace that with a small number of computers that do a lot of things across a lot of functions, own that entire software stack, and build all the code to run all those things,” he said. “But it’s very, very hard to do.”

Newer automakers like Rivian have adopted zonal and centralized E/E architectures more rapidly than legacy automakers have in part because they've had a blank slate to start with and much smaller portfolio of vehicle models, Basich said. “For Rivian, they didn’t have legacy vehicle platforms and operations to maintain, and as a result it has taken several years for legacy OEMs to turn the ship around,” he added.

But legacy OEMs have zonal E/E architectures in the production pipeline. “For example, GM plans to launch its next-generation E/E architecture in the Cadillac Escalade IQ in 2028,” Basich said. And legacy and new automakers face different challenges, he added. “For new OEMs, their challenge has been scaling up production, whereas for legacy OEMs, their challenge has been deploying advanced E/E architectures across large product portfolios.”

Buy Rather Than Build

One path for legacy car companies is choosing to partner with newer automakers to buy rather than build a zonal E/E architectures for SDVs and AIDVs. Stellantis recently announced partnerships with Applied Intuition and Qualcomm to speed up the automakers’ tech efforts, and in late 2025 Volkswagen formed a $5.8 billion joint venture with Rivian for the same purpose.

“It’s essentially taking our zonal ECUs and associated base-level OS and deploying that across a variety of form factors, price points, and brands within the Volkswagen portfolio,” Scaringe said. “We’ve been pretty open on saying we think of this as the first of what we hope are many deals like that,” he added. “That some of the traditional OEMs are behind and we have all the ingredients I’m sure will gain a ton of traction not only in the U.S. but also in the European market.”

Scaringe said legacy automakers will be faced with three choices as the industry moves from software- to AI-defined vehicles. The first is to do nothing and “risk shrinking because customers are going to want vehicles that have software-defined architectures and deeply integrated AI capabilities.” The second is to build it themselves, and the third is to partner with companies like Rivian.

EVs and new E/E and software architectures have by coincidence developed simultaneously, and as legacy automakers have pulled back on EV programs, some have doubled down on their legacy E/E architectures, Scaringe added. “That saves a lot on R&D, but it’s a very risky bet because the 2030s are going to come around, and the basic premise of those architectures was unintentionally formulated in the 1960s,” he added.

“It’s inconceivable that a car company can grow or maintain market share without a domain-based architecture and a software-defined vehicle without very high levels of autonomy,” Scaringe said. “We think those two things are going to increasingly become important for customers.”