Rivian's CEO and chief product officer RJ Scaringe spent most of 2021 launching three vehicles at once, into a supply chain that was fraying in real time. The R1T pickup, the R1S SUV, and a commercial delivery van all hit production within six months of each other during the post-COVID parts shortage. Looking back, he calls it the one thing he would have done differently. The lesson shaped everything that came after. Now the company is ramping its first mass-market vehicle, the R2, priced from $45,000 at a moment when the average new car in the United States costs around $50,000. Scaringe spoke at ACT Expo about the road from startup to 17,000 employees, and where he thinks transportation is headed over the next decade.
Rivian's strategy has always leaned heavily on building in-house: software, compute hardware, battery packs, motors, and power electronics. That approach requires a large engineering organization, which in turn requires the revenue volume to justify it. R2 is the vehicle that is supposed to close that gap. At the same time, Rivian has begun licensing its electronics and software platform to Volkswagen Group in a deal worth $5.8 billion, with deployment beginning in 2027 across Volkswagen's EV lineup. Scaringe's framing at ACT is that the two revenue streams, vehicles and software licensing, are built to reinforce each other. The Volkswagen deal also validates the in-house approach: one of the world's largest automakers decided it was easier to pay than to build from scratch.
Scaringe's explanation of "AI-defined vehicles" starts not with AI but with architecture. He traces the history from fully analog cars through decades of ECU sprawl, arguing that most vehicles on the road today still run hundreds of isolated computers with software written by dozens of separate suppliers, a structure that was never designed, it just accumulated. Rivian's zonal compute approach collapses that into a smaller number of high-compute nodes, which is what makes continuous updates and on-board AI integration practical rather than theoretical. Rivian built its own 800-TOPS inference chip and puts two of them in each vehicle, for 1,600 TOPS of onboard compute total. Foundation models trained on fleet data inform real-time decisions across the vehicle. On the commercial side, Amazon's growing fleet of electric vans, now past 30,000 units deployed, runs a dedicated compute partition where Amazon's own applications sit alongside the vehicle OS, including driver monitoring, route mapping, and safety systems. Scaringe also mentioned starting a separate industrial robotics company a few months ago, with Rivian as the first customer.
Bottom line: Scaringe is doing what few automotive CEOs will do in public: treating the vehicle as a platform and describing the software licensing business as the thing that makes the vehicle economics work at scale. Whether R2 hits the volumes that math requires is still an open question, but the strategic logic is more coherent than most of what you hear from legacy automakers right now.