The Rivian R2 is here, the reviews are good, and the conversation among reservation holders has narrowed to a single component: the lidar sensor expected on a later Gen 3 version of the car. Travis Ketchum thinks that focus is aimed at the wrong thing. In his video he argues that the sensor buyers are waiting for matters less to them than the autonomy computer they will never see, and that Rivian's own messaging about when either arrives has been confusing enough to leave people holding $100 reservations in the dark. He is not neutral about this. He bought a Tesla Model Y with hardware 3 in 2022, watched hardware 4 land less than a year later, and says he has spent the years since on the wrong side of an autonomy promise.
Lidar sends out laser pulses and times the return, producing a depth map that camera systems can be checked against. That distinction is the whole argument in driver assistance right now, with Tesla committed to cameras alone and lidar mounted on the Waymo vehicles the video points to. The video adds a wrinkle worth sitting with: if lidar ships on every R2 rather than on a few dozen validation cars, the entire customer fleet becomes ground truth data collection. There is a buyer consideration underneath all of this that the video touches only lightly. Software arrives over the air, but compute does not. A more powerful autonomy computer cannot be subscribed to later. It is a physical part, chosen at order time, and for anyone who keeps a car past the warranty that decision is effectively permanent. Reservation holders are being asked to make a hardware bet on software that does not exist yet.
Ketchum reports that a launch R2 carries essentially the same autonomy computer as the Gen 2 R1S and R1T from mid 2024, built on what he describes as dual Nvidia Orin chips. He describes that hardware as roughly comparable on paper to what Tesla was running for its own advanced features, and says the gap in results comes from software and training data rather than silicon. The Gen 3 change, he says, is Rivian moving to a chip it designed in house, which he calls RAP1, with up to four times the performance. He notes Rivian itself does not expect a meaningful difference for a few years. On timing, he says the messaging has ranged from late 2026 to early 2027, and that some reservation holders received emails pushing their order invitations toward February 2027. His own read is that a validation fleet may exist by the end of 2026 but no customer takes delivery of a Gen 3 car until well into 2027, and that the launch package, which he says includes a subscription free Autonomy Plus license, will not include the newer hardware. On data, he reports Rivian has roughly 60,000 to 80,000 qualifying Gen 2 vehicles collecting training samples against the 1.6 to 1.8 million cars Tesla delivers annually, a gap the R2 and cheaper models are meant to close. He also points at the Volvo EX60, which he says uses a similar Nvidia solution with far more conservative driver assistance plans, as evidence that the software, not the chip, sets the ceiling.
Bottom line: If you flip cars every three years, order the launch R2 and enjoy it. Ketchum did exactly that and says so plainly. If you are the kind of owner who is still driving the thing in 2033, wait the six to nine months. The sensor is the part people argue about because it is visible and has a name. The computer is the part that decides whether your car quietly improves for a decade or stops getting the good updates in year four. Rivian has earned some patience here. It has not yet earned a deposit on a timeline it keeps rephrasing.
Commentary on a third-party video. Figures and claims are as presented in the source and have not been independently verified. Spotted an error? Tell us and we will correct it.