technologyScore 60/100Watch
Rivian developing in-house RAP1 autonomy processor (800 TOPS per chip, 2 chips per vehicle = 4x NVIDIA baseline)—moving away from supplier dependency for Level 4 inference
RJ Scaringe· Rivian· Robotics· 2026-05-01· about Rivian, NVIDIA (RIVN, NVDA)
“The Gen 3 autonomy hardware suite includes our in-house RAP1 platform—our in-house inference platform at 800 TOPS per chip. We have two of those chips in the vehicle, so it is extremely powerful—a roughly 4x increase relative to the NVIDIA-based platform—and the inclusion of LiDAR, along with enhancements across the rest of the perception stack.”
Why it matters
Rivian's custom silicon development signals OEM intent to reduce reliance on NVIDIA for edge inference. The 800 TOPS design suggests potential custom architecture optimized for driving models, implying reduced demand for high-margin NVIDIA automotive chips in robotaxi/Level 4 applications.
Investment implication
NVIDIA's automotive inference segment faces potential headwinds from vertical integration trends. However, Rivian's reliance on advanced node process manufacturing (likely 5nm or below for 800 TOPS density) increases TSMC/Samsung foundry demand. LiDAR sensor suppliers and perception stack integrators (cameras, radar) remain critical.