NVIDIA DRIVE Thor is a centralized automotive SoC delivering up to 2,000 TOPS (FP4) on the Blackwell architecture, targeting L3/L4 autonomy by unifying ADAS, cockpit, infotainment, and parking on a single chip -- an 8x leap over DRIVE Orin's 254 TOPS. The $14B design-win pipeline (over 6 years, announced at GTC) is anchored by BYD, Geely, Mercedes-Benz, Volvo, Hyundai Motor Group, Toyota, and numerous Chinese EV OEMs. However, Thor's mass production was delayed from mid-2024 to mid-2025+ due to Blackwell architecture design flaws (Jensen Huang: 'It was 100% NVIDIA's fault'), causing key Chinese OEMs (XPeng, NIO) to accelerate in-house chip development.
The DRIVE Hyperion reference platform (featuring dual Thor SoCs) was adopted by BYD, Geely, Isuzu, and Nissan for L4 programs at GTC 2026. In autonomous trucking, the Aurora/Continental/NVIDIA partnership targets mass production of L4 trucks in 2027 using dual Thor SoCs. Mercedes-Benz CLA became the first production vehicle shipping NVIDIA's full AV stack (including Alpamayo reasoning AI) in Q1 2026. Despite the strong design-win pipeline, actual FY2026 automotive revenue was only $2.3B (+39% YoY) -- massively missing NVIDIA's earlier $5B target -- with growth decelerating from 72% in Q1 to 6% in Q4 as DRIVE Orin matured before Thor production fully ramped..
Automotive is a long-duration option
NVIDIA's automotive business remains small relative to data center but carries a large design-win pipeline. The key question is conversion speed, not market opportunity.
What is the actual ASP of a DRIVE Thor SoC vs DRIVE Orin? The 8x compute leap should command a significant price premium but no public pricing exists.