The ecosystem for sensors used in autonomous vehicles is growing fast across hundreds of companies, as Nvidia demonstrated as part of its GTC 2022 event.
Nvidia revealed its current Nvidia Drive Hyperion 8 software works with radar, camera, ultrasonic and lidar sensors from these suppliers: Sony, for camera; Luminar for lidar, Hella for radar, Continental for radar and Valeo for camera and ultrasonics.
Hyperion 8 runs on the Nvidia Orin SoC, which is in production and appearing in multiple vehicle models. Nvidia also announced Hyperion 9 at GTC 2022 which will run on Atlan SoCs starting in 2026. Hyperion 9 will add even more sensors than with Hyperion 8, increasing to support 14 cameras, 9 radars, 3 lidars and 20 ultrasonics in the sensors suite. It is too early for Nvidia to comment on the qualified sensor suppliers for Hyperion 9, the company said.
Nvidia has been careful to note that its software is built for a wide variety of use cases, so a single vehicle maker might end up using less than entire suite of all supported sensors.
While Nvidia’s sensor support is fairly vast, it is not the only player in the AV space, having been compared by analysts to Mobileye and Qualcomm. In one example alone, Global Market Insight put the size of the global automotive lidar market at $3 billion by 2024, an indication of the numbers investors in AVs foresee even before a fully autonomous vehicle is developed, one in which there is no steering wheel or pedals inside.
So far, the AV and assisted driving space for Nvidia is just 2% of overall revenues, but has room to grow as the AV industry grows, said Sam Abuelsamid, principal analyst at Guidehouse Insights.
Most of Nvidia’s vehicle technology is also shared across other verticals at Nvidia, a benefit often overlooked. The GPU cores in the Orin SoC run the same Ampere architecture found in other Nvidia chips, Abuelsamid noted.
“Tapping the automotive market is also about more than just selling the SoCs for ADAS and AD, but also about selling simulation systems, data centers for training, licensing maps and more,” he said.
The same approach applies to Qualcomm where Snapdragon Ride chips are derived from Snapdragon chips used in mobile. While Mobileye will be spun into an IPO by Intel, Abuelsamid expects Intel and Mobileye to work together closely in numerous ways. One connection will be tying Intel data centers to Mobileye customers much as Nvidia is doing. Also, Intel and Mobileye will collaborate on design of imaging radar and lidar sensors, he noted.
RELATED: Jags and Land Rovers will get Nvidia automated driver tech in ‘25