Qualcomm Enters the ADAS and AV Space
|
NEWS
|
In November 2021, Qualcomm announced that BMW will leverage their Snapdragon Ride platform to power future Advanced Driver-Assistance Systems (ADAS) and Autonomous Vehicle (AV) applications on their Neue Klass EV models, which are set to be brought to market in 2025. The announcement sees BMW expand their number of end-to-end autonomous vehicle platform suppliers, with their longstanding relationship with Intel Mobileye remaining intact. From Qualcomm’s perspective, the BMW deal signals the first success of their strategy to expand into the ADAS and AV domains.
Having already carved out a significant market presence in the infotainment and digital cockpit domains, Qualcomm is now looking to secure a foothold in the ADAS and AV domains, adding a new competitor to a space thus far dominated by Intel-Mobileye and NVIDIA.
The Era of Scalable, End-toEnd Autonomous Vehicle Platforms
|
IMPACT
|
Specific details regarding the Qualcomm-BMW deal were scarce in the November announcement. However, the scalable ADAS and AV solution will combine three elements—a machine vision Ride System-on-a-Chip (SoC), a high-compute centralized ADAS processor to host BMW’s driving policy software, and Qualcomm’s suite of Car 2 Cloud services to remotely manage the Ride SoC and ADAS processor. Although the announcement did not specify any upper limit to the level of automation that BMW will deploy in conjunction with Qualcomm, the Ride platform is capable of scaling from SAE Level 1 windshield-mount ADAS at ten TOPs, to 700+ TOPs configuration intended to enable Level 4 automation.
This further reflects the trend towards Original Equipment Manufacturers (OEMs) opting for semiconductor suppliers capable of supplying platforms that are both “end-to-end” and scalable.
End-to-end systems add AV software elements to the compute platform, either at the application level, through various modules that can be leveraged entirely, or in combination with third party/OEM in-house software. Qualcomm developed Snapdragon Ride to serve an ecosystem of third-party software stacks and will in future benefit from the incorporation of Veoneer’s Arriver software expertise. NVIDIA, having entered the autonomous driving space by positioning the parallel processing capabilities of their graphics processing units (GPUs) to enable embedded AI inferencing, has greatly expanded the range of software products and tools over time, including the DRIVE AV software modules and DriveSIM prototyping and validation tools. The vast majority of automotive announcements at the GPU Technology Conference (GTC) 2021 concerned software innovations, rather than hardware.
Scalability enables OEM customers to tap a single supplier to fulfill multiple ADAS/AV configurations across multiple models. For example, Intel Mobileye’s SuperVision L2+ systems leverage two EyeQ5 SoCs, while their fully driverless Mobileye Drive System features eight EyeQ5 SoCs. Similarly, the consistent GPU architecture of NVIDIA Drive SoCs means that current customers of the Hyperion 8 self-driving system will be able to upgrade the Orin platform to the next generation NVIDIA DRIVE Atlan.
Competitive Dynamics in the AV Space
|
RECOMMENDATIONS
|
Automakers rarely accept a single point of failure in their supply chains—an instinct that is only likely to be reinforced by the recent semiconductor shortage. As such, it was unlikely that any single supplier of AV platforms could become sole supplier to the entire industry. Nevertheless, BMW’s addition of a second AV supplier alongside their existing partnership with Intel Mobileye brings an end to years of consolidation and rationalization in the AV supplier space. Through their acquisition of Veoneer, and in the integration of the Arriver business unit, Qualcomm will gain deeper access to the ADAS and AV domain through Veoneer’s existing customer base, which includes Mercedes-Benz and Volvo, both of which currently have selected NVIDIA’s DRIVE Orin SoC as the basis of their future autonomous vehicle platforms.
Ultimately, automakers must find the right balance between scalable, end-to-end solutions which can fulfill multiple ADAS/AV configurations, and ensure that their ADAS / AV roadmaps are not dependent on any one supplier.
