Qualcomm's Snapdragon Summit Yields a Promising Evolution to the Digital Cockpit Through Strides in Compute Performance

At Snapdragon Summit 2024, Qualcomm unveiled new automotive compute platforms with the most powerful processors it has released to date, including the Oryon Central Processing Unit (CPU) (designed for 3X performance increases), Hexagon Neural Processing Unit (NPU) (designed for 12X performance increases), and Adreno Graphics Processing Unit (GPU) (designed for 3X performance increases). The new Snapdragon Elite tier of the Ride and Cockpit platforms was already been selected by Li Auto and Mercedes-Benz for future vehicle lines to power their digital cockpits and automated driving capabilities. This was announced along with a collaboration with Google that leverages Snapdragon Digital Chassis and Google Cloud and Android Automotive Operating System (AAOS), as well as the software built on top of it, to drive the development of Generative Artificial Intelligence (Gen AI)-enabled digital cockpits and, more broadly, a cloud framework for Software-Defined Vehicle (SDV) development. Both of these announcements in tandem hold strong potential for Original Equipment Manufacturers (OEMs) to move forward in their SDV transitions, but their stakeholders must consider several factors to make this potential a reality.

For example, the Cockpit Elite and Ride Elite are powered by the same System-on-Chip (SoC), capable of supporting 16 high-resolution displays with AI-enabled zonal audio and noise suppression, as well as 360° camera perception and fusion for up to 20 cameras with 16 Megapixels (MP) resolution. These cutting-edge specs will meet the demands of any OEM, but the use cases of such compute power need to be realistically established before integration into vehicles will yield any benefits. More clearly, the AI power of the platform will drive the integration of next-generation digital assistants that can provide occupants with proactive, contextually relevant, and personalized engagement and experiences.

The unification of separate domains in the vehicle, which traditionally would require an entire Electronic Control Unit (ECU) each, is a critical aspect of changing compute architectures in the automotive space, and this trend is a global movement toward fewer, more powerful pieces of hardware covered in ABI Research’s Accommodating Mixed-Criticality Compute in Digital Cockpit Domain Controllers report. Qualcomm’s Elite tier platforms deliver OEMs a compute solution that can consolidate multiple ECUs into domain controllers and, crucially, can be significantly upgraded over time due to the significant overhead in computing power that a vehicle will ship with at launch.

Qualcomm’s AI focus in these solutions is a sign of where the automotive industry is moving, and a strong complement to its high-powered Hexagon NPU is the Qualcomm AI Hub. This is a comprehensive Software Development Kit (SDK)/Machine Learning Operations (MLOps) toolkit to optimize software for its various hardware platforms, which can squeeze more efficiency out of the compute power via this more optimized software. This also means that bigger AI models can run on more constrained hardware, and that the barrier to entry for development of AI features is lowered, so more developers can get involved. For example, the AI Hub includes speech recognition models for digital assistants, or semantic segmentation neural network models for autonomous driving. Looking to the future, facial landmark detection models can be used for proactive driver fatigue recognition by intelligent AI copilots.

Despite the showcase of next-generation SoC platforms, which will certainly be enough to satisfy an OEM’s appetite for compute power, stakeholders in the industry must consider how features that they intend to bring to cars throughout their lifetime can generate revenue. Compute is no longer a barrier to the SDV, with the necessary processing power, AI acceleration, and energy efficiency all being up to par. Finding applications for AI and the emerging wealth of compute power should be a key priority—high-end gaming, for example, is an often-mentioned example of a feature that can be enabled by the innovations in automotive-grade silicon. However, the reality of monetizing it are more nuanced. Stakeholders at silicon vendors, Tier One suppliers, OEMs, and even in-car gaming companies have noted that occupants typically aren’t looking for intense and immersive gaming experiences in vehicles, and that rather more casual, short bursts of gaming sessions when the car is charging, or at a rest stop, are more popular. This makes high-end gaming a difficult feature to convert into revenue, although there are still many years before it will be implemented in vehicles, so there is time to gradually change consumer attitudes to more immersive forms of in-car gaming that will utilize the compute power and graphical capabilities offered by Qualcomm’s new platforms.

Even then, stakeholders must answer two questions:

1. Are there AI applications with a robust Return on Investment (ROI)? A recent ABI Insight, “Agentic AI is the Perfect Fit for Automotive” details one of these proposed applications.
2. Can OEMs consolidate organizationally at the same pace as their compute platforms? Teams across different domains are still struggling to collaborate efficiently, with ownership of different features often being misunderstood. The utilization of this compute power needs to be complemented by much more streamlined development, which is not possible in OEMs today.

One possible solution to the second question lies in Qualcomm’s collaboration with Google—allowing OEMs to lower development time for new features that could generate revenue through the Unified SDV Car to Cloud Framework. This uses Google Cloud-hosted automotive software development to make developers more productive and reduce time to market for the AAOS platform and services. Empowering its OEM customers in this way is critical to enabling the use of Qualcomm’s compute offering, and the rest of the supply chain should follow suit with new ways to help OEMs consolidate development across several domains, which traditionally were siloed in separate teams.