OEMs accelerate automated feature development with new Amazon EC2 DL2q instances, powered by the Qualcomm Cloud AI 100

At IAA MOBILITY in September 2023, Amazon Web Services (AWS) and Qualcomm Technologies, Inc. underscored the companies’ efforts to co-innovate purpose-built industry solutions with Qualcomm Technologies. We’ve now reached the first major milestone in the companies’ joint efforts with the general availability of new Amazon Elastic Compute Cloud (Amazon EC2) DL2q instances. Powered by Qualcomm® Cloud AI 100 for cloud AI inferencing. The Amazon EC2 DL2q instances serve as the first instance types to bring Qualcomm Technologies’ AI solution to the cloud on AWS.

Amazon EC2 DL2q instances are used by OEM engineers to develop then validate performance and accuracy of deep learning (DL) workloads for Qualcomm devices like AD/ADAS (Autonomous Driving/ Advanced Driver Assistance Systems). DL2q instances feature the same AI technology used in smartphones, autonomous driving, personal computers, and extended reality headsets, so they are ideal for development and validation of these AI workloads before deployment by OEM engineers. They can also be used to more cost-efficiently deploy DL workloads in the cloud.

Vehicle to Cloud Hardware Parity

To help the automotive industry transition into the software defined vehicle (SDV) era and achieve a higher degree of software development and deployment efficiency, AWS and Qualcomm Technologies are enabling maximum parity between the car and the cloud environment. Automakers will reduce time to market while bringing new features in the car to enable better user experiences.

As automakers migrate towards consolidated domain electronic control unit (ECU) architectures, they are adopting higher level automated driving features within new vehicle platforms. Developing highly automated driving features is a data-driven, compute-intensive process, requiring hundreds of petabytes of data to train advanced machine learning models. Within the development process, automakers collect data using vehicle test fleets, then transfer the data into on-premise data center infrastructures where it can be pre-processed, labeled, and stored. Developer teams and data scientists develop the code for the new automated driving models and features, then use machine learning algorithms to train the models using the millions of miles of data from the vehicle test fleets. This process is time consuming and expensive, running on specialized AI accelerators that automakers must purchase for are used within their on-premises infrastructure. Qualcomm’s Cloud AI 100 technology on AWS changes that, as hardware parity with Qualcomm chips in the cloud offers simulations closer to those in the real world.

Automotive Use Cases and Adoption at BMW

With up to eight Qualcomm AI 100 accelerators and 128 GB of total accelerator memory, customers are using DL2q instances to run popular generative AI applications, such as content generation, text summarization, and virtual assistants, as well as classic AI applications for natural language processing and computer vision. Thanks to the Qualcomm Cloud AI 100, DL2q will feature the Qualcomm AI stack portfolio, enabling customers ‘build once, deploy anywhere’ approach. To get started quickly, customers are using the AWS Deep Learning AMI (DLAMI), which comes prepackaged with Qualcomm’s Software Development Kits (SDK) and popular machine learning frameworks, such as PyTorch and TensorFlow. With the ability to alternatively store and use both physical and simulated data on AWS for feature development, customers are dramatically reduce costs and streamline their development time from weeks to hours. By migrating these tasks to the cloud, developers can also use over 200 AWS services to further scale and accelerate their overall feature development process.

“Convincing incumbent industries to change their standard operating practices is a herculean effort; however, the benefits of shifting at-scale software development to the cloud helps support and clearly demonstrates the future of development and design,” said Andrea Ketzer, Director of Technology Strategy for AWS Automotive. “We’re confident that by working with Qualcomm, we’ll advance software-defined development for customers across domains – from automotive to manufacturing and more. Combining our expertise creates an opportunity to bring new and exciting digital experiences to the world.”

As an industry pioneer, the BMW Group will be the first automaker to use the Amazon EC2 DL2q instances to help develop highly automated driving features within its next generation of vehicles, the “Neue Klasse,” set to launch in 2025. These joint efforts demonstrate how integrating cloud technologies into design and development of software-defined vehicles will help accelerate deployment timelines.

Benefits and Next Steps

“Working with AWS is empowering us to build on our established industry leadership in high-performance, low-power deep learning inference acceleration technology,” Rajat Sagar, Sr. Director, Product Management, Qualcomm Technologies, Inc. “Our work to date demonstrates the great potential of integrating cloud technologies into software development and deployment cycles. We look forward to continuing our work with AWS to help unlock new possibilities across all industries with solutions that will significantly reduce costs, while also enabling enhanced aptitude in deploying AI to respond to ever-evolving industry demands.”

For more information on how AWS helps support the enterprise cloud, please visit www.aws.com. To get started using the DLAMI and DL2q instances, visit the AWS Management Console, AWS Command Line Interface (CLI), and AWS SDKs.