To address the demand for localized AI applications, Qualcomm has unveiled a comprehensive suite of solutions, including localized AI device frameworks and inference toolkits. These offerings are designed to support enterprise deployments of localized AI applications and cater to the computational needs of AI inference across various verticals. At the same time, they prioritize data privacy, deployment flexibility, and cost reduction.
Qualcomm’s initiative responds to the prevalent reliance on cloud-based AI deployment or the high costs associated with procuring specialized hardware. By leveraging Qualcomm’s solution, organizations can achieve a balance between data privacy and computational efficiency while significantly reducing operational expenses.
The solution primarily utilizes Qualcomm Cloud AI Pro and Ultra-grade accelerator cards, capable of supporting AI applications with parameter scales ranging from 10 billion to an impressive 70 billion. Despite their robust capabilities, these devices weigh only around 10 kilograms, or slightly under 15 kilograms, making them highly adaptable for deployment across diverse computational environments.
This solution is well-suited for applications such as computer vision, generative AI services, and localized AI agent services. Additionally, it includes essential tools, resources, and databases needed for service development, enabling enterprises and developers to swiftly build and integrate AI applications that span from localized deployments to cloud-based operations.
Notably, early adopters of this solution include AIoT edge computing providers such as Aetina, Honeywell, and IBM. The system is also adaptable for a wide range of automated scenarios, including retail stores, fast-food restaurants, shopping malls, dealerships, hospitals, factories, and production facilities.
