Qualcomm Unveils On-Device AI Roadmap at MWC Shanghai 2025
At the Mobile World Congress (MWC) Shanghai 2025, Qualcomm outlined its strategic shift toward on-device artificial intelligence, predicting that future AI interactions will predominantly occur on edge devices like smartphones, PCs, and cars. Vinesh Sukumar, Qualcomm’s VP of AI product management, described AI as the new user interface, evolving from generative models to autonomous intelligent agents. The company highlighted the benefits of edge-based AI, including improved speed, energy efficiency, and privacy. To support this transition, Qualcomm has developed a software stack optimized for compact models with 7 to 10 billion parameters, leveraging its Hexagon NPU for efficient processing. The strategy employs a layered architecture that offloads complex tasks to edge servers or the cloud only when necessary, ensuring low latency. Qualcomm reported enabling 200 to 300 AI applications annually and plans to scale developer tools and consumer features in the coming months. This move positions Qualcomm as a key infrastructure provider in the next generation of AI-driven user experiences, addressing hardware limitations and privacy concerns associated with cloud computing.
Wire timeline
Qualcomm Unveils On-Device AI Roadmap at MWC Shanghai 2025
At the Mobile World Congress (MWC) Shanghai 2025, Qualcomm outlined its strategic shift toward on-device artificial intelligence, predicting that future AI interactions will predominantly occur on edge devices like smartphones, PCs, and cars. Vinesh Sukumar, Qualcomm’s VP of AI product management, described AI as the new user interface, evolving from generative models to autonomous intelligent agents. The company highlighted the benefits of edge-based AI, including improved speed, energy efficiency, and privacy. To support this transition, Qualcomm has developed a software stack optimized for compact models with 7 to 10 billion parameters, leveraging its Hexagon NPU for efficient processing. The strategy employs a layered architecture that offloads complex tasks to edge servers or the cloud only when necessary, ensuring low latency. Qualcomm reported enabling 200 to 300 AI applications annually and plans to scale developer tools and consumer features in the coming months. This move positions Qualcomm as a key infrastructure provider in the next generation of AI-driven user experiences, addressing hardware limitations and privacy concerns associated with cloud computing.
TechNode