At the Snapdragon Summit 2026, Qualcomm unveiled a bold strategic pivot for the era of AI agents, introducing two new flagship mobile SoCs: the sixth-generation Snapdragon 8 Elite Super and Snapdragon 8 Elite. Built on TSMC’s cutting-edge 2nm process technology, these chips feature Qualcomm’s self-developed next-gen Oryon CPU—marking a milestone by pushing mobile CPU clock speeds to 5GHz for the first time. Enhanced performance is achieved through offset stacking packaging and Oryon FlexCache dynamic caching technology, while meaningful upgrades have also been made to the GPU, Hexagon NPU, low-power Sensing Hub, ISP, and connectivity capabilities.

These chips are tailored to meet the operational demands of Transformer architectures, large-scale Mixture-of-Experts (MoE) models, and AI agents. Notably, the Snapdragon 8 Elite Super can run MoE models with up to 30 billion parameters directly on mobile devices—a leap forward in on-device AI processing power.

Qualcomm envisions a future where personal computing evolves from being smartphone-centric to AI agent-centric. This shift will require on-device agents to possess capabilities like continuous perception, autonomous planning, cross-modal processing, and seamless cross-device collaboration. To support this vision, Qualcomm is accelerating the adoption of high-bandwidth computing architectures. Additionally, the company will open-source the Mojo programming language and the Max unified AI inference service engine from Modular—a firm it acquired earlier—to foster collaboration across on-device, cloud, and cross-device computing ecosystems, thereby addressing the emerging computational needs of the AI agent era.