Qualcomm CEO Cristiano Amon stated that AI companies are requesting smartphones capable of running 100-billion-parameter models continuously by 2028. These agentic models are expected to operate with personal context, enabling them to understand user intent and plan actions independently.
The current hardware capabilities, represented by the Snapdragon 8 Elite Gen 6 and Extreme Gen 6 platforms unveiled on September 22, 2026, support models of up to 30 billion parameters. These chips utilize 2nm technology and advanced memory management to handle AI workloads.
To bridge the gap to 100-billion-parameter models, Qualcomm is exploring techniques such as loading parameters from flash storage and utilizing low-power memory architectures. The company also employs always-on sensing hubs that process models of approximately 200 million parameters.
Amon emphasized that smartphones will remain central to personal AI due to the data context they hold. He noted that while AI capabilities could drive hardware demand, high memory prices and battery efficiency requirements remain significant engineering challenges for the industry.
Other companies are also targeting similar timelines, with reports suggesting that OpenAI is exploring hardware with comparable AI capabilities by 2028. However, no firm product commitments have been made, and the 2028 target remains a request from AI developers.
