Qualcomm CEO: AI Companies Want 100B-Parameter Phones by 2028
Qualcomm CEO Cristiano Amon revealed that AI companies are pushing for smartphones capable of continuously running 100-billion-parameter models by 2028. This target represents a massive leap from current on-device limits, which typically hover around 30 billion parameters. This ambitious projection highlights the rapid shift toward always-on, agentic AI that operates locally on mobile hardware rather than relying entirely on cloud servers. Achieving this would drastically improve data privacy, reduce latency, and enable seamless offline AI assistants. The transition to running 100-billion-parameter models continuously on handheld devices requires monumental breakthroughs in mobile NPU architecture, memory bandwidth, and model quantization techniques. Qualcomm's leadership emphasizes that future edge AI will focus heavily on proactive, autonomous software agents that execute tasks on behalf of the user.
## BACKGROUND
Edge AI involves running machine learning models directly on localized hardware like smartphones, laptops, and IoT devices instead of remote cloud datacenters. Parameter count is a general indicator of an AI model's capacity and complexity, with larger models typically offering superior reasoning and language understanding at the cost of significantly higher computing and memory requirements.