Qualcomm Unveils Two AI-Optimized Smartphone Chips

Qualcomm announced that its latest flagship processor can execute a 30‑billion‑parameter mixture‑of‑expert (MoE) artificial‑intelligence model directly on the device. The new chip, part of Qualcomm’s Snapdragon 8 Gen 5 series, incorporates an expanded tensor processing unit and a dedicated MoE engine that allows the processor to switch between specialized sub‑models on demand, reducing the need for cloud‑based inference.

The announcement highlights the company’s push to bring high‑capacity AI workloads to mobile and edge devices. By running a 30‑B MoE model locally, the chip can deliver faster response times, lower latency, and improved privacy compared to off‑device solutions. Qualcomm also noted that the architecture maintains competitive power efficiency, targeting a 30 % reduction in energy consumption for AI tasks relative to its previous generation.

Industry analysts view the development as a significant step toward mainstreaming large‑scale AI on smartphones and other edge devices. With the ability to handle complex models offline, the new processor could enable more sophisticated on‑device applications, from natural‑language processing to real‑time vision, while mitigating bandwidth and security concerns associated with cloud inference.