Qualcomm Unveils AI‑Focused Smartphone Processors, Flagship Chip Handles 30‑B Parameter Model On‑Device

Qualcomm revealed on Wednesday that it is shipping two new system‑on‑chip (SoC) families for premium smartphones, placing artificial‑intelligence performance at the core of the designs. The top‑tier processor is marketed as capable of running a 30‑billion‑parameter mixture‑of‑experts (MoE) model entirely on the device, eliminating the need for cloud‑based inference for many next‑generation AI applications.

The on‑device capability is intended to cut latency, lower data‑transfer costs, and improve privacy by keeping user data local. Analysts note that running such a large model without offloading to a server could enable real‑time language translation, advanced image editing, and personalized assistants that respond instantly, even in areas with limited connectivity.

Mixture‑of‑experts architectures split a massive neural network into multiple specialized sub‑models, activating only the most relevant ones for a given input. By supporting a 30‑billion‑parameter MoE, Qualcomm’s new chip pushes the boundary of what has previously been possible on mobile silicon, where the largest on‑chip AI models were typically under a few billion parameters. The company attributes the leap to a combination of a newer manufacturing process, a redesigned tensor accelerator, and a more efficient memory hierarchy.

The announcement arrives as rivals such as Apple and MediaTek are also racing to embed larger AI workloads in their latest chips. Smartphone manufacturers that adopt Qualcomm’s platform could differentiate their devices with features that rely on heavy on‑device computation, from sophisticated camera pipelines to context‑aware virtual assistants. Industry observers expect that the new SoCs will first appear in flagship models slated for release later this year.

Looking ahead, Qualcomm says the launch is part of a broader strategy to create a software ecosystem that lets developers target the expanded on‑device AI envelope without extensive re‑engineering. If the performance claims hold up in real‑world testing, the move could accelerate a shift toward more autonomous, privacy‑first mobile experiences and set a new benchmark for future generations of mobile processors.