Qualcomm announced an expanded strategic partnership with Hugging Face on June 24, aiming to onboard more than 3 million open AI models onto Qualcomm platforms including Snapdragon, Dragonwing, and Dragonfly. The deal also includes development of a Hugging Face Agent designed for hybrid orchestration of agentic AI workloads, essentially a tool that lets developers deploy models without manual integration headaches.

The expanded collaboration targets Hugging Face’s community of 16 million developers, giving them pathways to run open models on everything from edge devices to full-scale data centers. The Hugging Face Agent is built around zero-manual-integration deployment, meaning developers won’t need to hand-optimize their models for each Qualcomm platform. The agent handles the orchestration of agentic AI workloads across hybrid environments.

Qualcomm President and CEO Cristiano Amon framed the initiative as a significant milestone in democratizing advanced AI, making it more open and scalable across the full spectrum of computing.

Qualcomm and Hugging Face first partnered back in January 2022, when the collaboration focused on optimizing transformer models for Qualcomm’s Cloud AI 100 systems. In 2024, Qualcomm launched its AI Hub, which integrated with Hugging Face to support on-device model deployment. The new expansion scales that work horizontally across Qualcomm’s entire product portfolio.

On the same day as the Hugging Face announcement, Qualcomm revealed its acquisition of Modular, a company focused on AI software infrastructure from edge to cloud.

Qualcomm’s play here is hybrid inference: the idea that not every AI workload needs to go to a massive cloud data center. Some should run locally on a phone’s Snapdragon chip, some at the network edge, and some in the cloud. The Hugging Face partnership gives Qualcomm a funnel: if even a fraction of those 3 million models and 16 million developers start optimizing for Qualcomm hardware, the downstream effects on chip demand could be meaningful.