Key Takeaways:
Qualcomm is betting its chip-to-data-center portfolio can become the backbone for open-source AI, connecting Hugging Face's 16 million developers to a unified compute fabric spanning smartphones to server racks.
Qualcomm Technologies and Hugging Face expanded their partnership to let developers deploy open AI models across devices and data centers, uniting Qualcomm's chip portfolio with Hugging Face's 16 million-strong developer community. The collaboration, announced June 24, targets a new era of agentic AI and hybrid inference at scale.
"This engagement represents a major step forward in making advanced AI more open, scalable, and accessible," said Cristiano Amon, President and CEO of Qualcomm Incorporated. "By combining Qualcomm's leadership in high-performance, low-power computing with Hugging Face's vibrant developer community, we are enabling a new generation of AI applications that seamlessly span device and cloud."
The collaboration rests on three pillars: moving Hugging Face's storage and inference workloads onto Qualcomm's Dragonfly data center solutions, automating model onboarding across Qualcomm's Snapdragon, Dragonwing and Dragonfly platforms, and building a Hugging Face Agent for hybrid AI orchestration. Hugging Face hosts more than 3 million open models across every domain and modality. The agent will handle setup, optimization and deployment with zero manual integration work, according to the companies.
The deal positions Qualcomm to capture AI workloads at both ends of the compute spectrum — from a smartphone to a data center rack — challenging Nvidia's dominance in inference infrastructure. Nvidia controls an estimated 80 percent of the AI accelerator market, but Qualcomm's advantage lies in ubiquity: its Snapdragon chips power hundreds of millions of smartphones, while Dragonfly targets the data center. OpenAI and Broadcom this week unveiled their own custom inference chip, Jalapeño, as the industry seeks alternatives to Nvidia's GPUs.
"Increasingly the world is running on open and local models because they're more affordable than the big APIs and private by design," said Clément Delangue, Co-founder and CEO of Hugging Face. "Together with Qualcomm Technologies, using Modular software and tools, we're making it easy for our 16 million developers to run open models everywhere, from a device in your hand to a full rack in the data center, with agents that work across the compute continuum."
The first pillar connects Hugging Face's storage and inference services to Qualcomm Dragonfly data center products, creating a direct path from model experimentation to production deployment. The second pillar automates AI model onboarding across Qualcomm platforms — smartphones, PCs, wearables, industrial systems and automotive — using a single workflow. Hugging Face will also offer its PRO subscription to Qualcomm platform customers, providing premium storage and compute for building with open models.
The third pillar enables distributed agentic AI, where intelligent agents dynamically orchestrate models and workflows across on-device and cloud systems based on performance, cost, privacy and latency needs. Developers will access Modular's AI software components through the Hugging Face platform.
For Qualcomm, the partnership deepens its push beyond mobile chips into AI infrastructure — a market Gartner projects will reach $297 billion by 2027. Qualcomm shares trade at about 18 times forward earnings, a discount to Nvidia's 35 times, reflecting the market's skepticism about its data center ambitions. This deal could narrow that gap if Dragonfly gains traction with Hugging Face's developer base.
This article is for informational purposes only and does not constitute investment advice.
