Why AI Firms Are Turning to Qualcomm for On-Device Intelligence
Qualcomm CEO Cristiano Amon revealed that artificial intelligence companies are now developing their own smartphones and turning to Qualcomm for assistance in running large AI models directly on devices. Speaking at a technology event in October 2026, Amon noted a shift where AI firms are moving beyond software to design hardware, particularly phones capable of handling advanced on-device processing. He emphasized that these companies are asking Qualcomm to support the execution of AI workloads reaching up to 100 billion parameters, a scale previously reserved for data centers. The comments highlight growing collaboration between chipmakers and AI developers as edge computing becomes central to next-generation mobile experiences.
Breaking news
Town AI Secures $108 Million Series B, Valued at $1 Billion After $55 M Pre‑Series
Apple Prepares for Major Product Launch Event on October 28
Apple Releases Release Candidates for watchOS, tvOS and visionOS 27.2
Apple Limits iPhone Duo and iPhone 18 Pro Max to Reservation-Only Sales in Japan and South KoreaAmon explained that AI companies are seeking Qualcomm’s expertise in optimizing chips for power efficiency and performance when running massive language models locally. He stated that the demand is driven by the need for faster response times, better privacy, and reduced reliance on cloud connectivity. Qualcomm’s Snapdragon platforms, he said, are being adapted to handle inference tasks for models with up to 100 billion parameters through advancements in AI engines, memory bandwidth, and software stacks. This shift reflects a broader industry trend where AI functionality is no longer confined to servers but is being embedded directly into consumer devices like smartphones, enabling real-time translation, generative photo editing, and context-aware assistants without sending data to the cloud.
How Will Phones Handle 100B-Parameter AI Models?
Amon detailed that achieving this requires more than just raw processing power; it involves innovations in model compression, quantization, and heterogeneous computing across CPU, GPU, and NPU components. He noted that Qualcomm is working with partners to refine software tools that allow AI models to be efficiently deployed on mobile silicon without significant loss in accuracy. While current flagship phones typically handle models in the single-digit billions, the roadmap aims to scale up through iterative improvements in chip architecture and AI acceleration. Amon cautioned that full 100B-parameter execution may still involve hybrid approaches, with some layers processed on-device and others offloaded, but the goal is to maximize local computation for speed and security.
The integration of large-scale AI on phones could redefine user expectations, making devices more intuitive and proactive in anticipating needs. Amon suggested that future smartphones may act as personal AI hubs, managing tasks across health, productivity, and entertainment with minimal latency. He also indicated that Qualcomm sees this as a long-term opportunity to deepen its role in the AI ecosystem, moving beyond supplying components to co-developing solutions with AI innovators. As on-device AI matures, it could reduce pressure on data centers and enable new applications in areas like augmented reality, real-time language translation, and intelligent automation—all running securely within the user’s hand.
What Does This Mean for the Future of Smartphones?
Are AI companies actually building their own phones? Yes, according to Cristiano Amon, AI firms are beginning to design and release smartphones tailored to run their AI models efficiently, rather than relying solely on existing devices from traditional manufacturers.
Frequently Asked Questions
Can today’s smartphones run 100-billion-parameter AI models? Not fully yet; current devices handle smaller models, but Qualcomm is developing the hardware and software needed to support larger workloads through optimization and hybrid processing techniques.
Will on-device AI replace cloud-based AI entirely? No, Amon explained that a hybrid approach is likely, with sensitive or real-time tasks processed locally and heavier workloads still using the cloud, balancing performance, privacy, and practicality.
