What's Happening?
Qualcomm CEO Cristiano Amon has stated that AI companies are pushing for smartphones capable of continuously running 100-billion-parameter AI models by 2028. This represents a significant leap from current flagship chips, which can handle models around
30 billion parameters. Amon indicated that these AI labs desire 'agentic AI'—software that operates proactively on a user's behalf, requiring constant background operation rather than on-demand activation. This ambitious target comes amidst an existing memory supply crunch, which Amon acknowledges is already straining the industry. The demand for such advanced on-device AI necessitates new memory architectures and efficient chip designs to manage increased memory bandwidth and battery consumption without compromising device performance or user experience. Qualcomm's latest Snapdragon 8 Elite Gen 6 and Snapdragon 8 Elite Extreme Gen 6 mobile platforms, built on a 2nm process, are designed to run models above 30 billion parameters, but the 100-billion-parameter goal is more than triple that capacity.
Why It's Important?
This development signifies a major shift in the mobile technology landscape, pushing the boundaries of on-device artificial intelligence. The move towards 'agentic AI' could fundamentally redefine smartphone functionality, transforming devices from app-centric tools to proactive personal assistants. This will intensify competition among chipmakers like Qualcomm, Apple, and MediaTek for scarce memory investment and advanced manufacturing capacity, as they also contend with demand from data-center AI servers. The ability to run such large AI models locally on a smartphone could enhance privacy, reduce reliance on cloud infrastructure, and enable more personalized and responsive user experiences. However, it also presents significant engineering challenges related to power efficiency, thermal management, and memory architecture. The success of this endeavor will determine which companies lead the next generation of mobile computing and define the future of smartphone interaction.
What's Next?
The industry is expected to focus on developing more memory-efficient chip designs and advanced manufacturing processes to meet the 2028 target for 100-billion-parameter AI models on smartphones. Qualcomm and other mobile chipmakers will likely invest heavily in research and development to overcome the technical hurdles of power consumption, thermal management, and memory bandwidth. There will be increased competition for limited resources, including memory suppliers and advanced-node capacity, as both mobile and data-center AI demands escalate. While Qualcomm has not specified which companies are driving this demand, the push from AI labs suggests that major players are actively seeking these capabilities. The coming years will likely see prototypes and further announcements regarding the technological advancements required to achieve this ambitious goal, potentially leading to new product categories beyond traditional smartphones, such as AI-agent phones and always-on wearables.
Beyond the Headlines
The pursuit of 100-billion-parameter AI models on smartphones has profound implications beyond mere technological advancement. Ethically, the rise of 'agentic AI' raises questions about user autonomy, data privacy, and the potential for AI systems to make decisions on behalf of individuals. The continuous operation of such powerful AI on personal devices could lead to new forms of digital dependency and influence. Legally, the development and deployment of these advanced AI systems will necessitate new regulations concerning data handling, algorithmic transparency, and accountability for AI actions. Culturally, this shift could redefine human-technology interaction, moving towards a more seamless and integrated experience where AI anticipates and fulfills needs without explicit user prompts. The 'chip war' for agentic AI on-device highlights a broader strategic battle for control over the future of personal computing and the underlying infrastructure that will power it, potentially leading to a re-evaluation of what constitutes a 'phone' by the end of the decade.
