Qualcomm's Amon Says AI Developers Seek 100-Billion-Parameter AI Models on Phones by 2028
Qualcomm CEO Cristiano Amon says AI companies want phones running 100-billion-parameter AI models continuously by 2028, versus today's 30B limit.
AI companies are asking Qualcomm for smartphones that can run 100-billion-parameter AI models continuously by 2028, according to CEO Cristiano Amon. In an interview conducted on October 8, Amon said the requests come from companies at the forefront of AI, though he declined to name them. The gap is wide: Qualcomm's newest flagship platform tops out at about 30 billion parameters today.
What the request asks for
Amon described a model that stays loaded and runs all the time, rather than starting for a single task and shutting down afterward. The companies also want these models to understand personal context, reason through problems, and take autonomous action. Amon positioned the smartphone as the main hub for AI agents, ahead of wearables.
Snapdragon 8 Elite Gen 6 can run a 30-billion-parameter mixture-of-experts model locally, and Qualcomm's sensing hubs handle models up to 200 million parameters. Google's Gemini Nano-2, a current mobile model, uses about 3.25 billion parameters. A 100-billion-parameter target is more than three times the Snapdragon ceiling and roughly 30 times the size of today's on-device model.
Why 100-billion-parameter AI models strain phone hardware
Qualcomm has not promised a chip or product date for the requirement. A model of that size is estimated to need about 50 GB of memory, while the newly announced chips list 24 GB. Amon has pointed to memory, chip architecture, and energy consumption as the main hurdles, and the target ties directly to battery life and heat limits.
Local execution keeps prompts and personal context on the handset and removes the round trip to a data center. The cost is hardware that must hold tens of gigabytes of model weights and sustain constant inference without draining the battery. Cloud serving avoids that load but moves the data off the device.
The Samsung Galaxy S27 and OnePlus 16 are named among phones built on the current Snapdragon 8 Elite Gen 6, which means the 30-billion-parameter ceiling applies to the devices buyers can expect first. Any move to 100-billion-parameter AI models would therefore require a generational jump in both memory capacity and sustained power efficiency.
Why this matters
If the request is met, the phone becomes the place where agents run continuously, which would shift value toward memory capacity and power efficiency in premium handsets. Buyers should watch for memory specifications and battery claims in upcoming flagship launches, since Qualcomm has not yet committed to a product timeline.
Related Articles
- Nvidia Nemotron 4 open-weights model reshapes AI race
- Qualcomm Unveils Snapdragon 6 Gen 5 and 4 Gen 5 to Boost Midrange AI Performance
- Qualcomm Targets Hyperscale Market with CPU Built for Agentic AI Workloads
✔Human Verified
Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.