Qualcomm CEO Cristiano Amon says AI companies are 'starting to do phones' and are asking his company to support running 100-billion-parameter local large language models continuously by 2028. Amon made the comments in an interview with Alex Heath of Sources, where he also discussed China versus US AI sentiment, how AI is changing smartphones and wearables, Qualcomm's data center strategy, and what 6G will unlock.
What AI companies are asking for
Local LLMs at that parameter count are a different beast from the small models that run on phones today. Running one continuously means the model never sleeps — it's always on, always ready, always drawing power. That's a hardware problem as much as a software one, and it explains why AI companies are knocking on Qualcomm's door instead of just buying cloud credits.
📊 Market Data Snapshot
Amon didn't name which AI companies are making the request, or which phone makers might ship the first devices built around that spec. What's clear is that the demand is real enough for Qualcomm's CEO to put a date on it: 2028.
Why 2028 keeps coming up
The timing lines up with two other things Qualcomm is working on: the next generation of AI chips and 6G. Amon mentioned 6G in the same interview, and that's not a coincidence. Continuous on-device inference at 100B parameters needs bandwidth and latency that current networks struggle to deliver reliably. 6G is supposed to fix that.
So the phone in this scenario isn't just a phone. It's an edge node with a persistent AI workload, connected to a network fast enough to keep it fed with fresh context. That's a meaningful shift from the way smartphones work now, where most heavy AI processing gets offloaded to the cloud.
The crypto angle nobody's pricing
Decentralized compute and storage networks have spent years arguing that AI inference is moving from cloud to edge. Qualcomm's roadmap is a validation point for that thesis, even if the market isn't treating it like one yet.
Projects that sell distributed GPU power or decentralized storage could find a new demand stream if on-device LLMs need a data layer that doesn't live on a single company's servers. That's the pitch, anyway. Whether it translates into actual usage is a separate question, and it's one the AI-token sector hasn't answered convincingly so far.
For now, crypto traders seem to be ignoring the headline entirely. Bitcoin is grinding sideways on low volume, dominance is high, and altcoins are feeling the squeeze. The Qualcomm news is a multi-year story, not a today story.
What to watch
The concrete milestone is 2028. Between now and then, the questions worth tracking are which AI companies commit publicly to on-device 100B-parameter models, whether Qualcomm names hardware partners, and whether any of them say the word 'decentralized' out loud.
Amon has put a stake in the ground. The rest of the industry has two years to either meet it or quietly walk it back.
