Qualcomm CEO Cristiano Amon says AI companies are starting to make phones and want Qualcomm to enable continuous 100B‑parameter local LLMs by 2028
| Source: Techmeme | Original article
Qualcomm CEO Cristiano Amon says AI firms are moving into phones and are requesting Qualcomm enable continuous operation of 100 billion‑parameter local language models by 2028.
Qualcomm’s chief executive Cristiano Amon told reporters that AI firms are “starting to do phones” and are already pressing the chipmaker to make it possible to run 100‑billion‑parameter language models on smartphones continuously by 2028. The demand, Amon said, reflects a shift from cloud‑only inference toward on‑device generative AI that can operate without a constant internet connection.
The request matters because a 100‑billion‑parameter model would dwarf today’s on‑device offerings, which are typically limited to a few hundred million parameters. Sustaining such a workload would require breakthroughs in power efficiency, thermal management and silicon architecture. If Qualcomm can deliver a solution, it would give handset makers a new lever for differentiation, reduce latency, and address privacy concerns tied to sending user data to remote servers. It also signals that AI vendors see mobile hardware as a viable platform for the next generation of generative tools, a trend echoed in recent coverage of the challenges of local AI adoption.
What to watch next is whether Qualcomm’s roadmap includes a dedicated AI accelerator capable of the required compute‑per‑watt, and how quickly silicon partners – from OEMs to chipset designers – can integrate it into upcoming flagships. Industry observers will also track regulatory sentiment, especially the contrasting AI policies in the United States and China that Amon highlighted, to see how geopolitical dynamics shape the rollout of on‑device large language models. The next few product cycles will reveal whether the 2028 target is an aspirational benchmark or a realistic milestone for mobile AI.
Sources
Back to AIPULSEN