Qualcomm is setting ambitious long-term goals for on-device artificial intelligence. The company's CEO has hinted that future smartphones could possess the hardware required to run massive 100-billion-parameter large language models (LLMs) locally by 2028.
Executing models of this scale directly on mobile silicon would represent a significant leap for edge computing, cutting latency and reducing reliance on cloud infrastructure. However, the projection introduces significant hardware questions that currently lack concrete solutions.
Running such demanding models natively would necessitate vast amounts of ultra-fast memory. The CEO's outlook did not address the dual challenges of mobile DRAM supply shortages and the steep component costs associated with outfitting mass-market phones with enough RAM to support 100B parameter models.
For more details and full context regarding the announcement, check the original report on Wccftech.
