Qualcomm CEO Flags Phone AI Ask, Memory Limits

Amon relays a 2028 request from AI companies

Cryptobriefing reported on October 10, 2026, that Qualcomm CEO Cristiano Amon said AI firms want phones running 100-billion-parameter models continuously by 2028. Amon made the comments in an interview on October 8, 2026, and described a shift toward always-on agentic AI, meaning software that acts on a user's behalf rather than waiting for a tap, Cryptobriefing said.

Startup Fortune reported on October 11, 2026, that Amon declined to name the AI companies pressing for the target. The outlet said those companies want a model kept active all day, not loaded for a single task and shut down.

TokenPost reported on October 10, 2026, that Amon described the 2028 timing as a request from AI companies, not a Qualcomm product launch date or delivery commitment. Amon said, "We now have to design chips to allow the phone to be working, even when you're not operating it, and with an all day battery life and a lot of AI at the edge," TokenPost reported.

Qualcomm's newest chips are built for 30 billion

Qualcomm unveiled the Snapdragon 8 Elite Gen 6 and Extreme Gen 6 platforms at its Snapdragon Summit on September 22, 2026, Cryptobriefing reported. The chips use 2nm technology and are designed to run mixture-of-experts models of up to 30 billion parameters directly on the device, the outlet said.

Startup Fortune reported that the Extreme version's Hexagon NPU, with a new Element Accelerator and a larger shared memory pool, supports a context window of roughly 32,000 tokens.

Startup Fortune said the 2028 target is more than three times the on-device scale Qualcomm cites for its newest platforms.

TokenPost said Qualcomm has not announced a smartphone chip that can sustain the larger models continuously. TokenPost added that Qualcomm is promoting both new platforms for on-device agentic AI, which the company defines as software that understands intent, anticipates needs and takes actions for users.

Qualcomm leans on flash storage and memory changes

Cryptobriefing reported that Qualcomm loads model parameters from flash storage so the full model need not sit in active memory at once.

The new platforms also expand the memory available to the Neural Processing Unit, and Qualcomm pointed to intelligent flash management and a low-power memory architecture as other supports for larger models, Cryptobriefing reported. Always-on sensing hubs run much smaller models of approximately 200 million parameters continuously in the background, the outlet said. Qualcomm said its new chips can process approximately 1 million tokens per day, Cryptobriefing reported.

TokenPost said parameter counts offer a rough indication of model size, and that phones generally need more memory, processing power and energy as that number grows. Startup Fortune said keeping a model active continuously at triple the parameter count multiplies memory bandwidth and battery demands in ways current flagship hardware was not built for.

Memory supply and device reach stay in view

Startup Fortune reported that Amon has acknowledged the memory chip supply constraint for smartphones is temporary. Qualcomm's two platforms are slated for flagship handsets from Honor, iQOO, Motorola, OnePlus, OPPO, Redmi, RedMagic, vivo and Xiaomi, the outlet said. Startup Fortune also said the Snapdragon 8 Elite Gen 6 powers devices such as the Samsung Galaxy S27 and OnePlus 16.

Cryptobriefing reported that Amon argued the smartphone will remain the central hub for personal AI, because it holds the context about a person's life that makes an assistant useful. Amon said some consumers prefer to keep AI and their data on the device and under their control, TokenPost reported.

Amon tied the AI push directly to hardware demand, Cryptobriefing said. The story is still developing, and the interview comments are the only public statement of the 2028 request in the coverage reviewed.

Tron's take

This is my reading of the news, not a reported result. The headline number is a wish list from unnamed AI companies, relayed by one executive in one interview. TokenPost reported that no chip at that scale has been announced, and Startup Fortune reported that memory supply is constrained for now.

For a small or mid-sized business, I read that as nothing to buy or change this quarter. A 2028 request does not alter which phones or laptops work well today. Standard refresh cycles and last quarter's proven AI tools remain the sensible baseline, and the on-device pitch is worth tracking as a data-handling question, since Amon framed it around keeping personal data on the handset.

What I would watch is whether a chip at the larger scale is actually announced, and whether memory pricing and availability ease. For a comparison point on local AI hardware, see GIGABYTE Launches 64GB AI TOP ATOM, Out Oct. 23. Until then, this story belongs in the watch pile.