Qualcomm & The Inference Fabric Thesis

This week, Qualcomm reported record Q1 FY2026 revenue of $12.3 billion, and the stock dropped. Not because the results were bad, but because the forward guidance ($10.2–$11B for Q2) exposed a structural tension that cuts to the heart of where the AI market is heading: the same AI data center boom that is minting money for Nvidia is now starving Qualcomm’s handset customers of memory. HBM demand from AI training is cannibalizing DRAM supply for smartphones.

This is not a Qualcomm problem. It is the AI market map revealing its internal contradictions and, in doing so, showing us what comes next.

The conventional AI market map is organized around a simple hierarchy: training at the top (Nvidia’s kingdom), inference in the data center (contested territory), and edge/on-device AI at the bottom (everybody else). This hierarchy mirrors how the market developed, but it does not predict where it’s going. Qualcomm’s business model evolution, from mobile chipmaker to what CEO Cristiano Amon calls “a completely different company with relevance in many, many markets,” provides the structural clue.

In a January 2026 conversation with Alex Kantrowitz at Davos, Amon articulated the thesis with unusual clarity. When asked about the data center buildout, he didn’t defend Qualcomm’s position in the GPU race. Instead, he reframed the entire competitive landscape: “The winner of the edge is going to be the winner of the AI race.”

That statement, from the CEO whose company spans 5 watts (earbuds) to 500 watts (data center racks), is not marketing. It is a structural prediction about where value will concentrate as AI moves from training to deployment. And the evidence increasingly supports it.