Arm vs. Qualcomm - Jon Peddie Research

The Arm-Qualcomm dynamic is fundamentally a reallocation of value-chain control, not a licensing dispute. Arm's CSS strategy elevates it from architecture seller to subsystem provider, directly compressing Qualcomm's differentiation at the SoC integration layer. Adreno and Hexagon were Qualcomm's moat; if Arm bundles CPU+GPU+interconnect, that assembly premium gets structurally eroded. Technically, on-device AI inference is rewriting architecture priorities. NPU compute share is jumping from roughly 15% to 40%+. If Arm's Ethos can't match Hexagon's perf-per-watt, Qualcomm will accelerate NPU in-house and weaken Arm CPU's anchoring role. Meanwhile, Neoverse's server penetration via Graviton and Cobalt squeezes x86 — and if Qualcomm enters data center, it directly challenges Arm's server narrative. On compliance, roughly 35% of Qualcomm's revenue is tied to Taiwan, China fab capacity and mainland demand. Escalating export controls hit both advanced-node access and end-market shipments simultaneously. Arm's R&D spans UK, US, and China; any geopolitical friction inflates IP-delivery compliance costs across the chain. Apple is the biggest wildcard. M4 and A19 prove high-end de-Arming is viable, forcing Qualcomm to build irreplaceable AI inference capability or accept a mid-tier integrator role. MediaTek leverages CSS to lower entry barriers, squeezing Qualcomm's mid-range margins further. The 12-to-24-month tell: does next-gen Snapdragon still adopt Arm CSS, or shift to an Arm CPU plus in-house GPU/NPU hybrid? If the latter, Arm's revenue model migrates from per-core to per-IP-block, reshaping ecosystem bargaining power. RISC-V crossing 5% penetration in edge AI is the real structural threat to Arm's franchise.