ARM Avoids Customer Conflict by Targeting Underserved CPU Segment

ARM licenses designs to customers like Apple (fully custom M1 from ARM ISA for consumer perf/battery) and AWS/Nvidia (Neoverse V2 cores like Graviton/Grace for faster server time-to-market). Customizing deep into microarchitecture risks delays—Apple's Intel-to-ARM took 3 years—and capital intensity erodes margins (Intel/AMD at 35-50% vs. ARM's 95%). AGI CPU doesn't compete: it enters total data center market directly, with pre-launch interest from OpenAI, Cloudflare, Meta (release H2 2026), letting licensees focus on peripherals like Nvidia's LPDDR5X/MVLink for AI.

Data Centers Shift to CPU-Heavy Mix for Agentic Workloads

Gigawatt-scale facilities (e.g., Stargate's 1.2GW Abilene site + 6x1GW) allocate power across CPU/GPU/RAM like a 700W PC budgets for gaming vs. editing. Pre-2023 training/inference favored GPUs; now agentic AI demands 4x more CPU cores (30M to 120M per GW) as agents proliferate. ARM's AGI CPU optimizes for this, validating demand via hyperscaler buy-in before TSMC production.

$15B Revenue Upside Faces Fierce Customization Competition

ARM projects $15B from AGI CPUs in 5 years, with 50% line margins blending to 75% overall (diluting from 95%). Growth aligns with AI demand, but hyperscalers build in-house (Google Axion, Microsoft Cobalt, Alibaba Echin) atop ARM/AMD while Ampear competes directly. Success hinges on ARM's design edge plus superior production speed/integration to outpace custom forks and x86 holdouts.