Arm Seen Dominating $50 Billion AI Inference Market as Workloads Reach Two-Thirds of Compute
Updated
Updated · The Motley Fool · May 23
Arm Seen Dominating $50 Billion AI Inference Market as Workloads Reach Two-Thirds of Compute
2 articles · Updated · The Motley Fool · May 23
$50 billion in AI inference chip demand this year could favor Arm over Nvidia, AMD, Broadcom and Intel, as analysts argue its energy-efficient designs are better positioned for the next phase of AI.
Two-thirds of AI computing power is expected to go to inference in 2026, up from 50% in 2025, while McKinsey sees data-center inference workloads rising from 21 gigawatts last year to 93 gigawatts by 2030.
Arm already sits inside key inference hardware: Nvidia’s Grace and Vera server CPUs use its architecture, and Google, Amazon, MediaTek, Qualcomm and Apple also rely on Arm designs across data centers and devices.
20% annual royalty growth through fiscal 2031 is one pillar of Arm’s expansion, helped by Armv9 royalty rates that are nearly double Armv8’s and by a new push into selling its own silicon.
25 billion in fiscal 2031 revenue is Arm’s target, more than 5 times its trailing $4.7 billion, with non-GAAP EPS projected above $9.00 from $1.77 in fiscal 2026.
With AI's future tied to Arm, is the industry ignoring the massive risk of its concentrated supply chain?
Arm promises energy-efficient AI, but will this simply fuel an even greater surge in global power consumption?
By selling its own chips, is Arm risking its neutral status and alienating its most important licensing partners?