Arm Holdings PLC (NASDAQ:ARM) is moving beyond its traditional licensing business, seeking a larger footprint in the data center CPU market amid surging demand for AI compute, according to UBS.
At its “ARM Everywhere” event, the company unveiled a 136-core CPU claiming roughly double the performance per watt of comparable x86 chips, targeting AI workloads and non-Nvidia XPU server head nodes.
Arm argues that emerging “agentic AI” workloads are reshaping system architecture. As orchestration, control, and system-level functions become more CPU-intensive, servers may increasingly deploy one or more CPUs per GPU, boosting CPU content per rack and widening Arm’s addressable market.
Customer interest appears strong. Eight companies, including Meta Platforms, OpenAI, and enterprise players such as SAP and Cloudflare, have already engaged with Arm, despite the product still in early ramp. First-generation chips are sampling now, with production slated before year-end; follow-on generations are under development.
The company projects its total addressable market expanding from roughly $2.4 billion in FY26 IP/CSS revenue to $24 billion including merchant CPUs, and potentially surpassing $100 billion by FY31—well above prior industry estimates. UBS highlights that the new CPU business, layered atop Arm’s high-margin licensing model, could generate nearly $15 billion in revenue by FY31, roughly 15% of the projected AI server CPU market.
Financially, Arm expects revenue of $7.8 billion and $3 or higher EPS by FY28, rising to $25 billion and $9-plus EPS by FY31. Chip gross margins are projected to improve from mid-30% to around 50%, supporting operating margins above 30%. UBS notes that while the stock currently trades at about 50 times FY28 EPS, the multiple falls to roughly 16 times on FY31 estimates, making the long-term growth story more compelling.
Software compatibility is less of a barrier than in previous cycles. Arm emphasizes that AI workloads simplify transitions from x86, and unifying CPUs with DPUs and SmartNICs on a common instruction set reduces friction for large-scale deployments.
Cloud AI represents the most significant growth opportunity, with Arm estimating the market could expand from $210 billion in FY26 to $1.2 trillion by FY31, far outpacing Edge and Physical AI segments. Potential royalty streams from partnerships with SoftBank and OpenAI could provide incremental upside.
Arm’s strategy marks a significant evolution: from an IP licensor to a vertically integrated participant in AI infrastructure. While execution risks remain, especially scaling the chip business against entrenched x86 competitors, the company’s early traction and expanded TAM suggest a credible path for long-term growth.