Shares of semiconductor design company Arm Holdings (ARM) continued to rise on Wednesday, reaching a record high. This followed optimistic comments from Wall Street analysts, who believe the company will be a major beneficiary of the growth in agentic artificial intelligence (agentic AI).
Bernstein analyst David Dai raised his price target for Arm to $300, implying upside of approximately 17% from current levels. Dai believes that this leading semiconductor company will benefit from a shift in AI computing paradigms.
Currently, AI model training relies heavily on graphics processing units (GPUs) designed by companies like Nvidia. The parallel processing capabilities of GPUs allow them to break complex computational tasks into many smaller tasks and execute them simultaneously, making them highly suitable for model training. However, the rise of agentic AI is shifting the computational focus from model training to inference—the process of using a trained model to make predictions or generate content. This shift is driving demand for central processing units (CPUs), which offer cost advantages for inference tasks.
In March of this year, Arm launched its first data center CPU, a high-performance chip specifically designed to support agentic AI workloads. Dai noted that Arm stands out in the server CPU market due to its unparalleled energy efficiency. He estimates that the server CPU market could grow fourfold by 2030, reaching $137 billion, with Arm capturing a significant share of that market. Accordingly, he forecasts that by 2030, Arm’s sales and profits will increase more than fivefold, reaching $26 billion and earnings per share (EPS) of $9.83, respectively.
McKinsey estimates that by 2030, inference will surpass training as the dominant workload in AI data centers, accounting for more than half of all AI computing capacity. By the end of this decade, inference workloads are expected to represent 30% to 40% of total data center computing demand.
Arm’s business model includes charging customers licensing fees and collecting royalties on each chip designed using its intellectual property (IP). Arm maintains a solid customer base, including Apple, Nvidia, Amazon, and others, all of which use its IP to design chips for smartphones, personal computers, and data centers. Due to the high energy efficiency of its designs, flagship smartphones and custom AI processors that prioritize performance-per-watt widely adopt Arm technology. Companies such as Google, Nvidia, Amazon, Microsoft, and Marvell Technology are all using Arm technology to design their AI data center CPUs.
Counterpoint Research predicts that Arm is poised to dominate the server CPU and custom AI processor market within the next three years, capturing 90% of that space. More importantly, Arm is no longer limiting itself to an IP licensing and royalty model. In March of this year, the company announced it was developing its own chip—the Arm AGI CPU—specifically designed to support agentic AI and inference workloads.
In the just-concluded fiscal year 2026 (ended March 31), Arm’s revenue grew 23% to $4.92 billion, with healthy growth in both licensing and royalty revenues. The company projects that by fiscal year 2031, the addressable opportunity for data center CPUs will reach at least $100 billion, with total revenue potentially jumping to $25 billion and EPS reaching $9.00—a significant increase from $1.77 in fiscal year 2026.
Taken together, the primary drivers behind Arm Holdings’ rising stock price are expectations of increased demand for agentic AI and AI inference. Leveraging its energy efficiency advantages, Arm has secured a favorable position in the server CPU market. Coupled with strong customer demand for its self-developed AGI CPU chip, the company’s revenue growth prospects over the next several years are substantial. If Arm achieves its financial targets as expected, the stock still has further upside potential.
