For 35 years, Arm Holdings (ARM) has operated largely behind the scenes. Its chip architecture powers more than 99% of the world’s smartphones, yet the company itself has never made a physical chip—it only licensed the blueprints.
That changed this week. At an event in San Francisco, Arm unveiled its first-ever in-house silicon: the Arm AGI CPU. The move marks a historic pivot for the British chip designer, as it transitions from a pure intellectual property (IP) licensor to a player that both designs and sells its own physical silicon. With demand for AI infrastructure surging and the battle for computing efficiency intensifying, Arm’s strategic shift is being closely watched—and raises a critical question for investors: can it replicate Nvidia’s trajectory?
Arm’s traditional business model has long been built around licensing. Whether it’s Apple or Qualcomm using Arm’s architecture to build custom chips, or other manufacturers purchasing off-the-shelf designs and paying royalties, Arm has never directly produced silicon.
That model is now evolving. Mohamed Awad, Arm’s head of cloud AI, described the new Arm AGI CPU as “the first production silicon from Arm, designed for AI infrastructure at scale.” The chip is “ruthlessly optimized” for artificial general intelligence (AGI), featuring up to 64 CPUs and roughly 8,700 cores. According to Awad, it delivers “two times the performance-per-watt than you can get from an x86 rack.”
The shift reflects the broader AI industry’s obsession with efficiency. As power consumption becomes the key constraint in data centers, Arm’s inherently efficient architecture is gaining new appeal. “That means twice as much performance in the same footprint, in the same power,” Awad said, noting that customers are increasingly looking to maximize returns on every dollar spent on AI infrastructure.
If Arm’s new chip is the arrow, its ecosystem of early adopters will determine how far it flies.
Meta Platforms (META) played a central role as both co-developer and the first large-scale deployer of the Arm AGI CPU. The chip is designed to work alongside Meta’s in-house MTIA (Meta Training and Inference Accelerator) chips, with the two companies committing to collaborate across multiple generations of the Arm AGI CPU roadmap.
But Meta is far from alone. Arm’s inaugural customer list includes Cloudflare, F5, OpenAI, SAP, and SK Telecom—spanning content delivery, cloud security, enterprise software, and telecommunications. The lineup reflects strong early validation for Arm’s silicon push, with customers spanning the AI supply chain from cloud infrastructure to application layers.
Arm said the move is intended to offer partners a full range of options—from IP and subsystems to its own silicon—aimed at accelerating AI deployment.
Arm already counts Apple, Nvidia, Amazon, Google, and Microsoft among its major customers. To date, more than 350 billion Arm-based chips have shipped, and the Arm ecosystem includes over 22 million software developers.
But branching into silicon production represents a fundamental shift. Arm is no longer just an upstream supplier; it now competes—at least in part—with some of its own customers. Still, the company is focused on the incremental opportunity: AI infrastructure. Arm estimates the AI CPU market alone could reach $1 trillion, and the company hopes self-developed silicon will help it capture a larger slice.
Financially, Arm remains on solid footing. The company’s latest results show gross margins of 97.5% and net margins of 17.15%, with a debt-to-equity ratio of just 0.06—signaling strong profitability and a clean balance sheet. However, operating margins have declined at an average annual rate of 13.8% over the past five years, pointing to rising costs and increasing competition in its core IP licensing business.
For investors, Arm’s hardware push brings valuation front and center.
The stock currently trades at a trailing P/E above 180 and a P/S ratio above 31—well above historical medians, reflecting elevated market expectations. Yet its forward price/earnings-to-growth (PEG) ratio stands at 0.57, below the threshold of 1 that often signals undervaluation. If Arm delivers on the growth implied by its AI ambitions, the current price may not yet reflect its full potential.
Analysts are cautiously optimistic, with a consensus price target of roughly $149.70. Technical indicators show the stock’s RSI at 66.93, nearing overbought territory—a sign that near-term volatility is possible.
Arm’s move into self-developed silicon marks one of the most significant strategic shifts in its 35-year history. By transitioning from an enabler to a direct participant in the AI hardware market, the company is positioning itself to capture more value in the AI value chain.
Early support from Meta and a roster of blue-chip customers provides near-term validation. The longer-term question is whether Arm can grow its new silicon business without disrupting its core IP licensing model—and whether that combination can fuel a Nvidia-like run.
For investors, a PEG of 0.57 offers a foundation for value-focused narratives, but a triple-digit P/E serves as a reminder: replicating Nvidia’s path won’t come without challenges.
