Arm launches first in-house CPU with Meta as anchor customer
Strategic pivot from pure licensing to competing with Intel and AMD in server processors
Arm Holdings has launched its first proprietary central processing unit, marking a historic departure from its decades-long business model of licensing chip architecture to other companies. The new Arm AGI CPU, designed specifically for data center artificial intelligence workloads, will debut with Meta as its lead customer and co-developer.
Shares of the $140.6 billion chip IP licensor rose 1.3% to $138.66 in Tuesday trading, hovering near its 200-day moving average of $138.13. The stock has rallied 73% from its 52-week low of $80, though it remains 24% below its November peak of $183.16.
The Arm AGI CPU features up to 136 Arm Neoverse V3 cores per processor, with 6 gigabytes per second of memory bandwidth per core at sub-100 nanosecond latency. Operating at a 300-watt thermal design power, the chip supports high-density 1U server chassis for air-cooled deployments, allowing up to 8,160 cores per rack, while liquid-cooled systems can achieve more than 45,000 cores per rack.
Arm claims the new processor delivers more than double the performance per rack compared to x86 platforms dominated by Intel and Advanced Micro Devices. The company estimates this efficiency can lead to up to $10 billion in capital expenditure savings per gigawatt of AI data center capacity.
"The rise of AI agents is increasing demand for CPUs, particularly for reasoning, coordination, and data movement in AI systems," Arm said in its announcement. "Data centers are expected to require more than four times the current CPU capacity per gigawatt for agent-driven applications."
The partnership with Meta represents a significant win for Arm as it seeks to establish its silicon products business. Meta is working with Arm to optimize the social media giant's infrastructure for its family of applications, complementing Meta's existing custom silicon strategy through its Meta Training and Inference Accelerator (MTIA) chips.
Beyond Meta, Arm has secured commitments from a roster of technology companies including Cerebras, Cloudflare, F5, OpenAI, Positron, Rebellions, SAP, and SK Telecom. Original equipment manufacturers ASRock Rack, Lenovo, Quanta Computer, and Supermicro are preparing systems with the new processor, with early availability already underway and broader shipments expected in the second half of the year.
The strategic shift represents both an opportunity and a risk for the Cambridge-based company. Expanding into silicon production opens a significantly larger addressable market—Arm estimates data centers will require more than four times current CPU capacity for agent-driven applications—but also creates potential competitive tension with existing licensees such as Qualcomm that have built their own data center processors using Arm's architecture.
Analysts remain broadly bullish on the stock, with 26 of 41 analysts rating Arm a buy or strong buy and an average target price of $151.86, according to current market data. However, the company's valuation remains elevated at 176 times trailing earnings, though the forward price-to-earnings multiple contracts to 61.4.
Arm's pivot to silicon production comes amid intensifying competition in the data center processor market. Intel and AMD continue to dominate x86-based server chips, while custom silicon from cloud providers including Amazon Web Services, Google, and Microsoft has eroded their market share in recent years. The emergence of specialized AI processors from companies like Nvidia and Cerebras has further fragmented the landscape.
For Arm, the move into production silicon is the most significant strategic evolution since its acquisition by SoftBank in 2016 and subsequent public listing in 2023. The company has traditionally generated revenue by licensing its processor architecture and collecting royalties on chips shipped by customers. Adding its own silicon products creates a new revenue stream that could accelerate growth but also fundamentally changes its relationship with existing partners.
The success of the Arm AGI CPU will depend on adoption rates beyond the initial partner ecosystem and the company's ability to scale manufacturing and support operations. With the data center chip market projected to grow substantially through the end of the decade as AI workloads expand, Arm is betting its future on capturing a meaningful share of that expansion through both its traditional licensing business and its new silicon products division.