Arm flexes its chip-making muscles (with Meta’s help)

• Chip designer Arm has expanded its horizons and become a chip maker for the first time
• Its first product is a CPU designed for agentic AI infrastructure
• The chip has been co-designed by Meta, which already plans to deploy it, as do many others, including SK Telecom

Arm is no longer just a leading chip designer that licences its intellectual property to semiconductor manufacturers – it’s now also a chip maker itself, and its debut product is a central processing unit (CPU) designed for agentic AI infrastructure.

It’s a move that has been expected for some time and one that puts the UK-based company into more direct competition with Intel and AMD: Indeed, Arm claims its debut product offers “more than 2x performance per rack compared with x86 platforms, enabling up to $10bn in capex savings per GW of AI datacentre capacity” (though this is all according to its own estimates).   

Arm hasn’t made this move in the hope it will win some business – it has a lot of big name support for its strategy and, indeed, its initial chip product has co-developed with big tech giant Meta (more on all this later). 

It’s also worth remembering that Arm is majority owned by SoftBank Group, the Japanese tech and investment giant that is run by Masayoshi Son, one of the big hitters in global AI infrastructure, and Arm would not have made this move without Son’s blessing.   

So what has Arm announced? It has unveiled the “next evolution of the Arm compute platform, extending into production silicon products for the first time in the company’s history,” the company noted in this announcement. 

“This begins with the launch of the Arm AGI CPU, an Arm-designed CPU for AI datacentres, built to address a rising class of agentic AI workloads,” it added. There’s a lot of information and data about the product in Arm’s extensive press release and in this blog.  

Meta will deploy Arm’s chip to “optimise infrastructure for its family of apps and, working alongside Meta’s own custom silicon called Meta Training and Inference Accelerator (MTIA), enabling more efficient orchestration in large-scale AI systems.” The two companies announced last October a “strategic partnership to power the next era of AI”, and say they are now “committed to collaborating across multiple generations of the Arm AGI CPU roadmap”.

“Delivering AI experiences at global scale demands a robust and adaptable portfolio of custom silicon solutions, purpose built to accelerate AI workloads and optimise performance across Meta’s platforms,” said Santosh Janardhan, head of infrastructure at Meta. “We worked alongside Arm to develop the Arm AGI CPU to deploy an efficient compute platform that significantly improves our datacentre performance density and supports a multi-generation roadmap for our evolving AI systems.”

Meta isn’t the only company supporting Arm’s move. The chip firm says it has “confirmed additional commercial momentum” with partners such as Cerebras, Cloudflare, F5, OpenAI, Positron, Rebellions, SAP, and SK Telecom. “These customers will deploy the Arm AGI CPU for key agentic CPU use-cases, including accelerator management, control plane processing and cloud and enterprise-based API, task and application hosting.”

To help meet the needs of such customers, Arm is partnering with leading equipment and device manufacturers, including ASRock Rack, Lenovo, Quanta Computer and Supermicro, with early systems available now and broader availability expected in the second half of the year.

Arm added: “More than 50 leading companies across hyperscale, cloud, silicon, memory, networking, software, system design and manufacturing are supporting the expansion of the Arm compute platform into silicon. That momentum includes industry leaders, such as AWS, Broadcom, Google, Marvell, Micron, Microsoft, Nvidia, Samsung, SK Hynix and TSMC, alongside many others.”

Suk-geun (SG) Chung, CTO and head of AI CIC at SK Telecom, stated in this blog: “SK Telecom is expanding into large-scale, full-stack AI inference datacentre infrastructure, which includes Arm AGI CPU and Rebellions AI accelerator chip. By bringing together our sovereign A.X foundation model with inference-optimised AI servers, we are ready to deliver it to [the] world while elevating our AIDC competitiveness.” 

Arm will have to tread carefully, though, as its new product is, on paper at least, competing with existing products developed by some of its partners, including the Nvidia Grace CPU and AWS Graviton.  

Rene Haas, CEO at Arm, stated: “AI has fundamentally redefined how computing is built and deployed. Agentic computing is accelerating that change. Today marks the next phase of the Arm compute platform and a defining moment for our company. With the expansion into delivering production silicon with our Arm AGI CPU, we are giving partners more choices, all built on Arm’s foundation of high-performance, power-efficient computing, to support agentic AI infrastructure at global scale.”

The move is set to transform Arm’s financial performance. The company expects meaningful revenues from the new product to be reported in the financial year that end in March 2028, by which time it believes AGI CPU sales will have totalled about $1bn. But then those revenues are expected to ramp up, with Arm estimating that its CPU product sales will be in the region of $15bn for the financial year that ends March 2031. By comparison, Arm’s total full year revenues for the fiscal year that ends next week are set to be less than $5bn. 

Not surprisingly, investors, as well as customers, are excited about the potential of Arm’s first foray into making its own processors, and its share price gained more than 9% in value to hit $147.63 in pre-market trading on the Nasdaq exchange on Wednesday morning. 

- Ray Le Maistre, Editorial Director, TelecomTV