ARM is designing proprietary AI accelerator chips, with Meta positioned as an anchor buyer.

Arm Holdings Plc., the U.K.-headquartered chip design company, is reportedly planning to launch its first-ever complete semiconductor after securing Meta Platforms Inc. as one of its first customers, according to an exclusive report by the Financial Times. This move would bring Arm into direct competition with many of its existing customers for the first time. Historically, Arm has licensed its instruction set architecture—a blueprint for building powerful and energy-efficient chips—to others rather than manufacturing its own hardware. More recently, the company has begun selling more complex core designs that customers can customize.

Arm's role as a neutral intermediary has earned it the nickname "Switzerland" of chip technology firms, as it has traditionally dealt impartially with major customers including Apple Inc., Nvidia Corp., Qualcomm Inc., Intel Corp., Amazon Web Services Inc., and Microsoft Corp. However, the massive investment surge in artificial intelligence infrastructure is proving too tempting to ignore. Meta alone has announced plans to spend up to $65 billion on AI infrastructure this year, while Google plans to invest $75 billion and Microsoft will spend $80 billion. Though much of Meta's spending will go toward Nvidia's graphics processing units, substantial funds will also purchase central processing units from Intel and Advanced Micro Devices Inc., and Meta is believed to be developing its own chip designs. The Financial Times reports that Arm is specifically looking to design a CPU for servers, rather than compete with Nvidia and AMD in the GPU market.

The broader strategic context for Arm's shift involves its majority shareholder SoftBank Group Corp. and founder Masayoshi Son's vision for AI infrastructure dominance. SoftBank founder Masayoshi Son appeared at the White House alongside U.S. President Donald Trump, OpenAI Chief Executive Sam Altman, and Oracle Corp. founder and Chief Technology Officer Larry Ellison to unveil Project Stargate, which plans to invest $500 billion in AI infrastructure, with SoftBank and Abu Dhabi state fund MGX providing much of the funding and Arm as a technology partner. During Arm's latest financial earnings call, CEO Rene Haas cited Meta's, Google's, and Microsoft's spending commitments as major opportunities, stating, "No one is pulling back." Additionally, SoftBank is currently attempting to acquire Ampere LLC, a chipmaker backed by Oracle that produces CPUs for data center servers, which the Financial Times indicates is critical to Arm's plans. Arm's stock has already increased 29% this year and has a market capitalization exceeding $173 billion after the company went public following Nvidia's failed 2020 acquisition attempt for $40 billion.

Industry analysts warn of significant risks associated with this strategy. Rob Enderle of the Enderle Group noted, "Competing with your licensees is a good way to lose those licensees if you're not careful, and this does represent a risk to those customers who license Arm technology. It means they'll be competing with Arm itself while also using its technology, but of course, Arm will have a significant advantage as it owns that technology." Holger Mueller of Constellation Research Inc. described the move as "a risky gamble for the company," explaining that any shift from pure licensing revenue would dilute profit margins and expose Arm to supply and production issues and greater stock volatility. Mueller added, "This will expose Arm to supply and production issues and greater stock volatility due to the roller-coaster nature of chip market sales, and it will upset its existing customers. It's surprising, but in any case Arm's management is aware of all of this, so it must be able to see an upside." However, Enderle suggested the lucrative AI industry could generate enough direct sales to offset licensing revenue losses.

The Financial Times suggests Arm may announce its plans by summer or possibly even earlier, though the company has declined to comment on the report. This development aligns with broader industry trends, as Reuters reported four days ago that OpenAI is developing its own chips to reduce reliance on Nvidia's GPUs, already in advanced design stages with plans to transfer designs to Taiwan Semiconductor Manufacturing Co. for experimental production.