Arm is building its own chip, and Meta is waiting in line

For decades, Arm has been the ultimate neutral arms dealer in tech. It designed the architecture that powers everything from smartphones to servers, licensed the blueprints to everyone else, and stayed strictly out of the actual chipmaking business.

Now, that neutrality looks like it is coming to an abrupt end.

Arm gets into the silicon business

Arm could launch its very first in-house chip as early as this summer, according to a report from the Financial Times citing people familiar with the matter.

Producing its own physical silicon marks a massive strategic pivot from Arm's traditional model of licensing designs to partners like Nvidia and Apple. Instead of just handing over schematics, Arm wants to ship finished products.

Sources told the Financial Times that the new chip will be a server CPU built specifically for large data center environments. Even more intriguing, it is reportedly being designed with deep customization in mind for hyperscalers like Meta, which is already said to be in line.

This setup puts Arm on a direct collision course with Intel and AMD, the two legacy giants that still supply the overwhelming majority of server CPUs to Big Tech.

Rumors about this move have been building for a while. Back in May, Nikkei Asia reported that Arm was setting up an internal AI chip unit, targeting a working prototype by spring of this year and volume production by the fall.

Arm declined to comment on the report, while Meta did not immediately respond to requests for comment.

Son's $500B master plan

Behind this push is Masayoshi Son, founder of Arm's majority owner, SoftBank Group. Son reportedly views in-house silicon as a critical pillar in his sweeping artificial intelligence infrastructure roadmap.

That vision recently became much clearer with the announcement of Project Stargate, a half-a-trillion-dollar U.S. AI infrastructure initiative alongside OpenAI's Sam Altman and Oracle's Larry Ellison.

SoftBank and OpenAI are co-leading the massive project, with SoftBank taking on the financial heavy lifting, OpenAI overseeing technical operations, and Son serving as chairman.

The venture plans to deploy $500 billion over the next four years to build out domestic AI infrastructure for OpenAI. It kicks off with a staggering $100 billion initial round, backed by OpenAI, SoftBank, Oracle, and Abu Dhabi-based investment firm MGX.

Why it matters

For enterprise buyers and hyperscalers like Meta, getting direct access to fully customized Arm server CPUs is an enormous win. It offers tailored performance per watt and gives them massive pricing leverage over Intel and AMD.

For the semiconductor industry, this changes the fundamental rules of engagement. We think Arm is taking a huge gamble by crossing the line from supplier to direct competitor. Selling chips directly to Meta while continuing to collect licensing checks from Nvidia, Qualcomm, and Apple is bound to create serious friction across its customer base.

At the same time, it looks completely rational from SoftBank's vantage point. If you are spearheading a $500 billion compute buildout with Stargate, collecting a small royalty on each core is chump change. Son wants to capture the full economic value of the silicon that powers the next era of compute.

The real test will be execution. Licensing abstract blueprints is very different from managing real-world silicon supply chains, packaging, and high-volume yields. Arm is stepping onto a battlefield where margins are hard-won and failure is unforgiving.