Arm Holdings has entered the chip-selling business for the first time, unveiling an AI data-centre CPU it is branding around the agentic AI era and projecting the product line will add approximately $15 billion in annual revenue within five years.
The announcement marks a clean break from the business model that has defined Arm since its founding: licensing chip architecture to third parties rather than competing in silicon markets directly. Analysts quoted across multiple outlets described the pivot as the most significant shift in the company's history.
Meta has been named as a partner in developing what both companies are calling a new class of data-centre silicon, lending immediate commercial credibility to a product that has yet to reach volume production.
Arm's shares rose between 15 and 20 percent on the day of the announcement, with the range reflecting intraday volatility across trading sessions. The move drew attention partly because it repositions Arm as a potential direct vendor to hyperscalers rather than simply a royalty collector upstream of them.
Barron's noted that Nvidia's stock also moved higher on the day, with analysts arguing the new Arm CPU targets inference and agentic workloads rather than the GPU-accelerated training infrastructure that underpins Nvidia's dominant position. The two product categories are complementary for now, though the longer-term competitive implications depend on how quickly Arm can scale the new line.
Arm's existing revenue base is heavily weighted toward mobile, where its architecture powers virtually every smartphone processor. The data-centre push is an attempt to replicate that dominance in a faster-growing market, though the $15 billion projection is a five-year target against a backdrop of intense competition from established x86 incumbents and custom silicon efforts from the hyperscalers themselves.
