Arm just reported $1.29 billion in Q1 FY27 revenue, up 22% year over year, with data center royalties more than doubling. The number that matters more is buried in an IDC citation: spending on Arm-based accelerated server platforms has nearly doubled in the past two quarters and surpassed x86 platforms. That is the first time Arm has crossed that threshold, and it happened without announcement.
The mechanism is Neoverse and what Arm is calling CSS (Compute Subsystems), which bundle CPU, interconnect, and memory controller into licensable blocks that hyperscalers can drop into custom silicon without redesigning from scratch. The last 500 million Neoverse cores shipped in nine months, a pace that took years to reach in earlier generations. At the same time, the AGI CPU, launched in March 2026, has already pulled forward more than $2 billion in customer commitments across FY27 and FY28, with Qualcomm now entering the AI datacenter CPU market on the Arm Dragonfly C1000. Google committed Axion as the host CPU for its TPU systems. AWS locked in Graviton5 for Meta's agentic AI fleet. Microsoft expanded Azure Cobalt 200.
The decision that x86 had locked in AI server infrastructure is now the wrong prior. Hardware teams building AI accelerators, inference clusters, or edge-to-cloud pipelines that assumed x86 as the host substrate should run that assumption through the current numbers. Arm-based host CPUs are now the majority platform for AI-optimized server spending, and the royalty rate per chip is higher on Armv9 and CSS than on prior generations. The pricing model that treats Arm as the low-power mobile play and x86 as the serious server play is at least two years behind the market.