Arm Sees Agentic AI Demand Surge as AGI CPU Customer Orders Top 2 Billion Dollars by Fiscal 2028

Arm says demand for its new AGI CPU has accelerated much faster than initially expected, with the company now reportingmore than 2 billion dollars of customer demand across fiscal 2027 and fiscal 2028. That is more than double what Arm communicated when it introduced the product in March, and it marks one of the clearest signs yet that agentic AI is reshaping how cloud and AI infrastructure buyers think about CPUs, not just accelerators.

There is one important correction to the framing around the number. Arm did not double its formal revenue forecast to 2 billion dollars by 2028. What the company actually said is that it now has more than 2 billion dollars in customer demand for the AGI CPU across fiscal 2027 and fiscal 2028, while it remains on track toward its previously stated silicon business forecast discussed at Arm Everywhere. In other words, demand has surged, but the realized revenue path still depends on how much supply Arm can secure and deliver. Recent reporting from major financial outlets says supply remains the main constraint stopping Arm from immediately lifting revenue guidance in the same proportion.

That demand jump is being driven by a growing customer list that now includes OpenAI, Cerebras, Positron, and Rebellions, all of which Arm says are integrating the AGI CPU alongside accelerator based systems. Arm is positioning the chip as the orchestration and coordination layer for the agentic AI era, where CPUs handle accelerator management, control plane tasks, data movement, and broader infrastructure coordination while GPUs and NPUs focus on the heavy model execution layer.

Arm also highlighted Verda, a European AI cloud provider, as a recent deployment partner for agentic AI orchestration. At the same time, Arm says commercial systems based on the AGI CPU are already available to order from ASRockRack, Lenovo, Quanta, and Supermicro, showing that the company is trying to move beyond concept validation and into real product availability.

The AGI CPU itself is strategically important because it marks Arm’s shift from being primarily an IP and platform company into directly offering production silicon for the data center. Arm officially launched the AGI CPU in March as its first Arm designed data center CPU, built on the Arm Neoverse platform and aimed specifically at large scale agentic AI infrastructure. The company says customers can now adopt the Arm compute platform through IP, Compute Subsystems, or complete silicon, while keeping a shared architecture and software ecosystem.

Arm’s broader cloud position helps explain why this new silicon push is gaining traction so quickly. In the same quarterly results, the company said Arm based CPU compute now represents about 50 percent share among top hyperscalers, with Amazon, Google, and NVIDIA already integrating Arm based CPUs as head nodes alongside accelerator systems. That is a meaningful claim because it suggests Arm is no longer fighting just for relevance in the AI data center. It is already becoming part of the default infrastructure stack for some of the industry’s largest deployments.

From a market standpoint, the message is clear. AI infrastructure demand is broadening beyond accelerators, and CPUs are regaining strategic importance as agentic workloads become more distributed, persistent, and orchestration heavy. Arm is trying to capture that moment by offering a purpose built CPU that slots directly into AI first data center designs, and the early commercial response suggests the company has found a real opening. That conclusion is an inference based on Arm’s disclosed customer demand, partner list, and hyperscaler market share commentary.

The bigger test now is execution. Arm has the ecosystem momentum, the customer list, and the demand signal, but converting more than 2 billion dollars of demand into shipped silicon revenue will depend on supply chain readiness and production scale. If it can deliver, the AGI CPU could become one of the most important new products in the agentic AI infrastructure race over the next two years.

Do you think Arm’s AGI CPU can become a real control layer standard for agentic AI data centers, or will hyperscalers still prefer building around their own custom CPU strategies?