Is Arm’s First Silicon a Hedge Against Royalty Deceleration?

The transcript made clear that licensing strength is being driven by what Rene Haas described as "high-value licenses for next-generation Arm technologies." That likely means compute subsystems and the Neoverse roadmap, signed by hyperscalers and merchant silicon partners locking in multi-year architecture commitments. That is structurally durable revenue, and it explains why Arm could guide Q1 licensing up another roughly 20% YoY without flinching. CFO Jason Child noted that non-GAAP OpEx of $734 million came in roughly $10 million below guide, suggesting management is finding modest leverage even as R&D investment runs hot for the AGI CPU roadmap. The royalty story is more complicated. Management framing emphasizes the doubling in data center royalty, with named customer signals across Google's Axion CPU paired with TPUs, AWS Graviton scaling, Microsoft Cobalt, and NVIDIA's Vera, plus near-saturation share in DPUs and SmartNICs. That all reads cleanly. What gets less airtime is that the rest of the royalty base, mobile, embedded, automotive, appears to have softened materially in the quarter, which is why the blended royalty number landed well below the licensing line. The Q1 guide of "around 20%" royalty growth implies a snapback, and that is the number we will be triangulating against handset and embedded peer prints. On capital allocation, full-year non-GAAP operating expenses ran materially elevated versus prior year, and GAAP operating margin compressed in tandem. That is the price of the silicon pivot. Management is asking investors to trust that the AGI CPU revenue ramp from FY27 onward will more than absorb the spend.

Based on what we are observing, the next four quarters will reframe Arm from an IP house into a hybrid business. The production ramp of the AGI CPU is the single variable that determines whether the FY31 revenue target gets pulled forward or pushed out. The supply-constrained $1 billion FY27 outlook is the conservative anchor. Whether the realized number runs above or below that line will depend on TSMC 3nm allocation, packaging capacity for the 136-core Neoverse V3 design, and Meta's deployment cadence at the Hyperion site and adjacent gigawatt facilities. We will be tracking three signals over the next two quarters: first, whether NVIDIA's May 20 print and Vera attach commentary confirm the Arm royalty trajectory implied by Jensen's trillion-dollar Blackwell-plus-Rubin order book; second, whether the Q1 FY27 royalty line snaps back to the guided level, which would resolve the cloud-versus-everything-else question; and third, whether AGI CPU 2 timing shifts in either direction as supply visibility improves. The licensing tailwind is real, the silicon optionality is meaningful, and the royalty deceleration outside cloud is the story management is least eager to discuss.