Arm Holdings (ARM) — 2026Q4 FY2026 Earnings Call Analysis

Demand Doubled, Guide Held at Half

They have $2B of AGI CPU demand and are guiding to $1B because they can't secure wafers, memory, or packaging.

Thesis: The IP business is compounding cleanly: ACV +22%, data center royalty doubling two years running, 49% operating margin with OpEx growing slower than revenue. The chip business is a zero-control call option. Arm is fabless with no leverage over TSMC wafers, HBM, or packaging, yet the stock's upside case is priced off a $15B FY31 chip target that depends on supply Arm doesn't own and demand from hyperscalers who build competing Arm CPUs.

Verdict: HOLD — Conviction: MEDIUM

Catalyst: Q4 FY27 chip revenue landing at or above the $90M guide. Jason said Q3 will give a firmer estimate on Q4 delivery. Either the supply materializes and the beat-and-raise thesis holds, or it doesn't and the chip narrative cracks.

Key Risk: Hyperscaler CapEx guidance turning. The entire data center royalty growth, which offsets mobile's negative unit growth, is levered to AWS, Google, NVIDIA, and Meta sustaining current buildout pace. Jason conceded the cycle has run three years and said 'who knows how much longer it goes' while guiding 20%.

The Tell: Jason Child: 'this AI investment super cycle has now gone on for 3 years. So who knows how much longer it goes, but it's at least going to happen for the next year.' The CFO hedging the duration of the cycle that drives the entire stock thesis, while simultaneously anchoring a 10% long-term floor against today's 20% guide. He told you the gap between guided growth and sustainable growth is wider than the multiple implies.

Detected Patterns

Capacity Ceiling: Demand doubled to $2B but guide held at $1B. Rene confirmed supply covers only $1B including wafers, memory, packaging, and test. Arm is fabless IP with no control over TSMC wafer starts or HBM allocation. They cannot scale what they sell.
Backlog Quality: The $2B is described as customer demand and interest, not contracted backlog. Supply for it is not yet secured. Buyers are hyperscalers who also build competing Arm CPUs. William Blair raised channel tension directly; management answered with ecosystem goodwill, no contractual protection.
Mix Shift Spin: Chip business runs at 30% gross margin against an IP business at ~65% operating margin. Jason frames it as profitable due to compute die leverage, but as silicon scales it dilutes blended margins. The 'largest business by FY31' framing masks the margin dilution.
Beat and Raise Machine: Third consecutive year of >20% growth. Beat revenue midpoint, OpEx $10M under guide, ACV +22% running ahead of reported license revenue. Q1 guide at $1.26B implies continued 20% growth with royalty and license both guided at 20%.
Structural Demand Shift: Data center royalty doubled two consecutive years. Google replacing x86 host with Axion on TPU 8t/8i, NVIDIA 256-chip Vera rack, AWS scaling Graviton, SAP moving core database to Arm. ~50% CPU share at top hyperscalers, ~100% in DPUs and SmartNICs.
Rising Customer Switching Costs: All three largest accelerator platforms run Arm CPUs. Software already ported. Rene stated software friction to new Arm deployment is minimal. Cloudflare, F5, SK Telecom design wins extend beyond hyperscaler concentration.

Friction Level: MODERATE_FRICTION — Both sides agree the demand exists. Bull reads the supply hold as sandbagging; bear reads it as a structural ceiling for a fabless IP company that controls zero wafer starts, HBM supply, or packaging capacity. The disagreement is whether the $2B converts or is cancellable soft interest from buyers who also compete with Arm's silicon.

Report not found

The report data is no longer available. Please return to the archive.

Demand Doubled, Guide Held at Half | Silicon Signal