ARM Holdings Reports Q4 FY2026 as AI Royalties Drive Stock to 84% YTD Gain

For most of the semiconductor industry’s history, the chip IP business operated quietly in the background. Companies paid Arm Holdings a modest licensing fee, built products on its architecture, and paid royalties on units shipped — a percentage of chip revenues so small that it barely registered against the costs of silicon design, manufacturing, and packaging. That model is breaking down, and Arm’s fiscal Q4 2026 earnings, released Wednesday after market close, mark the latest milestone in a dramatic reshaping of semiconductor economics.

Analysts entered the report expecting $1.47 billion in quarterly revenue and adjusted EPS of $0.58 — both records for the company, representing 19% revenue growth year-over-year. But the topline numbers undersell the structural shift underneath them.

What’s Driving the Revenue Surge

ARM’s business has historically tracked smartphone and IoT volumes. Royalty revenue depended on unit count: more phones meant more ARM royalties. That equation has changed in two ways.

First, the average royalty rate per chip has risen substantially as the industry migrates from ARM’s older v8 architecture to v9. The newer architecture carries royalty rates that analysts estimate are roughly twice those of v8, meaning that royalty revenue grows even without a unit volume increase. As Apple, Qualcomm, and Samsung ship v9-based chips in flagship devices, ARM’s per-unit economics improve structurally, not just cyclically.

Second — and more significantly for recent quarters — ARM has become the dominant architecture for data center computing, a market where individual chips carry far higher average selling prices and therefore generate far larger royalty checks than smartphone processors.

In Q3 FY2026, the most recent quarter before Wednesday’s report, ARM disclosed that data center royalties grew triple digits year-over-year. That trajectory reflected the rapid deployment of ARM-based custom silicon by hyperscalers: Amazon’s Graviton 4, Google’s Axion, Microsoft’s Cobalt 100, and Alibaba’s Yitian 710 are all ARM-based designs, and all began ramping in volume through 2025 and 2026.

The AI inference workload driving much of this data center build-out is particularly well-suited to ARM architectures, which offer better power efficiency per computation than the x86 designs they are displacing in the rack. At the scale hyperscalers operate — millions of servers, each running continuously — power efficiency translates directly to billions of dollars in operating cost savings.

Compute Subsystems: The Strategy Behind the Numbers

Less visible in the headline royalty figures but increasingly important to ARM’s growth trajectory is the success of its Compute Subsystems (CSS) program. CSS represents a shift from ARM’s traditional licensing model — providing architecture specifications that customers implement themselves — toward providing pre-integrated subsystems that customers can drop into chip designs with minimal customization.

For customers, CSS dramatically reduces design time and cost. For ARM, it commands higher licensing fees and, crucially, higher royalty rates on the resulting chips because the design complexity ARM bears has increased. In Q3 FY2026, ARM reported 21 CSS licenses across 12 companies, with five customers already shipping CSS-based chips into production. That pipeline is significant: as those chips ship in volume, they generate royalty streams at rates substantially above the blended portfolio average.

The CSS program also gives ARM a path into markets where it previously had limited presence. Automotive chips, custom AI accelerators, and networking processors are all areas where CSS customers are actively shipping or preparing to ship, adding revenue streams that were negligible three years ago.

The CPU Ambiguity

In March 2026, ARM announced that it would build its own CPU for the first time in the company’s three-decade history — a chip that would compete directly with the products its customers build using ARM’s architecture. The announcement created immediate tension with ARM’s major licensees, some of whom questioned whether ARM was transforming from a neutral IP supplier into a competitive threat.

ARM’s leadership has been careful to frame the CPU initiative as complementary to the licensing business rather than substituting for it. The company describes the chip as targeting specific deployment scenarios — edge AI inference, in particular — where it can demonstrate performance and power characteristics that might not otherwise be achievable using standard CSS. The goal, ARM argues, is to expand the market for ARM-based silicon, not to capture revenue that would otherwise go to licensees.

Wall Street has been skeptical but constructive. ARM stock’s 84% YTD gain through the pre-earnings period reflects confidence that the royalty rate expansion and data center ramp more than compensate for any licensee relationship friction. How ARM characterizes the CPU initiative in Wednesday’s earnings call will be closely watched.

Full-Year and Forward Guidance

Analyst consensus for the full FY2026 year points to royalty revenue nearly doubling versus FY2025, driven by the combination of v9 migration, data center ramp, and CSS adoption. Licensing revenue growth has been more episodic — large contract signings can spike a quarter’s numbers significantly — but the underlying trend is upward as customers pay more for higher-value design services.

For Q1 FY2027 guidance, analysts will be listening for signals on several fronts: the pace of hyperscaler custom silicon ramp in the second half of 2026, the trajectory of v9 penetration in mid-range and value smartphones (which represent vastly larger unit volumes than flagships), and any update on the own-CPU program’s commercial timeline.

The larger question hanging over ARM going into Wednesday’s report is whether the company can sustain its royalty rate expansion story as the industry catches up to the v9 transition and as hyperscalers complete their initial custom silicon ramps. If ARM can continue growing royalty rates through CSS adoption, new markets, and the CPU initiative, the FY2026 results may look modest in retrospect. If the rate expansion plateaus, revenue growth returns to being a function of unit volumes — a slower, less exciting story.

Either way, the quarter represents the clearest picture yet of what the chip IP model looks like when AI infrastructure spending is the dominant driver of semiconductor demand.