Arm Holdings: AGI CPU Demand Tops $2 Billion as Supply Confidence Builds, But Smartphone Royalty Growth Slips to High Teens
Q1 fiscal 2027 earnings call, July 29, 2026
Arm Holdings delivered a record first quarter of fiscal 2027, with revenue up 22% year-over-year to $1.29 billion, but the more consequential story for investors is what CEO Rene Haas and CFO Jason Child revealed about the trajectory of the company's new Arm AGI CPU silicon business and the widening gap between blowout cloud AI demand and a softening smartphone market.
AGI CPU demand doubles the original target, supply risk eases
Arm introduced its Arm AGI CPU product in March with a stated $1 billion revenue opportunity across fiscal 2027 and fiscal 2028, backed by roughly $2 billion of pipeline demand. Ninety days later, that pipeline has grown to "north of $2 billion," according to Child, with new customers added in both the U.S. and China. More importantly, Arm says it has now secured the manufacturing capacity to fully support the original $1 billion commitment and has made progress toward securing supply for the incremental demand above that.
Haas was direct about what changed: "Our ability to secure that supply for the north of $1 billion, our confidence has increased in the last 90 days. And I would say it's across all of those areas... wafers, substrates, test capacity, memory." When Bank of America's Vivek Arya pressed on why an extra $1 billion of supply should be so hard to secure in what he called a $50 billion market, Haas pushed back on the framing: "It is a very, very, very tight market across everything... I think even my compatriots who are in the CPU business for a living and have their own fabs have not been able to supply the demand."
Gross margin guidance for the AGI CPU business is unchanged for now, staying in the high-30% to low-40% range through fiscal 2028, with a path to 50% over "a couple of years" as Arm brings more of the manufacturing process in-house rather than relying on ASIC partners. Child flagged that recent industry-wide price increases on components are still being digested and could affect that math; a fuller update on both revenue and margin is promised at the Q3 print, once Arm has visibility into fiscal Q4 and fiscal 2028.
TAM estimates for CPU-based AI compute have exploded since March
Perhaps the most striking data point on the call was the scale of upward revision to total addressable market estimates across the industry. When Arm first floated a $100 billion CPU TAM for the AGI CPU category at its March event, Haas noted "there was some surprise around it" because prior industry consensus sat closer to $50-60 billion. Since then, peers have cited numbers as high as $200 billion, and Child pointed to a most recent estimate of $220 billion "sitting out there from a variety of sources" — a figure AMD also referenced at its own Analyst Day last week, primarily driven by agentic AI workloads.
Haas's explanation for the re-rating centers on agentic inference economics: "The agentic workloads are essentially capacity constrained in terms of throughput by the number of CPUs you have. So as inference demand goes up... and as agentic demand goes up... that means an increase in CPU demand." Arm isn't changing its own published targets yet, but both executives acknowledged the March numbers "may have been conservative." Child added that current supply constraints should ease materially by calendar 2028-2029, with wafer and memory capacity estimated to expand 70% to 100% over that period — implying the real ceiling on Arm's silicon ambitions is a 2026-2027 supply problem, not a demand problem.
Smartphone royalty growth downshifts from 20% to high teens
The offsetting negative in the quarter is smartphones. Memory price inflation is squeezing handset bills of materials, and Child confirmed the damage has broadened beyond the low end: "initially the expectation was just going to affect the lower end of the market, and now we are seeing all parts of the market, even some upper and mid-tier being affected." As a result, Arm now expects full-year royalty growth closer to the high teens rather than the roughly 20% it guided to last quarter, and Q2 royalty growth is guided to just 13%, down from prior expectations.
Arm's insulation from broader smartphone unit declines comes from architecture mix — the shift of customers to Armv9 and compute subsystems continues to lift royalty rates per device even as shipments fall. Haas noted the company expects double-digit royalty growth in smartphones even as the underlying market is projected to decline by double digits. Child was careful to frame this as a timing issue rather than a structural one, expecting a stronger back half of the year and a return to the 20%-plus royalty growth range in fiscal 2028 and beyond as cloud AI overperformance continues to offset the handset drag.
Neoverse data center momentum remains the core growth engine
Away from the new silicon business, Arm's underlying Neoverse data center franchise continues to compound quickly. Cumulative Neoverse core shipments have surpassed 1.5 billion, with the most recent 500 million shipped in just nine months versus six years for the first billion. Data center royalty revenue more than doubled year-over-year for a second consecutive quarter. IDC data cited by Haas shows spending on Arm-based accelerated server platforms has nearly doubled over the past two quarters and has now overtaken x86 platforms — a notable inflection point.
The customer roster reinforcing this shift reads like a who's-who of hyperscale infrastructure: Nvidia has brought its Vera CPU into production, claiming up to 50% higher performance and 2x better energy efficiency than comparable x86 systems; Google's Axion is now the host CPU for its latest TPU systems; AWS plans to deploy "tens of millions" of Graviton5 cores for agentic workloads; Microsoft expanded Azure Cobalt 200 on Neoverse CSS; and Qualcomm is entering the AI data center CPU market with its Arm-based Dragonfly C1000.
Licensing strength and the SoftBank contribution
License and other revenue hit a Q1 record of $574 million, up 23% year-over-year, aided by renewed long-term agreements with major hyperscalers, automotive and robotics customers, and handset OEMs. Of that total, $193 million came from Arm's technology licensing and design services agreement with SoftBank, with Child guiding to a roughly $200 million quarterly run rate from that relationship for the rest of the year. Annualized contract value, the metric Arm prefers investors watch given lumpy deal timing, grew 13% year-over-year — still above the company's long-term royalty growth framework, according to management.
Where the AGI CPU fits competitively
Responding to Barclays' Tom O'Malley, Haas confirmed the Arm AGI CPU is designed to address all three sub-segments of the AI compute buildout — traditional server, head-node AI clusters, and agentic applications — pointing to existing customers spanning Cerebras and OpenAI on the head-node side and Meta, Cloudflare and Oracle on the agentic and general-purpose server side. On the competitive threat from vertically integrated players building their own accelerators alongside custom CPUs, Haas noted Nvidia's NVLink Fusion architecture creates an explicit path for Arm-based CPUs, including third-party designs, to plug into Nvidia's Rubin accelerators, leaving room for Arm's CPU-only model to coexist with proprietary silicon stacks rather than being crowded out by them.
On product roadmap specifics, the current Arm AGI CPU ships with 128 cores, trailing Amazon's 192-core Graviton5, though Haas signaled more cores are coming given that higher core counts improve throughput efficiency for agentic workloads. Silicon revenue will be broken out as a separate reporting line once it crosses 10% of total revenue, which Child expects to occur in fiscal 2028, with initial shipments beginning at the end of this calendar year.