TL;DR · 30-second read
The Short Version
Arm has long made its money selling chip blueprints to other companies. Now it is selling a chip of its own, built for the data centers behind artificial intelligence.
It says that chip could bring in up to $2 billion by March 2028. The first commercial shipments come early next year, so reaching that figure means earning almost all of it in about 15 months.
Arm says orders are coming in and credits extra manufacturing supply for its growing confidence. The real test is whether it can get enough chips made in time.
Arm Holdings reported revenue of $1.44 billion for its fiscal second quarter ended September 30, 2026, up 27% from a year earlier, according to a shareholder letter furnished to the SEC on a Form 6-K on November 4. License and other revenue grew 50% to $773 million, royalty revenue grew 8% to $670 million, and data center royalties doubled.
BigGo Finance reported that Arm raised its AI chip revenue target to $2 billion and that the licensing figure was a record. Arm’s letter says its confidence in achieving “up to $2 billion” in Arm AGI CPU revenue over fiscal 2027 and 2028 has increased over the last 90 days, with first commercial deliveries on track for fiscal Q4 2027.
Executive Summary
Arm now describes itself as running two businesses on one technology base. The first is its established IP business, which licenses chip designs to other companies and collects royalties when those designs ship. The second is a new silicon business, in which Arm sells its own data center processor, the Arm AGI CPU, directly to customers.
The quarter’s reported growth came from the first business. Licensing and royalties together make up the $1.44 billion in revenue, while the AGI CPU has yet to reach first commercial delivery. Arm’s up-to-$2 billion figure is therefore a forward statement about a product that begins shipping in the quarter ending March 2027.
That timing matters for anyone planning around the chip. Arm says purchase orders from multiple customers extend through calendar 2027, and it credits additional supply and customer mix for its increased confidence. Together, those statements point to manufacturing capacity in fiscal 2028 as the main operational test of the goal.
Five Quarters to Reach a Two-Year Ceiling
Arm’s figure covers two fiscal years, or eight quarters, but its first commercial deliveries are scheduled for the eighth-to-last quarter’s opposite end: fiscal Q4 2027, which runs from January through March 2027. That leaves at most five quarters of commercial shipments in the window, four of them in fiscal 2028. Reaching the full amount would average about $400 million a quarter across those five quarters. That is our arithmetic rather than Arm’s schedule, and shipments would be expected to build over time rather than arrive evenly. Still, $400 million equals about 28% of the company’s total revenue in the latest quarter.
The wording also deserves attention. “Up to” makes $2 billion a ceiling, and the sentence describes increased confidence rather than stating a new, higher number. Arm attributes that confidence to “additional supply and customer mix” and adds that it remains “focused on expanding capacity to meet growing customer demand.” Arm describes demand as extremely strong and points to purchase orders running through calendar 2027. On the company’s own account, then, the factor that has moved most recently is the ability to deliver.
This does not prove that demand is settled, because the statement about purchase orders carries no dollar value. It does show who is most exposed to the supply question: customers holding those orders, Oracle Cloud Infrastructure with its plan to offer AGI CPU instances in 2027, and investors modeling a new revenue line. Their planning depends on how quickly capacity arrives during fiscal 2028.
Licensing Is the Larger Number, and It Comes From the Established Business
The 50% licensing jump is the quarter’s most striking growth rate, but it does not measure chip sales. License and other revenue reflects what customers pay for access to Arm technology, and at $773 million it made up roughly 54% of the $1.44 billion total. Arm links the increase to customers managing growing system complexity and tighter power limits as they combine CPUs with GPUs and NPUs, the specialized processors used for graphics and AI work.
Arm calls licensing “an important leading indicator of future royalty growth,” because a design licensed today can earn royalties once it reaches production. The latest royalty figures show why that future conversion matters. Total royalties rose 8% even as data center royalties doubled, which means the rest of the royalty base grew more slowly than 8% in aggregate and may have declined. Data center strength is real on Arm’s figures, but it has not yet lifted total royalties at anything close to the pace of licensing.
A single quarter’s licensing result is a signal rather than a run rate, and the timing of deals can move it. Arm’s 6-K also separates license and royalty revenue between related and nonrelated parties, a split worth reading alongside the 50% figure. The move beyond IP into direct AI silicon spending is under way commercially, with orders in hand, but it does not yet account for the growth in these results.
A New Seller in a Market Its Licensees Already Serve
Arm’s letter says Arm-based server share at the top hyperscalers, the largest cloud operators, rose from 15% in calendar 2024 to more than 50% in calendar 2025, and it expects more than 60% by the end of calendar 2026. It credits ramps across Google Axion, AWS Graviton, Microsoft Cobalt and NVIDIA platforms. Those are partners’ chips, which earn Arm licensing and royalty income. The AGI CPU adds Arm’s own product to the same broad market.
Arm presents the two as complementary, saying agentic AI, in which software agents run tasks persistently, raises demand for both partner processors and its own. Oracle offers a concrete test. NVIDIA’s Arm-based Vera CPU is generally available on Oracle Cloud Infrastructure, and Oracle plans to offer AGI CPU instances in 2027, which Arm describes as bringing Arm-designed and partner-designed processors together on a common platform.
The commercial question is where the AGI CPU’s demand comes from. Sales to operators that would not otherwise build a custom chip would expand Arm’s market. Sales that displace a partner’s Arm-based processor would swap a royalty for a chip sale, with different economics. The customer mix behind the orders will help show which pattern is emerging.
The Power Claim Asks a Rack Benchmark to Carry a Lot
Arm’s letter cites a Morgan Stanley projection of a 38-gigawatt global power shortfall through 2028, then points to Signal65 testing that showed the AGI CPU delivering approximately twice the performance per rack of AMD’s EPYC 9755. It says this “could reduce the projected global power shortfall by half.” The first part is a test result against one competing processor. The second is a much larger extrapolation.
Performance per rack measures how much computing fits in a rack, which is not the same as computing per unit of electricity. Getting from that benchmark to a halved global shortfall would also require assumptions about how much of the affected capacity runs CPU workloads, how closely real workloads match the test, and how widely the chip is deployed. That last assumption sits uneasily beside a product whose commercial deliveries have not started and whose stated revenue goal is capped at $2 billion through fiscal 2028.
Power efficiency is a legitimate purchasing argument, and data center operators working within fixed electrical capacity have good reason to weigh it. Those operators are best served by evaluating the rack-level result against their own workloads and power budgets, while treating the global figure as an illustration of what could happen rather than a forecast.
Background
Arm Holdings plc, based in Cambridge, England and listed on Nasdaq, built its business on licensing processor technology and collecting royalties on chips that use it. The company says its platform now spans core IP, advanced compute subsystems and purpose-built silicon, and it reports an ecosystem of more than 30 million developers. Its fiscal year ends March 31, so fiscal 2027 closes in March 2027 and fiscal 2028 in March 2028.
In data centers, Arm technology underpins processors that large cloud and AI companies develop for their own platforms, including AWS Graviton, Google Axion, Microsoft Cobalt and NVIDIA’s Vera. The Arm AGI CPU, launched in March, marks the company’s move into delivering Arm-designed silicon directly to customers alongside that partner business. Source: [Arm Holdings Q2 2027 Earnings Call] Arm Raises AI Chip Revenue Target to $2 Billion as Licensing Surges 50% to a Record $773 Million, BigGo Finance’s report on Arm’s fiscal second-quarter earnings call. Primary sources: Arm Holdings Form 6-K filed 2026-11-04 (quarterly report for the period ended September 30, 2026); Arm Holdings Form 6-K filed 2026-11-04 (November 2026 report and exhibit index); Exhibit 99.1: Arm Holdings plc Reports Results for the Second Quarter of the Fiscal Year Ending 2027; Exhibit 99.2: Arm Holdings shareholder letter for the second quarter of fiscal 2027.Sources