Ahead of Consensus.
Intelligence across tech and capital markets, for investors, policymakers, and business leaders.
9 minute read
Arm Holdings plc closed the first quarter of its 2027 fiscal year with numbers that would flatter almost any technology company on earth. Revenue for the three months ended June 30 reached $1.289 billion, up 22 percent from a year earlier, a first-quarter record for a business that only recently learned what it feels like to clear $1 billion in a single quarter. Royalty income rose 22 percent to $715 million and licensing revenue climbed 23 percent to $574 million, each a first-quarter high, evidence that neither half of Arm’s dual business model is carrying the other. Non-GAAP earnings reached 45 cents a share, comfortably ahead of the 40-cent consensus Wall Street had settled on.
The cash story is arguably more compelling than the income statement. Operating cash flow hit $902 million, and free cash flow surged 343 percent to $665 million, aided by favorable timing in collections and tax payments unlikely to repeat every quarter. Still, the underlying trajectory is unmistakable. Arm closed the period with $3,058 million in cash and cash equivalents, $830 million in short-term investments, and shareholders’ equity of $8,630 million. Few companies growing this quickly can also claim a balance sheet this unencumbered. Fewer still trade at a valuation that already assumes as much.
The headline figures, however, were never really the point of this earnings season. In March, Arm unveiled the AGI CPU, a 136-core data center processor co-developed with Meta and the first piece of production silicon the company has ever sold directly, a genuine departure after three and a half decades spent licensing designs rather than building them. Management had sized the near-term opportunity conservatively, at $1 billion across fiscal 2027 and 2028. Four months later, customer demand has already exceeded $2 billion, double the initial opportunity identified, and Arm says it has secured the manufacturing capacity for the first $1 billion while working to expand supply further.
The economics of that pivot deserve more scrutiny than they have received. Arm projects AGI CPU revenue will reach approximately $15 billion by fiscal 2031, with first-generation chips carrying gross margins in the high 30 percent to low 40 percent range through fiscal 2028, expanding toward 50 percent over time as more manufacturing moves in-house. Set against a licensing business that runs gross margins in the high 90s, that gap is not a rounding error. It is a deliberate trade: some of the industry’s best unit economics, exchanged for a foothold in a market Arm has never had to compete in before, on the belief that owning silicon, not merely the blueprint behind it, is what the AI era ultimately rewards. Management’s own long-range target of roughly $25 billion in total revenue by fiscal 2031 depends on that bet paying off without hollowing out the licensing revenue that still funds most of today’s profit.
Beneath the silicon headline sits a quieter, arguably more consequential trend: the steady migration of hyperscale computing onto Arm’s architecture. Data center royalty revenue more than doubled year over year, and cumulative Neoverse core shipments surpassed 1.5 billion, with the most recent 500 million units shipped in just nine months, a pace that took six years to reach the first time around. The acceleration is not happening in isolation. Nvidia’s Arm-based Vera CPU entered full production during the quarter, while Google, Amazon Web Services, Microsoft, and Qualcomm all expanded their own Arm-based data center platforms, a roster that now spans custom cloud silicon, commercial CPUs, and at least one new entrant built explicitly to contest the AI infrastructure market.
The mechanism behind that momentum is straightforward, even where its implications are not. A data center chip built on Arm’s instruction set, whether licensed directly or embedded inside a hyperscaler’s own design, commands a materially higher royalty than the smartphone silicon that built Arm’s original business. As AI workloads spread outward from the cloud, toward laptops, vehicles, and eventually machines operating with little direct human oversight, that same logic, in which efficiency counts for as much as raw capability, is what Arm is counting on to keep widening its reach well past the server room.
None of it was enough. Shares had already fallen 8.1 percent to close at $224.89 on the day of the report, within a 52-week range running from $100.02 to $452.70, as broader anxiety over AI infrastructure spending weighed on chip stocks ahead of the print. Once results landed after the close, beating estimates on both revenue and profit and carrying second-quarter guidance of $1.38 billion, itself a record at the midpoint, shares still slid nearly 7 percent further in after-hours trading.
Two threads explain the disconnect. Arm signaled that smartphone royalties are expected to soften in the coming quarter, a detail that complicates any narrative in which AI demand alone can carry results indefinitely, given that smartphone royalties have been one of the steadiest sources of growth as Armv9 adoption lifts the average price Arm earns per chip. And Arm walked into the print trading at roughly 81.6 times forward earnings by at least one estimate, a multiple that leaves scant room for anything short of an emphatic surprise. Analysts, notably, have not abandoned the name; the average price target across roughly fifty covering analysts sits above $300, even after a punishing month. What moved the shares, in other words, was not doubt about Arm’s business. It was doubt about whether any single quarter, however strong, could justify what the market has already paid for it.
Two unresolved disputes will color how investors read every quarter Arm reports from here. The Federal Trade Commission opened a formal antitrust investigation in May, examining whether Arm intends to withhold or weaken the architecture licenses that customers including Apple, Qualcomm, and Nvidia depend on, now that Arm sells chips competing directly with some of those same customers. Separately, a US District Court entered final judgment against Arm in its long dispute with Qualcomm over Nuvia’s licensing rights, upholding a 2024 jury verdict in Qualcomm’s favor. Arm has said it intends to appeal, while Qualcomm’s own countersuit against Arm, alleging breach of contract and anticompetitive conduct, remains pending.
Neither case threatens this quarter’s numbers. Both point to a tension Arm cannot simply engineer its way around: the license that makes it indispensable to the entire semiconductor industry is now competing against products built on that very license. How regulators and courts resolve that contradiction, far more than any single quarter’s royalty rate, will determine whether Arm’s move into silicon becomes the durable second engine management envisions, or the complication that eventually slows the business it was never meant to threaten.