Arm's licensing model is moving from quiet infrastructure to antitrust flashpoint as AI chip demand makes its technology more valuable and more contested.
Arm Holdings is facing a U.S. antitrust test over how it licenses the chip technology that sits underneath much of the modern computing market. That may sound like a narrow fight over contract language, but it is really about who gets to build the next generation of AI hardware, and on what terms.
For decades, Arm's strength was that it did not usually look like a rival. It designed processor architecture and intellectual property, then licensed that technology to companies that made chips for phones, data centers, cars and connected devices. Apple, Qualcomm, Nvidia, Amazon and Microsoft could all build around Arm because Arm was the shared layer underneath them. The more neutral that layer appeared, the more valuable the ecosystem became.
That is why the scrutiny matters. As Bloomberg recently reported, Qualcomm has taken its licensing fight with Arm to antitrust agencies in the United States, Europe and South Korea, accusing the company of restricting access to core technology and moving away from the open model that helped it become so deeply embedded in the industry. Arm has denied wrongdoing and has framed the dispute as a commercial fight with Qualcomm.
The tension is not just legal. It is commercial. Arm wants to capture more of the value created by its designs, especially as AI pushes demand for efficient processors in cloud and edge systems. In February, Arm said quarterly revenue rose 26 percent year over year to $1.24 billion, with royalty revenue up 27 percent to $737 million and license and other revenue up 25 percent to $505 million. Those numbers show why Arm has more leverage now. They also show why customers are watching every licensing change closely.
The awkward part is that Arm's biggest customers increasingly look like the companies it may compete with. TechCrunch reported in March that Arm introduced its first in-house chip, the Arm AGI CPU, a production-ready processor aimed at AI data centers, with Meta as lead partner and launch customers including OpenAI, Cerebras and Cloudflare. That is a major change for a company long known for selling blueprints rather than finished silicon.
There is nothing automatically wrong with Arm moving up the stack. Companies evolve. Markets change. But the antitrust question becomes sharper when a supplier controls technology that customers need, then begins selling products closer to what those customers are building themselves. A licensing negotiation that once looked like a standard commercial discussion can start to look like a gatekeeper deciding which rivals receive what they need to compete.
The U.S. Federal Trade Commission has already shown how it thinks about Arm's position. In its 2021 challenge to Nvidia's proposed $40 billion acquisition of Arm, the agency said Arm technology was a critical input for competition in several chip markets, including Arm-based CPUs for cloud computing providers. The FTC also said licensees shared sensitive product information with Arm because it was seen as a neutral partner, not a rival chipmaker. Nvidia later abandoned the deal in 2022.
That history gives regulators a ready-made frame. If Arm was too important to be controlled by Nvidia because rivals depended on access to its technology, regulators may now ask whether Arm's own business shift creates a similar risk. The concern is not only price. It is whether access, timing, technical support or product roadmaps can be used to tilt the market.
Qualcomm made the fight harder to ignore
Qualcomm's dispute with Arm gave regulators a concrete example to examine. The fight began after Qualcomm acquired Nuvia in 2021, bringing in custom CPU technology that later became central to Qualcomm's push into higher-performance computing. Arm sued Qualcomm in 2022, arguing over license rights tied to Nuvia. Qualcomm won a major ruling in Delaware in late 2025, with the court rejecting Arm's remaining claims and confirming that Qualcomm and Nuvia had not breached the relevant architecture license, according to Qualcomm's public statement.
That court win did not end the broader argument. Qualcomm has separately accused Arm of conduct aimed at hindering innovation and favoring Arm's own products over long-standing partners. South Korea's Fair Trade Commission inspected Arm's Seoul office in November 2025 as part of scrutiny over licensing practices, with The Korea Times reporting that the investigation followed Qualcomm's complaint to regulators in Korea, the United States, Europe and elsewhere.
For AI chip startups, the stakes are practical. Most young semiconductor companies do not have the time or capital to rebuild the full software and processor ecosystem from scratch. Arm offers a path into markets where power efficiency, developer support and compatibility matter. If that path becomes more expensive, more restrictive or less predictable, smaller companies could be pushed toward fewer options before they ever reach scale.
That is where RISC-V enters the conversation. The open instruction set has been building support as companies look for more control over their own roadmaps. It is not a simple Arm replacement, especially in markets where software maturity and proven performance matter. But antitrust pressure on Arm could make RISC-V more attractive to hyperscalers, startups and governments that dislike being dependent on one licensing chokepoint.
The market should not expect a quick reset. Antitrust reviews move slowly, and licensing models in semiconductors are technically dense. Arm will argue that it invests heavily in technology and should be paid fairly for that work. Its customers will argue that the industry depends on access that remains neutral enough to support competition.
The important thing to watch is whether regulators treat Arm as just another chip IP company or as infrastructure for the AI hardware economy. If they choose the second view, the outcome could shape not only Arm's contracts, but the bargaining power of every company trying to build silicon for the next phase of computing.
Also read: Pixal3D makes image to 3D feel closer to a working pipeline • Americans are turning against AI data centers near home • Pope Leo makes AI warfare a governance test for startups
