Cambridge-based chip designer Arm has run one of the most successful business models in tech. It draws blueprints for chips for manufacturers and collects the royalties, whilst never making a chip itself.
The company licenses its instruction set architectures (ISAs) and processor core designs to companies, including industry heavyweights such as Apple, Qualcomm, NVIDIA, Samsung, and MediaTek. These companies reportedly pay Arm between $0.10 and $2.00 per chip to use its designs. This arrangement has made Arm indispensable without making it a threat to the companies that are funding it. That was until March 2026, when Arm announced the AGI CPU – a 136-core data centre processor co-developed with Meta, built on TSMC’s 3nm process, and capable of running at 300 watts.
Within eight weeks of the announcement, the US Federal Trade Commission (FTC) opened a formal antitrust investigation into whether Arm intends to use its control over the foundational blueprint of modern computing to tip the market in favour of its own processors. The FTC is specifically concerned that Arm will refuse or degrade the quality of CPU blueprint licences to third-party chip designers while competing against those same designers in the market for chips. The FTC has issued a formal document preservation demand, which means it is actively gathering evidence rather than monitoring from a distance.
Arm has said that its AGI CPU is an extension of its platform strategy, which is designed to give partners “the broadest set of options to build on Arm” rather than to replace them. It has contributed reference server designs to the Open Compute Project and said that its existing Neoverse CSS roadmap continues in parallel, so that “all Arm data centre customers move forward together.”
It is worth noting that the FTC isn’t accusing Arm of wrongdoing; rather, it is saying that Arm is now in a position where doing something wrong would be an almost rational business decision. However, its investigation suggests that, whatever Arm’s stated intent, the structural conflict of interest now exists regardless.
What is an ISA?
An ISA is the language a chip speaks. It defines which operations a processor can perform and how software communicates with hardware. Every application, operating system, and AI model ultimately runs on top of an ISA. Choosing one is therefore a critical decision in chip design, because the software ecosystem builds itself around it over time. If companies ever need to unwind that relationship, it can prove time-consuming and expensive. Apple’s migration from Intel’s x86 to its own Arm-based M-series chips, for example, took years and billions of dollars.
There are three top ISA contenders at the moment.
x86:
- Owned by Intel and AMD
- Dominates personal computers and traditional data centres
Arm:
- Owned by Arm
- Dominates smartphones, tablets, and increasingly AI infrastructure
- Originated as an academic research project at UC Berkeley
- Dominates IoT and consumer electronics
Unlike Arm and x86 architectures, RISC-V is entirely open-source and royalty-free, meaning that anyone can use it, customise it, build chips around it, and extend it without paying licensing fees to any company. The RISC-V International standards body, which governs the architecture, relocated to Switzerland in a deliberate move to place the standard outside the reach of any single government’s export controls or trade policy, positioning it as a shared global resource.
The rise of RISC-V
By the end of 2025, an estimated 20 billion RISC-V cores were in operation globally, with market revenues projected to grow at a compound annual rate of 34% toward $10.77 billion by 2030. The IoT market accounts for roughly 30% of that installed base. By January 2026, RISC-V had reached approximately 25% global processor market penetration. In 2022, the European Union committed €270 million to indigenous RISC-V chip development. China has integrated it into its national semiconductor strategy as the primary vehicle for reducing dependence on Western-controlled IP.
Arm has admitted that RISC-V is now one of its biggest competitors. Compounding that pressure, in 2025, Qualcomm filed antitrust complaints against Arm with the FTC, the European Commission, and South Korea’s Korea Fair Trade Commission, accusing the company of restricting licence access and withholding technology. South Korean regulators then conducted unannounced office inspections of Arm. Separately, Qualcomm had previously won a breach-of-contract case against Arm over processor designs acquired through Qualcomm’s purchase of startup Nuvia. In December 2025, Qualcomm also acquired Ventana Micro Systems, a RISC-V specialist.
With the launch of its own chip, Arm has reportedly attracted customer demand exceeding $20 billion within six weeks – constrained only by TSMC wafer allocation rather than market appetite – with OpenAI, Cerebras, Cloudflare, and SK Telecom among the launch partners. That level of demand is notable, but it has also prompted questions about whether Arm has concluded the data centre opportunity is large enough to justify competing directly with the customers funding its existence.
What are Arm’s options?
Arm has not yet done anything wrong, and it is not without options. Its Neoverse server architecture has captured roughly 50% of hyperscaler compute, and it still dominates the smartphone market. Its software toolchain ecosystem puts RISC-V at a competitive disadvantage. However, if Arm continues to pursue chip sales, it risks deepening the conflict-of-interest concern driving the FTC investigation. If it retreats from chip sales, it abandons the data centre revenue rationale that justified the AGI CPU. And if it considers open-sourcing elements of its ISA, or adopting a tiered model that removes royalty friction for commodity applications while preserving premium licensing for high-performance cores, it would require Arm to fundamentally reimagine the business model it has operated for more than 30 years.
The investigation also raises a broader question of whether the foundations of modern computing should be proprietary at all. The reasons companies choose one architecture over another are specific to their circumstances, but RISC-V holds one structural advantage: it does not need to win a performance benchmark against Arm. It only needs to be good enough, available to everyone, and free from legal and financial friction. Whether that threshold has been crossed – or is close to being crossed – is a question the market is actively asking. What is clear, however, is that Arm must now think carefully about its next move.
