Latest / The Pivot Podcast with Fexingo: How Businesses Change Direction and Survive Disruption / How ARM Forced a New Playbook for Chip Licensing
Transcript
- Lucas: So, ARM Holdings. It's the company that designed the chip architecture inside pretty much every smartphone you have ever owned — iPhones, Samsung Galaxies, the whole Android world. For decades, it ran a licensing model that was so profitable, so elegantly simple, that competitors barely even tried to copy it. Luna: And then something changed. Because I remember reading that ARM's model was almost like a royalty fountain — a few cents per chip, billions of chips a year. Why would you mess with that? Lucas: Exactly. But the fountain started to look less reliable around 2021, when the UK government blocked Nvidia's $40 billion acquisition of ARM. That deal falling through left ARM owned by SoftBank, which needed to cash out. And simultaneously, a serious threat emerged: risc v, an open-source instruction set architecture that anyone can use for free. Luna: Right, risc v. It's not a company you buy chips from — it's a standard, like an open-source alternative to ARM's proprietary blueprints. And it's been gaining traction in data centers and IoT. Lucas: Exactly. So ARM had to pivot — not because its product was bad, but because its licensing model was about to become obsolete. And the pivot is fascinating because it didn't involve a new chip. It involved a new contract. Luna: What did they change? Lucas: Traditionally, ARM licensed its architecture and then let customers like Qualcomm or Apple design their own custom cores based on that architecture. ARM made money on the front-end license and then a small royalty — typically around 1 to 2 percent of the chip's selling price. But under the new model, ARM started offering what's called a 'version-based' license. Instead of a perpetual architecture license, customers now pay for a specific version — say, ARM v9 — and if they want to upgrade, they pay again. Luna: So it's more like a software subscription, but for hardware blueprints. That's a big shift for a company that used to sell a license once and collect royalties for a decade. Lucas: It is. And the numbers bear out the urgency. ARM's IPO in September 2023 valued the company at about $52 billion. But in its prospectus, ARM disclosed that its royalty rates on new cores had effectively increased — for v9, the royalty rate is reportedly around 9 percent, compared to the historical 1 to 2 percent. That's a massive jump. Luna: Nine percent? That's almost an order of magnitude higher. How did customers react? Lucas: Qualcomm, which is one of ARM's biggest licensees, actually sued ARM in 2022 over licensing terms, claiming ARM was trying to restrict how its chips could be used. The case is still ongoing. But the interesting thing is that ARM is betting that its customers have no real alternative — at least for high-performance mobile cores. risc v is good, but it's not yet as power-efficient or as software-compatible as ARM. Luna: So ARM is essentially using its monopoly position to raise prices, while simultaneously trying to make itself indispensable through tighter integration. That's a risky pivot — it could push customers to accelerate their risc v efforts. Lucas: Right. And there's evidence that's already happening. In 2024, a consortium of major tech companies — including Google, Intel, and Qualcomm — launched the risc v Software Ecosystem project to improve software compatibility. That's a direct response to ARM's pricing power. Luna: But ARM isn't just a licensing company anymore, right? They've also started offering 'Compute Subsystems' — essentially pre-validated blocks of IP that include the CPU cores, memory controllers, and interconnects. That's a more integrated product. Lucas: Exactly. That's the second part of the pivot. Instead of just selling a blueprint, ARM now sells a semi-customized building block that reduces design time for customers. It's higher value per license, and it also increases switching costs. If a customer builds a chip around ARM's compute subsystem, they can't easily swap in a risc v core later without redesigning everything. Luna: So the pivot is two-pronged: raise the price on the existing model, and then offer a new, stickier product that makes the old model less relevant. But isn't that a gamble that the new product will be adopted fast enough before customers defect? Lucas: It is. And the timeline is tight. ARM's revenue in fiscal year 2025 was about $3.2 billion, growing roughly 20 percent year-over-year. But the growth is coming from the new licensing model — architecture licensing revenue was actually down. So the pivot is working in the sense that ARM is making more money per chip, but it's also shrinking the pool of customers willing to play ball. Luna: Let me play devil's advocate: Could ARM have simply stayed the course, kept the old model, and relied on the sheer inertia of its ecosystem? Why pivot at all if you're still profitable? Lucas: Two reasons. First, risc v is free and improving every year. If ARM didn't change its model, it risked becoming a commodity — just another instruction set that customers could eventually replace. The pivot is a way to differentiate. Second, SoftBank needed a higher valuation to exit. The old model produced steady but slow growth. The new model is designed to maximize revenue per chip, which boosts margins and stock price. Luna: So the pivot is partly about defending against disruption, partly about financial engineering. But what's the concrete example you'd point to that shows this working? Lucas: I'd point to Apple. Apple has been designing its own custom arm based chips since the M1 in 2020. Under the old model, Apple paid a small royalty per chip. Under the new model, Apple likely negotiated a custom deal that includes both version licensing and compute subsystem elements. We don't know the exact terms, but Apple's continued investment in ARM architecture — they just announced the M4 Ultra in March 2026 — suggests that ARM has successfully locked in its most valuable customer. Luna: But Apple also has the resources to switch to risc v if they wanted. They already design their own cores. What stops them? Lucas: Software compatibility. Apple's entire software ecosystem — iOS, macOS, all their apps — is compiled for ARM. Switching would require either emulation, which incurs performance loss, or a multi-year transition. So ARM's pivot is essentially monetizing that switching cost. It's a smart move, but it only works as long as ARM remains the best option for performance per watt. Luna: And that's where risc v could eventually catch up. There's a startup called Tenstorrent, led by legendary chip architect Jim Keller, that's building risc v based AI accelerators. They've already secured deals with companies like LG and Samsung for future chips. Lucas: Right. And that's the real tension. ARM's pivot buys time and money, but it doesn't eliminate the existential threat. In fact, by raising prices, ARM is giving risc v a stronger value proposition. The question is whether ARM can use its head start to innovate fast enough that customers never feel the need to switch. Luna: So what's the one thing a business leader should take away from ARM's story? Lucas: That a monopoly can still be disrupted, and when it is, the pivot has to be about changing the terms of the relationship — not just the product. ARM didn't invent a new chip. It invented a new contract. And that contract is designed to make customers dependent on a faster treadmill. The question is whether customers will stay on it. Luna: Or build their own treadmill. Lucas: Exactly. And that's the bet ARM is making — that building a new treadmill is harder than paying for the upgrade. For 2026, it looks like the right call. But in semiconductors, five years is an eternity. Luna: And with that, we'll leave it. Thanks, Lucas. Lucas: Thanks, Luna. For our listeners, next time you pick up your phone, remember: the chip inside is a story of a company that had to change everything without changing the product. We'll be back next week with another pivot.