Arm’s quest for royalties has a sting in the tail
Softbank's drive for higher payments would probe the weaknesses of the Arm lock-in.
There comes a time in every technology company’s evolution, assuming it grows old enough, where management turns round to its customers and declares: “Hey, that’s my money!”
This is where Arm, or at least its parent Softbank, has now landed. Given that Arm already collects a royalty on each core used by its chipmaking customers, Softbank has apparently decided according a report by the Financial Times that its best course of action for collecting more money is to demand it from the customers of its customers.
Twenty years is also roughly the birth-to-peak lifecycle of a processor architecture that has made it to the top. IBM’s System/360 mainframe looked triumphant in the mid-1980s, two decades after forging the early computer industry. No-one got fired for buying IBM in those days. Yet, at the time, the Intel x86, which ironically benefited from IBM’s patronage, was chipping away at the foundations to become the pre-eminent computer architecture of the two succeeding decades.
Then along came Arm, which went from being the choice of just one mobile phone maker in the mid-1990s to being the processor to beat by the end of the 2010s. Softbank’s purchase of Arm for £24bn came about 20 years after Nokia incorporated the little-known processor into its products. At that point, Softbank pointed to a future where there would be little Arms everywhere in everything all at once. In this bright future, worth $30bn of 2016 money, the idea was you’d even find them baked into bricks, attached to strain sensors, to gauge how safe a building is.
In 2023, the idea of the internet of things (IoT) is still there but it has taken a definite back seat to what kind of revenue stream Arm might be able to extract from its mainstay business: cores that go into SoCs that go into mobile and consumer devices. Dissatisfied with the few per cent it tends to get on the silicon, Softbank chief Masayoshi Son reportedly wants the device makers to stump up cash for running software compiled for Arm processors. This is likely to be a smaller percentage but one based on a significantly larger purchase price.
The argument for? Well, as is traditional in information technology, it’s because you can. IBM back in the day capitalised on the software lock-in. IBM’s decision to unbundle software from its systems sales was forced on it by the US Department of Justice and made it possible for specialist software companies to grow. But as they focused on the leading architectures of the time because porting software to other machines over time proved increasingly onerous, it became easier for IBM to charge premium prices for its hardware. That lesson was not lost on Intel, nor on many others in IT. Architectural compatibility lets a computer designer build a big, deep moat and milk the proceeds, should it want to.
Today, it seems Arm, or at least Softbank, wants to. Those systems companies have built big businesses that effectively run on Arm. So why not extract some additional rent?
This is where we come to the arguments against. Lock-ins work great when you have a preservable monopoly. In mobile devices, that is true to a large extent. Apple’s iOS is built on Arm. As is Google’s Android, which continues to hold the larger share of that market. But only a couple of months ago, Google said it would add the open-source instruction set architecture (ISA) of RISC-V to its Android ports.
That is not a problem for Arm today or tomorrow. It will likely be a major headache looking further out. Migrating software is a somewhat easier process than it was a few decades ago. Much of the software is implemented using so-called managed-code languages, such as the Java-like Kotlin. Even accelerators needed for graphics and machine learning are moving to interpreted or dynamically compiled languages that do not care all that much about what the target architecture looks like. This is helping to spawn an explosion in RISC-V variants developed by companies who all want to be the next Arm. And it looks like Softbank just gave them a helping hand by encouraging systems companies to look at other options beyond Android running on Arm.
Then you have China, where many of the phones are made. US sanctions mean the country has an even stronger incentive to shift to RISC-V. Implementations of RISC-V are affected by the same sanctions, but as an open-source ISA, Chinese companies are free to make their own versions. That is not an option that is open to them on Arm. Most of those manufacturers have not been sideswiped in the same way as Huawei, as it is blocked from using Android as well, but the incentives for them to make RISC-V work in mobile handsets are getting stronger all the time.
A further problem with the change in Softbank’s position lies in the huge diversity of systems into which Arm cores go. An environment where a processor designer collects a larger share of the final system value is broadly incompatible with the idea of Arm everywhere. How exactly do you price the royalty for systems where the value is all in the service and not the hardware. Does Arm give it up on the basis that 0.1 per cent of free is not a lot of money? Does it try to argue that it should get rent for the rent charged for that service? And how far does it get before the end customer decides that as it controls the software that runs on those devices, it will take the RISC-V option, thank you very much?
Similarly, cloud computing is a market that Arm wants to take off Intel. But trying to extract royalties from server blades is a seriously backwards step when you do not have a monopoly position to exploit in the first place, especially when you consider that the transition from x86 to Arm takes advantage of exactly the same managed-code software that would enable a shift to RISC-V.
Then there is the perverse-consequences problem. Who will the systems companies talk to in order to recover the money they are expected to pay out as an Arm tax? Most likely, the chipmakers who believed pushing Arm would give them access to new markets previously dominated by the likes of Intel. They too would have a strong reason for moving to an architecture where this kind of rent seeking would be far more difficult. Again, that points to RISC-V.
Arm has the key advantage of being in a position where it can offer mature, proven designs and that would protect it for some years. But when an entire markets winds up being incentivised to bring you down? That’s when disruption beckons.
Assuming the licensing change happens ahead of the IPO, it might be enough to make Arm look more attractive to institutional investors expected to take it off Softbank’s hands. They may in turn find their long-term expectations a more than a little optimistic, much like the IoT-powered rationale that drove the $30bn purchase in 2016.
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.