By Mike Feibus
What’s at stake?
Arm filed late August a lawsuit against Qualcomm Inc. and Nuvia, Inc. for breach of license agreements and trademark Infringement. But the lawsuit might not be as straightforward as it first appeared. At stake here is a new Arm processor developed by Qualcomm for laptops, which leverages technology innovated by Nuvia, now owned by Qualcomm. So, here’s the billion-dollar question: why is Arm — with the prospect of collecting royalties for each of Qaulcomm’s performance-packed processors — suing to prevent Qualcomm from making the chips?
After a couple months of head-scratching speculation over what motivated Arm to sue one of its best customers, a recent court filing from defendant Qualcomm adds some long-awaited context. Specifically, it’s now apparent that Arm is suing Qualcomm over its recent $1.4 billion Nuvia acquisition because it’s going to cost Arm a lot of money.
Arm wants to keep that from happening, of course. But if what Qualcomm asserts in the filing is true, there isn’t anything Arm can do about it except maybe delay the inevitable, as it’s apparently trying to do. Because, according to Qualcomm, what makes Nuvia so attractive are proprietary architectural techniques that can be applied to any processor instruction set architecture, or ISA, not just Arm.
Of course, there’s never a good time for a revenue shortfall. But this is a particularly challenging period for Arm. It is dusting itself off after the failed Nvidia buyout, a possible lifeline that formally collapsed in early February because regulators viewed it as a threat to the ISA’s posture of neutrality. The company has been chasing new sources of licensing revenue – both in new areas like servers as well as from higher per-unit royalties in existing markets – albeit with limited success.
At the same time, the 32-year-old instruction set licensing house must take care not to lean too hard on existing customers, as many of them have an eye on RISC V, an emerging processor ISA built around a 21st-century open-source development model, as a potential long-run alternative.
Cataclysmic change in landscape
Nuvia was founded by processor engineers from Apple and Google, with the aim of building amped-up server processors based on Arm. At the time, the startup touted better peak performance at lower power than the latest offerings from Intel and AMD.
Qualcomm isn’t much interested in the server market. But it had an eye toward building its own Arm CPUs rather than simply dropping stock Arm processors into its SoC’s as it has been doing. And it seemed that Nuvia’s crack design team would help catapult its designs ahead of the pack. Hence, the acquisition.
Arm offers two types of licenses: technology license agreements, or TLAs, and architectural license agreements, or ALAs. Chip makers with TLAs incorporate stock Arm processors into their products, while those with ALAs design their own based on the instruction set. TLAs are more expensive because Arm is doing more of the work.
Qualcomm has both types of agreements in place with Arm, though the lion’s share of its products today include stock Arm processors covered by its TLA. If the wireless pioneer were to build its own processors under its ALA – as it now plans – then it would pay much less per chip to Arm.
The Nuvia coefficient
For its part, Nuvia was planning to build custom server chips under its own ALA. But that wasn’t the least bit threatening to Arm, for a couple reasons. First, Nuvia was applying its proprietary design techniques to Arm-based processors targeting the server market, which is far smaller on a unit basis than Arm’s traditional markets. Plus, the startup was small, with no product revenue to speak of. So every server chip Nuvia sold would have been additive. And under the terms of its ALA, it had to pay Arm far more per unit than Qualcomm.
But in the hands of Qualcomm, Nuvia is a clear and present danger to Arm’s bottom line. The startup’s technology will be used to transform the foundation of Qualcomm’s $37.7 billion chip business from stock Arm processors to custom versions. That could help boost efficiency and slash costs – with a chunk of the savings coming in the form of lower royalty payments to Arm.
That could have ripple effects beyond Qualcomm as well, should the chipmaker manage to take share from other Arm TLA licensees with its custom SoCs.
On the surface, the Arm-Qualcomm suit has all the earmarks of a plain-vanilla contract dispute about money. Qualcomm is on the verge of paying far less per chip to Arm than it has. And that doesn’t sit well with Arm.
But in the broader context of Arm’s standing as the keeper of a foundational architecture that is pivotal to myriad markets, a row like this was inevitable.
For years now, licensees like Broadcom, MediaTek, Nvidia, NXP and Qualcomm have been content to incorporate stock Arm processors into their products, and differentiate with their own companion building blocks, like DSP, AI, networking and image processing. But serial underinvestment on the part of Arm has been testing licensees’ resolve.
A failing of the last-century licensing model for processor-level compatibility is that it rewards underinvestment in the short term. Once committed to the ISA, licensees have little recourse but to license and build compatible products. So the licensor will make more because it will collect the same in fees but spend less on furthering the technology.
Prolonged underinvestment can come back to bite the ISA gatekeeper in the long term, however. Indeed, Arm’s underinvestment already is giving RISC-V an entrée into its markets that the open-source ISA may one day exploit.
In the meantime, Apple’s success in building its own custom Arm processors no doubt convinced Qualcomm to reassess its own make-or-buy decisions. And the attempted acquisition of Arm by Nvidia, a licensee with divergent interests and priorities than Qualcomm, also had to stoke a reexamination of the company’s priorities.
In that sense, Arm has much more to be concerned about than its royalty revenue stream from Qualcomm. In a very real sense, it is the IP house that is on trial here, not Qualcomm. And just by filing suit against one of its biggest customers, it may already have lost.
Mike Feibus is president and principal analyst of FeibusTech, an industry analyst firm located in Scottsdale, Arizona. He is a well-recognized and oft-quoted expert on connected health and fitness, AI, cloud, smart home, connected car, augmented reality/virtual reality, privacy and security. Feibus is also a regular technology columnist for USA TODAY and CIO Magazine. He is a member of the Editorial Advisory Board at The Ojo-Yoshida Report.