Getting Chippy: ARM Versus Qualcomm (NASDAQ:QCOM)

When You Sue One Of Your Best Customers

ARM, which is currently owned by Softbank, sits at the center of the chipmaking universe, though they make no chips for sale. Their customers are Qualcomm (NASDAQ:QCOM), Apple (AAPL), Google (GOOGL), Nvidia (NVDA), Samsung, Intel (INTC), AMD (AMD), Amazon (AMZN)… well, I could keep going, but I think you get the picture. Cumulatively, 230 billion chips have shipped with ARM IP in them.

Qualcomm is one of ARM’s best customers. ARM IP is the foundation of their most important products: their Snapdragon systems-on-a-chip [SoC] for smartphone, PC, IoT and automotive.

It’s always odd when a company feels compelled to sue one of their best customers, but here we are:

The timing is also sort of head-twisting. ARM is soon to IPO, though we await final details. Qualcomm is about to debut their new line of Snapdragons, which is at the heart of the suit. Neither company can afford for this to go on for long.

But make no mistake: ARM’s lawyers took a maximalist position on the licensing issue at the heart of the suit, and threw in a trademark infringement claim for good measure. Qualcomm is under new leadership since June 2021, but their previous leadership was well-known for playing hardball in these kinds of matters. This could spiral badly for both companies if they are not careful.

But ARM likely believes that Qualcomm has the bigger incentive to settle this quickly, and they are probably right.

Who Is ARM?

Qualcomm is more well known. If you have an Android smartphone, there’s a good chance that a Qualcomm Snapdragon SoC underlies it. If you have a 5G phone of any sort, it is almost certain that Qualcomm IP is part of that. But the two most important parts of Snapdragon SoCs, the CPU and GPU, are mostly designed by ARM, and licensed by Qualcomm just like every other smartphone SoC designer not named Apple. Qualcomm’s smartphone SoCs are currently differentiated by their leading 5G radio, which is on the SoC. They would like to add to that, and that is at the core of the suit.

ARM was originally a joint venture of Apple, LSI and Acorn (ARM originally stood for “Acorn RISC Machine”). Apple needed a low power chip to run the Newton PDA, which debuted in 1992. Now, ARM chips run just about every smartphone and tablet in the world, every new Mac, much of IoT, and are increasingly seen in the data center.

ARM does not make chips. They make IP for others to license and make chips from. The core of ARM IP is the instruction set architecture. This is the base level at which hardware and software communicate. On top of that they build designs for low power consumption CPU and GPU cores, and reference designs for chips. Their niche since the smartphone revolution really took off is being the neutral Switzerland of the tech industry, licensing the same IP to everyone. I don’t ever recall them suing a big customer like this.

Customers like this because it creates an ecosystem that can be built upon, with a lot of great network effects. The biggest example is every Apple product is now based on the ARM instruction set, except the low volume Mac Pro. But Apple sells their own fully customized cores, something Qualcomm would like to do, starting next year. ARM is asking them to destroy all that IP.

A Timeline

This is just events relevant to the case.

2010-2015: Apple had already moved to better custom cores for the iPhone SoC based around the ARM instruction set, and Qualcomm was worried that they were falling behind (they were). They acquired the second kind of license from ARM, the architecture license. This allowed them to build their own designs like Apple for their Kyro cores for smartphone SoCs, which debuted in 2015.

2018: This did not go well, and in 2018 they semi-reversed course. Now Kyro cores are not fully customized like Apple’s, but rather “semi-customized,” or modified stock ARM cores. The CPU cores are not much differentiated from a competitor like MediaTek.

2019-2020: After taking Apple’s chip design from “Hey, these things aren’t bad!” to the best in the world, Apple’s chip design lead Gerard Williams left the company at the end of 2019. He wanted to make an ARM data center chip with custom cores, and Apple has no use for that. Williams recruited two old Apple colleagues who were then at Google to join him in founding Nuvia. They were working on the data center chip that Apple had no interest in. They also had an architecture license, but theirs was for a data center chip. Note the past tense.

2021: Qualcomm buys Nuvia and makes Williams SVP Engineering. They will get to that data center chip, but the first order of business is reinvigorating those Snapdragons which had fallen behind Apple’s badly in the intervening years.

I was pretty excited when that happened, and I immediately fired off an article titled “Qualcomm’s Acquisition Of NUVIA Is A Huge Move.” I still think so. It got pretty good readership here at Seeking Alpha, but it turns out some of those readers were ARM lawyers:

They quoted that article as evidence that Qualcomm needed help to get back up to Apple’s level. They did and still do. Qualcomm is sampling Snapdragons built around Nuvia IP this year and we will start seeing them in the wild in 2023. I strongly suspect that they have a deal with Microsoft (MSFT) for a Surface laptop sporting one of those Nuvia-Snapdragons for holiday season 2023. It may even get fabricated on TSM (TSM) 3 nanometer, their latest node.

In June 2021, Cristiano Amon became CEO of Qualcomm. Amon is an electrical engineer who was previously President. In that role, he was in charge of M&A technical integration, among other things.

2022: The Nvidia-ARM sale fell apart for all the reasons I predicted when it was announced. Softbank announced intentions to IPO ARM instead. Longtime CEO Simon Segars stepped down, replaced by Rene Haas, who was President of the IP licensing segment.

So that brings us up to the complaint.

The Complaint

This is pretty twisty, but let me try and sum up ARM’s argument as quickly as I can.

• As I alluded to above, ARM has two types of licenses. Most licensees have a technology license, which is the core and chip designs. But anyone can do this, and it is hard to differentiate using the same stock designs as everyone else.
• Qualcomm also has the other type of license, the architecture license. Far fewer customers have one of these. The terms of Qualcomm’s license were for Qualcomm to make custom smartphone CPU cores based off the ARM architecture like Apple does. ARM lawyers got pretty snarky about that part. The whole tone of the recitation was “Qualcomm wasn’t good enough to do this on their own, so they bought a startup who could.” For the record, it’s true.
• Nuvia had an architecture license to make custom data center cores. In ARM’s telling, the price was low because it was a startup headed into an uncertain market.
• As ARM claims is a standard contract term, these licenses are not transferable, and have to be renegotiated if the IP developed off them is re-assigned to a new entity, like in a merger. That clause is there to protect them from this exact situation, where they give better terms to a startup headed into an uncertain market than they would to Qualcomm.
• So the important things are that Nuvia’s license was for a custom data center chip, and that it is non-transferable without a new agreement.
• ARM says they told Qualcomm about this, and that Qualcomm’s smartphone architecture license does not also cover Nuvia’s IP.
• ARM terminated the Nuvia license on March 1, and informed Qualcomm that they had to destroy the Nuvia IP. It is their firm position that Qualcomm is not entitled to continue to develop anything that started at Nuvia absent a new license. This would effectively render the IP useless if a court agrees.
• Qualcomm’s didn’t do that and that’s how we got the lawsuit. ARM still wants to see the Nuvia IP destroyed, and some cash damages for trademark infringement while Qualcomm was breaching contracts.
• In May, Qualcomm sought to qualify the Nuvia-Snapdragons as ARM-compliant. ARM’s response was this lawsuit.

From the proposed remedies:

Upon termination, the Nuvia ALA [architecture license agreement] requires Nuvia to cease using and destroy any technology developed under the Nuvia ALA, as well as cease using Arm’s trademarks in connection with any technology developed under the Nuvia ALA.

Qualcomm shares Nuvia’s obligations under the Nuvia ALA in its capacity as Nuvia’s acquirer, and thus Qualcomm is likewise subject to the requirements of the Nuvia licenses’ termination provisions.

Throwing in the trademark infringement claim underlines the cost to Qualcomm if they proceed.

What Happens Now?

I wish I could tell you. ARM suing a big customer is sort of like Finland joining NATO, but that also happened in 2022. Let’s look at the moving parts.

• Both companies have new leadership. In their previous roles, both CEOs were integral to the core issues in this case.
• Softbank would like to IPO ARM soon. Already as overhang they have the seemingly endless saga of Allen Wu and the ARM China JV, which owns ARM’s Chinese licensing business. They have enough problems right now. They shouldn’t want this to still be going on when they IPO.
• But the assignment clause in ARM architecture licenses is very important to them, and is there to protect them from this exact circumstance. Despite the looming IPO, they need to let everyone know that the clause means something, and that they will go to court to enforce it if necessary, regardless of how terrible the timing is.
• Qualcomm is sampling these Nuvia-Snapdragons this year. They want them in products next year. They need to have them certified by ARM, and not open customers up to their own lawsuits.
• But Qualcomm’s lawyers are pretty notorious in these matters. Typically, they have been the litigant, and they have been very aggressive in the past in enforcing the terms of their licensing agreements. A friend joked that because I was quoted in the complaint, there is a very small but nonzero chance that Qualcomm deposes me over this. It’s funny because it’s true.

We have not yet heard Qualcomm’s response, but ARM started with a maximalist position, and I expect Qualcomm to do the same. Then negotiations begin. Maybe they are already negotiating behind the scenes. The danger here is that this spirals out of control, and actually winds up in court. That will put a cloud over the ARM IPO. I don’t even know what it means for Qualcomm, but I imagine a judge will enjoin them from commercializing these chips while the trial progresses. The uncertainty will deter customers, and this includes smartphone customers, their core business.

That puts a huge question mark on my long-term thesis for Qualcomm. We are in the middle of a very slow technology revolution in three parts:

1. Intel whiffs on iPhone, and ARM chips wind up in billions of cell phones instead of x86 chips. This is water under the bridge.
2. Powerful but power-efficient ARM chips begin to show up in PCs and expand design possibilities, especially in the largest segment, laptops. Apple used to be a 10% customer at Intel according to Bloomberg, and now all Macs but the lowest volume Mac Pro sport Apple Silicon based on the ARM instruction set. I am typing this article on one. The Windows part is next.
3. ARM chips start showing up in the data center, displacing x86 chips from Intel and AMD. For hyperscalers, ARM chips let them create a much denser data center and still lower Opex, and they bring other advantages for some types of loads like game servers. Amazon rolls their own, the Graviton3. Microsoft, Google and Oracle (ORCL) all use servers based on private Ampere’s Altra platforms in their cloud offerings. The economic benefit is too big to ignore.

So Qualcomm’s piece here is bullet 1, which is water under the bridge, and bullet 2, which is just beginning. Apple has given everyone the roadmap to making a great PC based around ARM chips. They run incredibly smoothly with battery life unmatched by comparable Windows laptops. Qualcomm and Microsoft have been working on this together for several years now, but their previous joint effort, the Surface X, was a failure. Both companies did not deliver — Qualcomm’s chip was too slow, and ARM Office was not ready yet.

But that was a few years ago, and now things have changed. Microsoft has finished their job, and both companies are counting on the Nuvia-Snapdragons to be Qualcomm’s piece. Like I said, I’m pretty sure they would like to have a Surface laptop for sale in 2023 that copies what Apple has done with the M1 Macs. This puts that in doubt.

AMD just reorganized their reporting segments into something much more useful, so we only have a single quarter of combined comparable data. But between AMD and Intel in Q2, they booked $9.8 billion in PC chip revenue, which annualizes out to $39 billion. Just as Apple walked away with 10% of Intel’s revenue and split it with TSM, Qualcomm can do the same, with an even larger chunk. Plus there are also new end markets in IoT and automotive. Qualcomm had $42 billion in revenue in the TTM.

Moreover, there are recent rumors that Qualcomm also wants to move into bullet 3, the data center. I always thought that had to be part of the pitch to Nuvia during merger negotiations. Gerard Williams left a very good job running the best chip design unit in the world and accumulating Apple shares to make a data center chip. I have to believe that part of the deal was that once they fixed Qualcomm’s problem — reinvigorating Snapdragon — they would turn to his data center chip. Williams is probably their most important employee right now, and has already demonstrated that he will leave if they don’t build a data center chip.

In Q2, Intel and AMD had $6.1 billion of data center chip revenue, a $25 billion annualized rate.

So ARM probably thinks that Qualcomm has more to lose if this drags on, and the bigger incentive to settle quickly, and on ARM’s terms. I agree, but it’s a dangerous game of chicken.

But in any event, this throws a big wrench into the whole long term Qualcomm thesis until it is cleared up. I sold all my semi stocks in early April. I have not been a Qualcomm shareholder since then, and I await better market conditions before I consider them again. But if they clear this up, and good market conditions return, they have a ton of runway.