Blog

Dec 12, 2023

Arm’s IPO will tell us how much AI hype matters

By Elizabeth Lopatto, a reporter who writes about tech, money, and human behavior. She joined The Verge in 2014 as science editor. Previously, she was a reporter at Bloomberg. If you buy something

By Elizabeth Lopatto, a reporter who writes about tech, money, and human behavior. She joined The Verge in 2014 as science editor. Previously, she was a reporter at Bloomberg.

If you buy something from a Verge link, Vox Media may earn a commission. See our ethics statement.

I love a circus, and the Arm IPO is kinda looking like a P.T. Barnum production. As our ringmaster, we have SoftBank, the current majority owner of Arm. On the high wire, with no net, we have AI. Riding an elephant, we have mobile. And doing a contortion act, we have US-China relations.

It’s been a slow time for IPOs, and the tech industry’s major silver lining is the AI boom. The bull case for Arm, the Cambridge, England-based company that designs chips, relies on its ability to move into the AI market. After all, in its current iteration, AI relies on massive data centers, stocked with state-of-the-art chips. But its most recent corporate filing raises some other concerns — and the question in my mind is what matters most to these investors. Finding out will be an important signal to other late-stage tech companies. Do the AI possibilities outweigh some of the significant drawbacks highlighted in the filing? How much does future potential count?

Let’s take it from the top. Arm is an intellectual property business about licenses and royalties; the company isn’t shipping silicon. Instead, its IP is licensed by companies such as Apple, Qualcomm, and Nvidia, which use Arm’s blueprints to design and fabricate their chips. That means evaluating the IPO means understanding Arm’s customers.

Even though it’s not really a household name, Arm’s products are for sure in your household and your pocket — it dominates the mobile market. But demand is slowing as consumers hold onto their phones longer, and according to the filing, the company’s revenue fell 1 percent to $2.68 billion in the fiscal year that ended on March 31st, 2023. Its net income for the quarter that ended in June — $105 million — was less than half of last year’s. This is not exactly the kind of growth we’re used to seeing in IPOs.

But it’s worth keeping in mind what kind of filing this is. We aren’t dealing with a special purpose acquisition company, or SPAC, where a company can make bold statements about its future. Arm has to talk strictly about what’s going on with it now and what’s happened in its past. The people who will be buying in on the IPO are gambling on Arm’s future. There are some hints about this in the filing — discussion of a 2022 restructuring feels like a delicate way to tell investors that Arm is a different company than it was the last time it was publicly traded in 2016.

Less constrained than Arm itself is Masayoshi Son, the founder of SoftBank. SoftBank bought a 25 percent stake in Arm last week from its Saudi-backed Vision Fund, which valued Arm at $64 billion. That’s about double what the Vision Fund paid for it in 2017, according to the Financial Times. That gives the Vision Fund’s investors a nice return, which it desperately needs.

The Vision Fund lost $30 billion last year, a comical amount of money. Some of its bad bets are famously bad: WeWork, which may not be a going concern for much longer, and FTX, a company that failed so spectacularly that the Enron expert said it was worse than Enron.

SoftBank itself did have $1 billion in investment wins in the quarter that ended in June, which was its first in a year and a half, so maybe things are looking up. But the company reported a net loss of about $3 billion all the same.

We can view the Vision Fund transaction two ways: first, as a way of giving Vision Fund investors a return, and second, as a way of ballparking a valuation for Arm. Figuring out a valuation is sort of more of an art than a science, but internal transactions — like that between SoftBank and its associated fund — are not necessarily as reliable as external transactions in figuring out what a company is worth.

But don’t take my word for it. Here’s the Financial Times: “Investors should take little notice of this figure.” The unbylined column figures that Arm’s actual value is closer to $30 billion, which is roughly what SoftBank bought it for in 2016.

Still, Son has been a booster. “We will aim for the biggest IPO ever in semiconductor history,” he said last year after an acquisition by Nvidia, valued at $40 billion, fell through. One way to keep a company’s valuation high is to sell relatively few shares. And so, with Arm, we’re seeing a low number of shares available for the general public. That stock float could be adjusted — the version of the filing I’m seeing could be superseded by something else, maybe after the road show — but as of right now, SoftBank is still the controlling interest.

Lately, SoftBank has been selling off investments; Arm is just the latest. The shares being offered to the public are a minority — and if the stock rises, SoftBank can sell more over time. Getting into Arm means getting into bed with SoftBank in a very serious way, after the company managed to miss the AI boat. Part of it was timing, and part of it was strategy: Son bet on small-scale AI rather than data centers, which is where much of the growth has actually taken place.

Okay, and what about Arm and AI? Maybe there’s long-term potential, but as it currently stands, there isn’t much evidence for it in the filing.

We know the mobile market is slowing, and we know that’s hit Arm already. But it’s maybe also worth noting how impressive Arm’s reach is. Something like 30 billion chips a year use Arm designs — that’s well beyond phones. We’re talking cars, for instance. Cloud computing. There’s also the internet of things, but that’s a relatively low-margin (and thus unimportant) business.

Because this is a licensing business, understanding Arm means understanding its customers. Arm’s top five customers accounted for more than half of its revenue last year, which means those customers have significant bargaining power. That puts some limitations on Arm’s ability to set pricing for its services — and it may need to raise prices to accommodate slowing demand in some of these sectors.

Arm’s clients — which include Apple, Amazon, Intel, Nvidia, and Google, among others — are considering buying into the IPO. “These companies’ interest is fueled by a desire to expand their commercial relationship with Arm, and make sure that their rivals do not gain an edge,” Reuters reports.

You can see why clients might want an edge here. Nvidia is already playing favorites with its chips — and making sure that Arm doesn’t do the same might be important to companies like Apple and Qualcomm since these chips are so ubiquitous in mobile devices. But with revenue from mobile falling, charging higher fees might be important, says Dylan Patel of SemiAnalysis. “There wasn’t a lot on risk and competition or how they plan to increase pricing” in the filing, he told me. “They’ve been telling people to expect higher pricing, especially smartphone customers.”

More pricing details may emerge in the pre-IPO road show, Patel says. Arm’s estimated royalties per chip range from 5 to 15 cents, says Ben Bajarin at Creative Strategies. “That needs to go up,” he says.

Arm’s CEO, Rene Haas, joined the company in February 2022 and has been wooing clients who operate data centers. That may be more profitable than the mobile market, particularly with the boom in artificial intelligence. Nvidia’s Grace chip uses Arm architecture already. “Servers really have more interesting upsides, just in terms of dollar amounts,” Bajarin says. “Arm’s got a lot of IP that can go into that.”

“SoftBank is pitching a story that’s not quite true,” says Patel. “That they’re in AI and they’re not.” Qualcomm and Apple use Arm heavily but not its AI IP, he says. “Arm has no direct AI business.” But Arm can help companies design chips that will let them integrate AI with everything else, he says.

More companies — including Google, Meta, and Amazon — are getting into making their own chips, which might mean a larger market for Arm to sell its designs. In its filing, Arm notes that many manufacturers are shifting away from “off-the-shelf” chips, which could be a growth opportunity. (Apple and Qualcomm have long used customized Arm designs, for instance; more recently, Google’s Tensor chips also customized Arm.) This may put Arm in a good position to benefit from the various attempts to compete with Nvidia’s AI chips. But it’s hard to see the future — this is more a question of where the winds are blowing than what Arm’s existing business is like.

On the other hand, there’s Intel and AMD duking it out, which add competitive pressure. “CPUs are kind of vulnerable with Intel and AMD now in a knife fight,” Patel says. If Intel’s new chips are a lot better and AMD continues to execute well, then the incentive for companies to design their own chips may fade. If a lot of these companies give up making their own chips and go back to buying them ready-made, Arm has fewer potential customers.

Arm China is Arm’s biggest customer, and since it’s an independent operation, Arm doesn’t have any control over it. Sure, SoftBank owns 48 percent of Arm China, but the majority is owned by investors connected to the Chinese government, which means that China’s in charge here. In the filing, Arm says it’s had trouble getting “timely and accurate” information from Arm China; further, Arm China is currently late on payments.

The risk factor section dealing with China is substantial. Because so many chips are sold there, Arm is particularly exposed to “economic and political risks” in the country. Arm China is the main way Arm accesses that market. And there have been incidents at Arm China before: in 2020, CEO Allen Wu was fired for conflicts of interest but then just kind of refused to leave. In 2022, new leaders were appointed and registered, which seemed to resolve some of the issues — but it’s overall a little… odd

One analyst told the Financial Times that the China risks disclosed in the filing were bigger than expected. Payments from China also seem to be falling, putting more pressure on Arm to make money elsewhere. With Arm China accounting for a quarter of Arm’s revenue, that’s a little hair-raising, particularly as the Biden administration continues to use AI chips as a cudgel in the China relationship.

On the other hand, in terms of risks, there’s “nothing that surprising” in the filing, according to Chris Miller, the author of Chip War. “They’ve been playing out in newspaper headlines,” he told me. “Every semiconductor company has, over the last five years or so, come to realize that US-China dynamics will have a major impact on their business.”

Arm isn’t uniquely exposed and might even be less exposed than companies that make chips directly, Miller says. If the US or China decides to punish one specific firm, Arm might actually be a little insulated because it sells architecture to so many businesses. Plus, it’s not like either side of the equation is going to cut themselves off from mobile devices, which is where much of Arm’s business is now.

The IPO market has slowed significantly in the last year, and tech company valuations fell as interest rates rose. Along with an expected IPO from Instacart, Arm will serve as an important bellwether for investor sentiment — because there are a lot of late-stage private companies waiting in the wings.

Arm might raise as much as $10 billion in the IPO, according to Bloomberg. That would make it the year’s largest IPO and might clear the way for other companies to go public as well. Databricks and Socure are among the companies that might follow in its footsteps, according to the Financial Times.

With so much of Arm’s AI potential in the future — and thus, not captured in its filing — sentiment really matters. There are questions about RISC-V, an open standard architecture, as a possible competitor (though Arm CEO Haas didn’t seem particularly concerned about it in an interview with The Verge last year).

There remain a lot of open questions for Arm as it embarks on its road show. That includes its valuation, which may be lower than SoftBank is hoping. The move to AI seems to depend on how well Arm can anticipate future demand. The China risks aren’t new, but they are newly pressing. On the other hand, if Haas can make a convincing case that the changes at Arm will let it grow into the more lucrative server market, expanding away from the saturated mobile market, SoftBank might get a win.

/ Sign up for Verge Deals to get deals on products we've tested sent to your inbox daily.