How Arm Powers Chips By Apple, Amazon, Google And More

Every single one of these devices has a chip inside that has the Arm CPU, and that Arm CPU is the brain of that device. There's a company nearly every chipmaker relies on that doesn't actually make anything tangible. Yet its blockbuster IPO in September valued it above $54 billion. Arm is open for its first trade. It's the largest public offering in over two years. Has brought confidence back to the IPO market. Computer chipmaker Arm Holdings getting a nice reception from Wall Street today. Rath

er than make chips, the UK-based company designs the architecture, then sells licenses for these instruction sets to companies that make central processing units, or CPUs. What exactly is an architecture? If you think about the CPU as the brain, those are the instructions or the commands, if you will, that tell the brain what to do, how to walk, how to run, how to lift an object. It also collects royalties on every chip shipped with its technology. Arm is the most pervasive CPU or brain in the h

istory of modern electronics. 250 billion of those brains that have shipped in our history. Over 30 billion of those brains ship every single year. Arm's customers include Apple, Nvidia, Google, Microsoft, Amazon, Samsung, Intel, Taiwan Semiconductor Manufacturing Company. In short, the world's biggest names in tech and chips. Most people think about a device. Then maybe if they're really sophisticated, they think about the chip, but they don't think about the company that came up with the origi

nal ideas behind how that chip operates. But once you do understand what they do, it's absolutely amazing. Arm chips are known for using less power than those made with rival x86, the older traditional PC and server architecture used by CPU giants Intel and Advanced Micro Devices. Now there's a surge in adoption of Arm. It's the basis for Apple's M-series of processors, which have replaced Intel chips in Macs; for Amazon Web Services' custom server chips; for Qualcomm's flagship Snapdragon chips

that are poised to break into the PC market. And now Nvidia and AMD are reportedly working on Arm-based PC chips, too. But in recent years Arm has also faced plenty of risks. About 20% of its revenue comes from China. Smartphones, which almost all contain Arm processors are seeing a major sales slump. And when Nvidia tried to buy Arm for $40 billion, the deal was blocked by regulators. That didn't go the way that everyone anticipated or hoped that it would. But the sun comes up the next day, ri

ght? And you have to be able to build from that. CNBC went to Arm in the UK and talked to CEO Rene Haas to find out how it became the year's biggest IPO with so many big name customers, in the face of low smartphone sales and geopolitical uncertainty. Arm was founded in 1990 by 12 chip designers working out of a turkey barn in Cambridge in the UK. It was originally a joint venture between Apple, Acorn Computers and VLSI, known today as NXP. The model was always to come up with a core design that

we would license to lots of different companies. Apple launched its early handheld Newton device on the Arm610 processor in 1993. Way ahead of its time: ran off a battery, had a display, which kind of goes to what the hallmarks of the company are. We were born running a device off a battery that was going to be low cost. So that was the birth of the company. For Arm, they received per unit royalty and they were able to take that money, reinvest it, hire more staff and grow and expand. Arm then

struck a deal with Texas Instruments, putting its processors in early Nokia mobile phones, beginning Arm's climb to become the dominant architecture in almost all smartphones today. There's probably 16 to 20 CPUs alone inside the smartphone. And it will do things like recognize your fingerprint, it'll manage the camera, it runs all the applications. Arm went public for the first time in 1998. Chief architect Richard Grisenthwaite was there. I joined when we were much smaller company and we were

about 100 people, and been very much involved in this tremendous transition that the company has gone through, expanding out from targeting one particular market area into a wide range of different computing environments. Arm grew rapidly in the 2000s, with the first touchscreen phones introduced in 2007, and the growth of connected home devices in the 2010s. Arm now has some 6,500 employees globally. About a sixth of our population is in the U.S. and a lot of very high quality engineering, both

architecture and microarchitecture, goes on there. And then we've got a lot of people in the UK and we've got people also in India and in France and in lots of other places around the world. In 2016, Arm went private when it was bought by Japan's SoftBank for $32 billion. Now CEO, Renee Haas was president of the IP products group at the time, spearheading diversification into emerging markets, including AI. PC and phone, automotive, data center and IoT. Every single one of those markets has AI

embedded in some way, shape or form. Stop, go. Arm says it has some 6,800 patents worldwide, with another 2,700 applications pending. Some of those are for Arm's Neoverse line for high-performance and cloud computing, which has helped it break into AI since its launch in 2018. Nvidia just partnered with Arm on their new Grace Hopper Superchip, which is designed to combine the GPUs of Nvidia plus the CPUs of Arm. Obviously huge for any AI-related things. That's actually got 72 of our Neoverse cor

es in it, plus their own big GPUs. By bringing those together and tightly coupling the way that Nvidia has with the Grace Hopper, they're able to come up with something that's something like 2 to 4 times the performance of what you'd get on an x86 system for a similar amount of power. But in 2020, Arm owner SoftBank needed cash. It had lost money on investments in companies like WeWork and Uber. So it struck a deal to sell Arm to Nvidia for $40 billion. Eighteen months later, the deal fell apart

, blocked by regulators and a consortium of Arm's biggest customers that also compete with Nvidia. It was a long long process. Disappointed it didn't happen just because we spent so much time on it. But I was very confident, whether it happened or it didn't happen, things were going to turn out great for us. Instead, SoftBank announced plans to take Arm public again and Haas took over as CEO. Arm made its second public debut on September 14th, climbing nearly 25% that day. It's fallen significan

tly since then. Although it did beat expectations in its first earnings report, revenue guidance came in lower than expected, sending its stock down 7% in extended trading. Arm makes money in two different ways by selling different kinds of licenses, charging companies a fee to access the Arm technology for building their chips. And you're collecting royalties for each of these chips? We are. Yeah. Any dollar figure you can put on? I don't have the numbers off the top of my head, but for us basi

cally, our model is 60% of our revenues come from the royalties and growing, 40% of it comes from our licensing business. Arm dominates in smartphone processors. But x86 architecture, developed by Intel in the 70s, still leads in computers and server processors. RISC-V is the other main competitor. The free open-source architecture has seen a recent growth in popularity, with backing from some of Arm's big customers like Google, Samsung and Qualcomm. They may have been seeking alternatives when

it looked like Nvidia was going to buy Arm. RISC-V sits a few years behind where Arm is at, and I don't think we're going to hear a lot about it right away. I do think in low power, in IoT and simpler designs, that RISC-V does have some traction. Rival x86, meanwhile, has more software developed for it than Arm, although that may be changing. The amount of software support is the thing that actually tends to determine the success or failure of that in the long run. Intel was very good early on w

ith getting a ton of software support for x86. Most servers were x86-based, but what's happened in the server market is that the software has been componentized. It's broken up into containers and things like that, and that makes it easier to run on other architectures like Arm. Arm's biggest differentiator is lower power consumption, something that's helping Arm break into the laptop market. Apple, for instance, says its Arm-based M3 chips give the newest MacBook Pro up to 22 hours of battery l

ife. Apple's move away from Intel was a big proving point for Arm's ability to diversify beyond smartphones. Nobody really believed, until Apple went all in and basically cut ties with x86 instruction sets and said, 'We are going to bet the future of the Mac on Arm.' And that was a huge inflection for the company. It was a change of the guard. And this isn't to say that Intel's future is in big trouble, but it certainly started to raise some question marks as to, well, if Apple can do it, can ot

hers? In September, Apple extended its deal with Arm through at least 2040. Qualcomm is another major customer making PC processors using Arm. Although now Arm is suing Qualcomm over the right to make certain chips with its technology. The issue started after Qualcomm acquired CPU company Nuvia in 2021, and with it, Nuvia's Arm license. Nuvia was actually supposed to be designing a server chip initially, so they had different terms with them, and so Qualcomm thought they could have the same term

s. Arm felt no, different companies have different terms. And it's boiled down to essentially that: legal discussions around what those terms ought to be. The case is set to go to trial in 2024, and Qualcomm remains committed to introducing Arm-based PC chips on Windows laptops in the next year. On the server side, Amazon Web Services is the big player making Arm-based chips for the data center. And that was the launch of Graviton. And really from there, Arm went from this mobile, low power, IoT

, automotive, specialty, embedded to holy cow, we can build next generation servers, PCs, and of course continue on this massive run of silicon for smartphones, all based on Arm. Arm chips have long been in cars, but it's now a rapid growth area with the rise of self-driving capabilities. And actually, it's one of the most computationally intensive tasks we've ever seen on this planet in terms of mass deployment of it. And what we need to provide is a standard platform to allow the world's softw

are developers to really concentrate on this incredibly hard task going forward. And for this reason, we've developed the AVA development platform, which we're showing here, which consists of 32 of our Neoverse cores. Arm is also seeing growth due to a trend of non-chip companies like Amazon and Apple turning to Arm to design their own custom silicon, allowing them to depend less and spend less on historic chip giants. We've gone primarily from companies who that's all they did to companies like

Microsoft and Apple and others who are now adding chip design as just a piece of their business, which means, you know, they've got a smaller team than entire companies built on that. And so you have to make that process easier and simpler. And that, for example, is where Arm is starting to move in terms of enabling the design of multiple components that connect together. They call it compute subsystems. While more companies are indeed making inroads into semiconductor design, the recent chip s

hortage exposed major concern over the fact that more than 90% of the world's chips are made in Asia. Now China and the U.S. are going back and forth imposing export controls on chip technologies, which could escalate to impact Arm's significant revenue that comes from China. In 2018, SoftBank broke off Arm's China business into an independent entity, Arm China, that's majority owned by a group of Chinese investors. It's essentially to allow us to not only grow our business in China, which is ou

r essentially base core business. We set up a distributor Arm, but at the same time, we also created an R&D arm that allows an independent entity to develop products specifically for the China market, some that are Arm-based but some that are not Arm-based. Haas says Arm has seen minimal impact from export controls so far. China's a good market for us, about 20% of our business. It's shifted over the years. It used to be largely mobile phone based. Now it's mostly around the data center and auto

motive. But Arm China was also embroiled in controversy for years, with SoftBank and Arm trying to oust the CEO of the China business, Allen Wu. Despite being fired twice, Wu refused to leave until last year. Now, several former Arm China employees are starting a new internal chip design company in China with backing from Shenzhen's government. Arm's stock slid more than 5% on the news. It's been very ugly and kind of messy and confusing. And now, of course, some of the people there have moved t

o start their own company. But again, a lot of Chinese companies have long standing relationships with Arm. So the expectation is they're going to want to work there because again, they have that huge base of software. If somebody creates a new architecture, they have to build the software, and that takes years and years and years. In February, Arm China laid off about 100 employees. Meanwhile, rival architecture RISC-V is gaining traction in China as an alternative to U.S.- and Europe-based tec

h that could end up banned in China. Geopolitics aside, Arm also faces risk from the major slump in smartphone sales. If they were still totally dependent on it, it would obviously not be great, but because they've diversified into these other areas, it gives them much more opportunity. That's why you see them focusing so much on things like Neoverse for servers, on autonomous driving with cars and on these embedded devices. So that's where they see the future moving ahead. We're not as impacted

as folks might think because one of the trends we've seen, particularly in smartphones, is more and more Arm processors that go into those phones. So for us, we've actually seen an increase in royalty per phone. Labor is another marketwide headwind. The world's chip leader, TSMC, is blaming a shortage of skilled workers for delays at its $40 billion fab under construction in Arizona. We had a shortage of talent and now it's hard. It's hard for Arm. It's hard for our whole industry. There's no w

ay that demand for semiconductors in the next 10 to 15 years will abate. It's only going to increase. So it's a pretty fierce talent war. That's why Arm is helping launch a Semiconductor Education Alliance to create new pathways for future talent pools, amid the increasing chip needs of generative AI and the growing list of companies like Nvidia working on Arm-based processors. What that's going to mean is increased demand for Arm technology everywhere, in all those markets that I mentioned, and

more and more compute capability because every single AI application needs a CPU. Every single one. You cannot do AI without a CPU. End stop.