The Hidden Force Behind the AI Revolution: Arm CEO Rene Haas

0:00We're at a very fascinating time in terms of where the winners and losers are going to be, particularly with business models. I was listening to your interview with Dan Ives and he was talking about the internet, whether this is 1995 or 1996. I think it's more like 1992. And the reason I say that is AI is so much larger than the internet. AI will touch every single piece of human interaction with almost any device or any piece of software or hardware. Will it prove to be that there's been overinvestment in certain areas? Almost certainly. I would say at some point, yes. Are we there now? I don't know.

0:37I mean, there's, if you can talk about, there's a, there's a valuation bubble, which I'm not going to comment on. And then there is an overinvestment, overinvestment model. I don't think we're overinvested at the moment when you look at large language models, but I do think large language models at some point in time top out in terms of their efficiency relative to the amount of capital we're required and more efficient or different models will be invented. One of the things that I'm very passionate about, as anybody has to be who's in a tech company, is there are no sacred cows. You have to look all the time at where is the market going, where can I provide value, and ultimately where do I need to take the company in terms of different directions.

1:16So we look at things like that constantly. So do I know the answer of what the model looks like in the future? No. Do I have a very good sense that ARM is going to be at the heart of it, 100%. Welcome to the Master Investor Podcast with me, Wilfred Frost, where we celebrate and learn from the success of the greatest investors, business leaders, and politicians in the world, giving you, our listeners, an edge. The Master Investor Podcast is sponsored by BNY Investments, LSEG, and Interactive Brokers. Please do remember the views expressed in this podcast are for general information purposes only.

1:54Nothing in the podcast constitutes a financial promotion, investment advice, or a personal recommendation. More on that in the show notes. My guest today, René Haas, is the CEO of perhaps the most important company in the UK, Arm Holdings. He's been CEO since February 2022 and at Arm since 2013. Before that, He spent seven years at NVIDIA and started his career back at Texas Instruments in 1994 and has spent essentially his entire career in semiconductors. He's also on the boards of AstraZeneca and SoftBank. Rene, welcome to the Master Investor Podcast. Thank you. Happy to be here. In 1994, you gave me 10 years of extra shelf life.

2:38Oh, did I? 1984. 1984. There we go. I got 10 years younger, but thank you. Do you prefer to be younger or have more experience in this framing of it? That's a good good. This morning, I'll take the more experience. Okay. Okay, perfect. Sorry for that. I got that wrong. Anyway, the key part, or the most important part of that is being CEO of Arm since February 2022. And I said in the intro, the most important company in the UK, I don't think it's talked enough about in this in this country. But I guess the first point before getting to the importance is to reiterate, you are a British company, a lot of fanfare and coverage that you listed when you IPO in New York.

3:15But you are you consider it a UK company, will always be a UK company? Yeah. I mean, the company was born in the UK, right? So we started in a barn in Cambridge. Our headquarters are in Cambridge. The vast majority of our employees in the company are in the UK. So by all accounts, we are, I'm proud to be a British company. And I guess I said that too as well, most important British company, that's, I guess, a slightly open debate. But my point on this is your chips, the chips that you design, appear in every device around us here in this room, but around the world? If we define importance by ubiquity and popularity and quantity, then ARM is extremely important.

3:59Inside this little room that we're in right now with a number of cameras and video players, et cetera, et cetera, probably 50, 100 ARM processors that are running all the machines. The ARM product is a CPU, which is the brain of every electronics device. So when you think about the importance of a brain to the human body, pretty important. And literally every electronic device uses ARM as its brain, whether it's a security camera, whether it's earbuds, data centers, as I said, inside this room, cameras. So by those definitions, yeah, we're quite important. So just for the uninitiated, another kind of setup question about what ARM does.

4:36What is a GPU? What is a CPU you just mentioned? And why does the NVIDIA Grace Blackwell sort of sum up the evolution of the industry in the last five or 10 years and highlight to our listeners why ARM is so central to everything and so tied to NVIDIA, which is the Goliath$4 trillion company that a lot of people have heard of? Right. Quite a mouthful there. So maybe a way to break it down is if you think about the semiconductor supply chain, it's quite disaggregated. What do I mean by that is there are companies that physically manufacture the chips, actually build the chips themselves. And for our world, there's really only three left that do it, TSMC, Intel, and Samsung.

5:21Then there are companies that design the entire system on chip. This is NVIDIA. This is MediaTek. This is Qualcomm. This is Apple. Myriad of companies who, end quote, build chips. But what they actually do is design the chip. And then the chip is manufactured, as I said, by TSMC or Samsung or Intel. So where does ARM fit in? We do the actual design of a component. And our chief component is the CPU. And we license that product to a company who's trying to put together an entire SOC. So again, we design the brain. We license the brain to someone building the chip. The chip is the body, the arms and the legs and the feet.

6:00And then someone puts it all together, the chip manufacturer. So where we sit in the value chain, Wilford, is we do the design of the CPU, which again, as I mentioned, is arguably the most important component in every digital device. Every single electronic device that runs software has to have a CPU. That's just by definition. So you said, well, what is a GPU? A GPU stands for graphics processing unit. Traditionally, what they've done is run graphics. They draw triangles. They run images. So in a PC, the central processor runs all the software, and the GPU is what actually puts the images on your display.

6:36Now, what's happened over time is, particularly with AI, GPUs are very good at certain problems, AI being one of them. So NVIDIA has had this huge explosion in growth around artificial intelligence using their GPU as the engine that does the processing, but that GPU can't run alone. So they have a product called Blackwell. That GPU needs a CPU. That CPU NVIDIA calls Grace is based on ARM. So when you think about Grace Blackwell, what is Grace Blackwell? It's a CPU based on ARM designed by NVIDIA using our intellectual property, using a GPU designed by NVIDIA and then ultimately built by TSMC. And that was such a great explanation and I think it's going to be super helpful to everyone, including me, even though I've researched this tons.

7:22You mentioned their intellectual property. I mean, just again to dwell on this point, you are really a pure IP company. You make these blueprints, these designs, as opposed to a manufacturing company. And in that sense, innovation is at your very core. 100%. The company started in the early 1990s as a design shop. And there was a product way back in the day called the Apple Newton. And that was a PDA way, way ahead of its time. That PDA needed a microprocessor, needed a CPU to basically run the display in the machine in the actual unit. Back in the 1990s, what was very important, it had to be very low power, had to run on batteries, and it actually had to be fairly low cost and very efficient on heat because you're putting this into a small device here and you don't want your hands to melt when you've got the PDA in your hand.

8:16There wasn't anything on the market that could fit that need. So there was a joint venture between Apple, VLC technology, and they went to a number of engineers who came out of a company called Acorn, early British computer company, and designed a custom microprocessor, a CPU for that PDA. That was the Newton. Now, the Newton failed. The joint venture didn't really go anywhere, but ARM was born from that innovation. And the founders of ARM had a very brilliant idea back in the early 1990s was, we're not going to build this chip based on ARM. we have this design that's very, very good at certain things, low power, chiefly, and very, very efficient.

8:59And we're going to license it to companies who want to build chips around it. Again, back in the 1990s, early 1990s, that was a crazy idea because there wasn't an ecosystem that allowed companies to really take the IP. There wasn't really a third-party ecosystem to design it. But what it did is it lowered the barrier for companies to actually adopt a microprocessor that a software ecosystem could be built on. So that's how the company was born. Robin Saxby, one of the founders of, not the founder of ARM, but one of the first CEO, he wanted to build a global standard for CPUs. And that was his vision for ARM, and it came true.

9:36And just to dwell again on the location, you know, half over half, I think you said, your headcount is in the UK, particularly in Cambridge. Is that still the right place in the world to do this sort of innovation work? Yeah, it's a wonderful question. Microprocessor design is really hard, really, really hard. And getting very, very good at it in terms of building products and generations and iterating is very, very hard. So when you build a critical mass of people in a location who can understand how the architecture is designed, can teach engineers how it's built, and can bring on new legions of engineers, it's huge.

10:15So we've been in Cambridge ever since we started, you know, early 1990s, so 30 plus years. There's a number of engineers. I think finally the last founders have retired, but there's a number of folks, Wilford, who are there, have been there 25 years, 28 years. My chief architect, I think he's been there 28 years, our head of engineering, 25, 26 years. It's hard to replace that critical mass. I mean, to pick that up and say, okay, we're going to move it to North Africa because the talent's cheaper there is it's nonsensical just in terms of attracting people and critical mass in terms of just getting the projects done.

10:50And what about young talent hiring up? We hire a lot of people. We have a graduate program, a few hundred a year, two to three hundred a year that are hired in here. Cambridge, Oxford, Imperial College. So it's a big effort for us. At the same time, I've heard you speak since coming in in 2022 about lacking some of the American hunger or mindset when it comes to scaling the company. Talk to me a bit about that and what you kind of observed and how you're trying to marry all of that. Obviously, your own background having spent a huge part of your time in Silicon Valley. Yeah. No, I think I've heard you on one of your podcasts talk about this since you lived in the States for a while.

11:30That's kind of why I launched the podcast. Yeah. So my career, early part of my career was with larger companies, but I always had an itch to do startups. So moved to Silicon Valley in the middle 1990s and did startups when I got there and then joined NVIDIA, which is a big startup, founder-led company, et cetera, et cetera. So my training and development and my DNA sort of morphed working with startup companies. I found that that, A, fit my nature well, the risk-taking, the entrepreneurship, trying something new. So when I joined Arm in 2013, wonderful company, and obviously was a start of it at one point in time, what I was struck by was a bit of we would go into meetings and we would talk about new projects and the first six or seven things we would talk about are all the things that could go wrong, right?

12:19All the things that could, why were the reasons that we shouldn't do this? And I don't know if that's a British thing or if that was an arm thing, but I've definitely sensed it working on both sides of the pond that for whatever the reasons are, and you were born here, you can probably say better than I can, there definitely is less appetite for risk. And maybe that comes from less of an appetite for failing. Whereas in Silicon Valley, if you fail to some degree, it's a badge of honor. Here, I didn't sense that as much and even an inside arm. So one of the things I've really tried hard to do since I took over in 2022 and even before that was inject some of that Silicon Valley appetite for risk, moving quickly, making mistakes, but still keeping the things inside the arm culture that are amazing.

13:09Super bright people, highly innovative, highly collaborative. So I've tried very hard to mix them both. I love that. And it's clearly worked. And the share price has performed very, very well. And this brings me to my next question, which is not a valuation multiple question, which maybe we'll come to. More, I guess I'm getting to the business model point, which is that Arm has been incredibly successful. Valuation today, $120 billion or so. But you mentioned a couple of other companies already that are as central, as exposed to this AI revolution as you are. TSMC is worth 1.5 trillion. NVIDIA worth four, four and a half trillion.

13:50I guess my question on this is, you're so central to the extraordinary success that they've had, is it a bit unfair that you're only worth$120 billion? Is that a business model question? Perhaps, but it's all relative. Your success has been huge as a company. Relative is the right point. I'll put this bit of context of what we've been trying to do in terms of arm, in terms of growing our position, which has been pretty dramatic over the last number of years. So our business model is licensing, and that essentially is customer wants to use its technology, they pay a license fee, and then on every unit shipped, we collect a royalty.

14:32The royalties have been very modest in the early years. And for some senses, that was a very good thing because what it did was it allowed the architecture to proliferate. Many people used it, low friction, et cetera, et cetera. But when I took over, I looked at, well, what are the things that we can do that can grow the top line and the bottom line, but also at its core, deliver more value? And one of the things we looked at was the IP as a CPU is one component, but there are many things that we deliver with the CPU to build a full system on chip. And these are different aspects of IP. We started to put them together into what we call compute subsystems.

15:10And this is blocks of IP. let's say Microsoft wants to build a 128 core chip. We actually provide the entire 128 cores together. We verify it. We validate it. We provide all that value to the customer. That drives much better royalty rates for us. And as a result, we put a very high focus on delivering more value and the royalties have really, really grown. So the proof point of that is 30 plus year old company. It took us 20 years to get to a billion in revenue. It took us another 10 to get to two. So now 30 years to 2 billion. It took us two years to get to three, one year to get to four. So we're growing.

15:53And that's a combination of the strategy has been working. And we've been able to get into the new market, such as Grace Blackwell, into data centers, into AI. So we'll get there. I was at NVIDIA 20 years ago, and our market cap was$20 billion when I started, and it was$20 billion when I left. It went up and down during the time. In our industry, it does take time. I don't want to play the valuation game, but in terms of ARM's long-term growth prospects, I'm incredibly bullish on it. It's a really interesting point, that accelerating growth. I guess the question that then always follows is whether or not you should, or you're considering manufacturing chips more yourself.

16:35And it's always sort of rumored you do talk about it a little bit. Is that something that fits with that and follows on from that question, or is it not really worth it? Yeah, it could. And to be, just to put a fine line on it, manufacturer chips, in other words, will I build a factory like TSMC? No. I mean, it's$30 billion of CapEx. Even my boss, Masa doesn't really have the appetite for that. But could we design what NVIDIA does, the full system on chip? Potentially, because some of these chips now, Wilford, if you look at all the intellectual property that's inside them, in some cases, 95 % of it is ARM.

17:10And if we can provide something to the market that allows our end customers to get to market faster and sooner with a better product, we'll look at it. One of the things that I'm very passionate about, as anybody has to be who's leading a tech company, is there are no sacred cows. You have to look all the time at where is the market going, where can I provide value, and ultimately, where do I need to take the company in terms of different directions? So we look at things like that constantly. And I guess the rumor or everyone's very focused on is OpenAI as a potential partner, and not least because of the kind of tie-up as well with SoftBank and MasaSund.

17:52Is that far down the line? Is it close? What's the kind of update there? I mean, they're a wonderful partner. One of the benefits of having SoftBank being a chief owner is alignment with your biggest investor, Masa. He's all in on AI as we are at ARM. He's very big on OpenAI, as are we. OpenAI is a magnificent company, and they're doing things all across the value chain. Obviously, what they're doing with the models is obvious, but they bought Johnny Ives' company, I.O. They've got some interesting ideas there. There's a lot of ways we can partner with OpenAI. And do you think it would upset your other core current customers like NVIDIA?

18:31One of our challenges, we were chatting a little bit earlier about upsetting different people in the ecosystem, everybody is an ARM customer. I mean, literally everybody is an ARM customer. If you were to ask me, well, who doesn't use ARM? I would have a hard time answering that question. So we think about that all the time. If we are going to go off and do X, is it going to upset Y? Give you an example, the two largest ecosystems, the three largest ecosystems for software are Android and Chrome, Windows, and iOS. They all run on ARM. So I have to think about what am I doing that upsets the Apple ecosystem versus the Microsoft ecosystem versus the Google ecosystem.

19:08So it's something we think about all the time. And I'd like to think we're pretty good at navigating through that.

19:19This podcast is sponsored by Interactive Brokers. Building wealth starts with the right broker. And Interactive Brokers helps you reach your goals with powerful tools, global market access, low costs, and unmatched financial strength. That's why the best informed investors choose IBKR. Learn more at ibkr.com forward slash master investor.

19:48You mentioned that these companies that you work with, all the greatest tech companies in the world. I just wanted to reflect on who you think are some of the great tech leaders at the moment. kind of two different sort of styles in terms of operationally, but in terms of innovation. I mean, first, who is the greatest innovator, do you think, alive today? Oh, boy. I'm going to make sure I don't insult anyone. I have a predisposition, of course, for the work that Jensen has done at NVIDIA. I think anyone who is still running a public company 30 plus years after with the level of energy and intensity and tenacity that he has is remarkable.

20:28And NVIDIA, as we mentioned, they were not an overnight success. Gaming company for many, many years. A lot of criticism for all the investments they were making in artificial intelligence and CUDA before it bore out. So, you know, Jensen would be on the list. But I think about the semiconductor ecosystem. I think Hawk Tan is brilliant. I think Lisa Su is brilliant. When I think about those large ecosystem players, each in their own way. I think what Satya Nadella has done at Microsoft is utterly remarkable. a company that didn't look like they could get out of their own way. And he was an inside guy.

21:03And I know Satya well, and I have the deepest respect for him. What he's done has just been amazing. I'm giving accolades to everyone. But I think what Sundar has done at Google is also amazing. A couple years ago, people maybe 18 months ago sort of left Google for dead in the AI race and felt like it's game over. It's all going to be around chat GPT and the work that they've done with Gemini, with DeepMind and Demis. So one of the things that's wonderful about technology is it's never dull. There's always someone to look out for. You were very skillful there on not upsetting any of your customers.

21:40The next part on who's a great operational leader, I mean, I was wondering whether you'd say Tim Cook and Apple are just another great quarter of earnings again last night. But what I think is so interesting when it comes to Apple is how core they were in ARM's history. And again, I heard you talking about this on the AQ2 podcast. And talk us through how all of a sudden things changed when the iPhone came along and how central you guys were to that. Yeah, so Apple has an amazing history of market makers, if you will, relative to the end products. And if you go back in time, they were historically, if you go back to Steve Jobs' time when they were building early Mac products, they were all based on Motorola and then something called PowerPC, which is now an extinct entity.

22:33But that was an IBM and Motorola joint venture, and Apple was involved in that as well in terms of using the product. And literally overnight in our world, they flipped over to Intel. And everything went away from PowerPC to Intel. And Intel at that time was the dominant supplier in the Windows ecosystem. And Apple did a lift and shift of their operating system away from PowerPC to Intel. Huge, huge event. So the penny drop moment for us was at the moment where Apple was looking at, we want to build a mobile device. All of our software, our complex software, is written on Intel for Mac OS. Do we take Mac OS and try to put it into a mobile device, a.k.a.

23:22an iPhone? Back then, they were building the iPod. The iPod was based on ARM because we're really good at that, right? runs off a battery, had a number of complex things to go off and do, but it wasn't a computer. But the big leap they made was we were going to build basically a handheld computer that looks like the capability of a MacBook, but fits in your hand, aka the iPhone. And the big debate inside Apple was, do we stay with Intel or do we try to go from the iPod, which is this ARM-based thing? And they chose ARM, which was a wonderful outcome for us, obviously. And that is where Arm became the global standard in smartphones.

24:01Android with Andy Rubin quickly followed. And next thing you know, in 2008, 2009, every smartphone on the planet, I'll be careful with the word every, is now based on the Arm architecture. It's so fascinating. And what I love about hearing that and got me thinking is you're ahead of the game on knowing what the next transformational product is that's about to coming out. I mean, as you said, smartphones transformed formed everyone's lives. I'm sure the work that went into it wasn't overnight. So you must have a two-year lead time on what's about to come. And I'm sure you can't tell me what's about to come, but I guess it comes back to why you're so bullish on AI generally, because you must know in a year or two years or three years, the next big thing that's coming.

24:49Yeah. I don't know if we know, but we have good instincts. And by that, I mean, And AI at the end of the day is a workload. It's a computer workload. And today, that computer workload almost exclusively runs in data centers. And you're training these large models, which takes a lot of compute. And you're also then running inference. In other words, basically now running the output of those training models. And today, that's all done inside the data center. And traditionally, and that's a loose word here because we're so new inside this, That's been a really 100 % GPU-bound problem. Yes, you need the CPUs.

25:26The CPUs work in conjunction with the GPUs. But people have looked at it and said, A, that's a GPU problem. And B, that's a data center only. That's not where it's going to go on a couple of different reasons. So let's just talk about the data center. As you start moving to more and more agentic AI, and these inferences are running agents, and these agents are working with other agents, those are much more CPU-bound problems. And those are CPUs talking to CPUs, launching the agents together. So I think what you're going to start to see in the data center, and we've seen it already on the Vera Rubin platform that was announced by NVIDIA, where the number of CPUs from platform to platform have gone up almost 6x.

26:07So back to the CPU relevance. So when we look at the trajectory of the CPU in the data center, we have a very, very good sense that's just going to continue to increase. Now, then you think, well, is AI only going to run in the data center? Well, of course not. And history has shown us time and time again is that ultimately workloads move to local devices. Local devices meaning the mobile phone, the PC, wearables, glasses, watches, something that's not invented yet. And what we have a very good sense is that they're going to move to these low power devices, not away from the data center in totality, but you're going to run AI locally.