Good evening everybody and welcome to FT House on Tuesday night here in Davos. I'm very pleased to be here with Rene Haas, the ARM CEO, a great British company. We're going to talk about everything under the sun, including Donald Trump's arrival in town tomorrow. Rene has been at ARM for a long time. He's been CEO for four years. He's on the board of SoftBank, the majority owner. And I can't think of a more hectic, dynamic, interesting time to be leading a global chip company and overseeing one of the main investors in AI and robotics across the world. So I think the best place for us to start is if you just kind of introduce the company and set the scene for us where ARM is post-IPO in the middle of this AI hype boom. Obviously, you got your start in making ultra-efficient chips for mobile phones, all of which you probably have a chip in your phones that you're holding up right now. But I want to hear where you are in the competitive landscape at the moment and where ARM is going next, where the next leg of growth is going to come from for the company.
Sure. No, thank you for having me. We're in the FT house. I'm going to assume that there is some working knowledge of ARM from folks who are from the UK. But for those of you who aren't familiar with our company, we're about a 35-year-old company based in Cambridge that is best known for the world's most ubiquitous compute platform. Our product is a CPU, which is the digital brain of every single device. And as mentioned, I can assure you everyone has ARM on their body today, either in their smartphone or some other device, the smartwatch, etc. ARM is the CPU that powers literally 100% of all the smartphones on the planet. We power PCs, we power smartwatches, we power dishwashers, we power cars. But we also power data centers. Nvidia is one of the largest partners that we have. The Vera Rubin platform that was just announced by Jensen at CES, the Vera CPU, that's all ARM. So we're everywhere. And now, this time of what we're seeing with artificial intelligence relative to literally every device having to run some level of AI workload is a combination of fantastic challenge but also opportunity. So we can chat more about that, but that's ARM in a nutshell. I think we calculated somewhere that the amount of ARM processors that have been shipped in our creation exceeds all the people who ever lived on Earth and then some, which I think is 100 billion plus.
Wow. Okay. Well, that's a great introduction, setting the scene about where ARM fits in the global tech landscape. It's very interesting. When you were at CES, you spoke to one of my colleagues and you talked about one of the overhangs that you're seeing on the market at the moment, which is a potential looming shortage in memory.
Yep. As you said, the world is moving from prioritizing smartphones and data centers just for normal cloud computing. But now we have this vast explosion in data center construction and a lot of the memory capacity is going towards components for those as opposed to components in smartphones, which naturally leads to the concern that either smartphone prices are going to rise to compensate for this lack of memory or these phone makers are going to see a hit to their margins. I mean, how are you thinking about this and do you think that this shortage can be, how long till this is cleared?
Yeah. So memory shortages in the semiconductor industry are nothing new. We've seen them before. Not the kind of dynamic that we're seeing this time. There's a type of memory called HBM memory, high bandwidth memory, which is largely used by these data center accelerators. The use for this high bandwidth memory has just exploded. And the reason for that is when you're doing an AI calculation and you're trying to calculate some of the key values that need to be referenced to spit out tokens, that reference is all held in this high bandwidth memory. And that's based on DRAM. So what we're seeing is there's such an insatiable need for this high bandwidth memory that the memory makers are shifting away their standard DRAM towards this high bandwidth memory, which for them is smart economics, but that puts a lot of pressure on the system for everything else that needs DRAM, which are smartphones, PCs, automobiles. So I think we're going to see some pain in the short term, but in our world, shortages are basically the catalyst for invention. And I think what we're going to start to see is a lot of people looking for alternative ways to build systems that aren't so reliant on this kind of high bandwidth memory. That's not going to happen overnight. That's going to take some time. But I also think it's necessary because as AI starts to move from the data center to the edge devices, and edge devices are essentially wearables or smartphones, the last mile that does the calculations, AI will start to run in those physical devices and that will require some different kind of memory technology because high bandwidth memory is too high powered to run inside a phone. So I think we're going to see some short-term pain quite candidly, but I think in the long term it'll be a good thing because it'll drive some new innovation.
And what about the price of everyone's Pixel and iPhones? Are they going to go up this year?
I would expect some of that to be honest with you. Typically the way DRAM shortages tend to work is that they get passed on all the way up to the end consumer. Now how that ultimately plays out in terms of elasticity of demand, that'll be interesting to see. I would expect though if things continue, you're going to have to see some level of pricing increases passed on. There's no way around it.
Yeah. And that segues into my next question. I mean, you've been here for a day now. This is your first Davos, I believe.
A lot of the questions that have come across this year, we've seen a few mini market panics. Like, what's your take on whether we're in an AI bubble? Like, is the hype exceeding the reality or...
Oh gosh. The bubble can be defined a number of different ways, whether there's a valuation bubble, whether there's a bubble in terms of this level of demand. The way I look at where we are with AI in terms of impacting our lives, if I used a baseball vernacular, maybe not a good analogy to use, I'll use football, which is a 90-minute match, we may be only in the first 10 minutes, maybe even less. And why do I say that? Folks here who work for a large enterprise or even a small enterprise, ask yourselves how much of the workflows that you have inside your company are actually driven by real AI. And I'm not talking about somebody in finance using ChatGPT to generate a forecast, I'm talking about real workflows inside the enterprise. It's literally nowhere. Yet you still have 800 million users using ChatGPT, which is an insane number. I do believe that enterprise is going to be a gigantic application space and it's really nowhere. Part of this is that it's not easy to transform it. When you think about workflows, whether it's around SAP or Salesforce or other things that you have inside the enterprise, transitioning those to full AI, that's going to take a lot of time. Now, are we seeing the spend potentially getting ahead of itself relative to demand? I think in the short term we'll probably see some of that, but in the long term, I think about the internet bubble that took place and I remember an investment that I made in the late 1990s. My broker at the time said, 'Hey, there's a company that their business model is they're going to deliver groceries that you can order on the internet, company's called Webvan.' I said, 'Well, that sounds like a good idea, who would not want groceries ordered on the internet?' Business model was a disaster. It was before 3G, 4G phones, you had no GPS, no mobile phones, so everything was done on the desktop. So Webvan was a colossal failure, but buying groceries online using the internet, not a bad idea. And AI dwarfs that in terms of its actual capability and efficiency. So no, when I hear that, I can't even get my head around it just in the sense of when I think about what a utility artificial intelligence will add to our lives.
So it doesn't surprise you that we're seeing private valuations for Anthropic at 350 billion, you know, OpenAI at 500.
You know, these are plausible numbers to you, in particular if they IPO this year.
If a decade or two ago when GE was the largest market cap company and someone would have told Jack Welch that Nvidia was going to be $5 trillion in 15 to 20 years, they would have said, 'What does Nvidia do?' Right? So no, I don't think, I mean, the numbers are crazy numbers relative to what we've seen, but I think what we've also seen, particularly with tech innovation, is that when we look back 10 years on something that's as profound as what we're seeing now, the numbers actually look small. And that's what I believe here.
Really well. I mean, when we were at Davos last year, you had Satya Nadella, the CEO of Microsoft, talking about spending 80 billion a year on largely data centers, right? Many of which will involve Nvidia GPUs and your CPUs. Microsoft is now forecasting 140 billion, you know, with similar escalation in numbers at Google and AWS, not to mention what OpenAI is spending on projects like Stargate, which you're involved in. Are you seeing any softening in demand as we head into 2026, in particular with all the geopolitical noise going on?
I think the numbers do need to square themselves on some level. You know, back to do we see any fall back or drawback in demand, back to my enterprise example, AI is largely not being used in the enterprise for real work and real work improvement. So that tells me that there's huge demand for AI capability, but on the flip side, these are large numbers. They are so large in terms of relative to even a Microsoft or a Google or a Meta or an Amazon with giant balance sheets at some point. How do you actually scale the capital for return? So will you start to see sovereign investments into, whether it's the US or whether it's other parts of the world? I think you're going to start to see models that start to evolve in terms of how to maintain that level of investment. And then there's always this race with China, you know, how to think about that. China has a number of things that the West doesn't really have that they're really good at. China's very good at access to energy. And China's, because they don't have an antiquated grid, there's not a lot of regulations when they need to put up a new power plant or build new transmission lines. And they build things very fast. So I think there is feelings amongst people either in government or in the market of how do we square this giant investment required, which is beyond the balance sheet of a lot of companies today, versus there's a race going on here in terms of how to get the productivity sooner.
Yeah. And people talking about this, in particular since we're sat here in Europe, there's usually a lot of grousing about how difficult it is to build things in Europe versus America. Whether this is, you know, getting the construction permissions, linking them up to the electricity grid, you know, what is the sense talking to your fellow executives? I mean, is Europe's stance changing on that? Are they true believers in AI like the Trump administration is in the US and are trying to remove barriers, or are we still seeing a little bit of reticence?
I definitely get a sense, and it's interesting, I was at the AI safety summit in Paris, which was I think a couple months after the Davos event, and at that time I sensed a reticence from Europe on investment. There was a lot around safety, around measures to control what was going on with AI, and the US was very much in the west along with the UK actually of continuing to invest heavily. This was a few months again as I said after Trump took office. Fast forward seven, eight months from that, I sense a difference from European leaders who do recognize that the AI opportunity is real and will have real benefits for society, but also how to measure that against not punishing the citizens who rely on the grid for energy, etc., etc., and if we build new energy sources that we don't see an impact to them. So, no answers, but I definitely get a sense that it's moved more towards the 'we need to solve this' as opposed to 'we're thinking about safety first versus solving.'
Yeah. So, as we all know, President Trump will arrive tomorrow. There's a lot of stuff going on, whether it's Greenland, you know, battling with its fellow NATO allies over the future of Greenland, you know, the regime change in Venezuela. As a global chip maker, are you worried about potentially more tariffs, restrictions on chips? You know, is a trade war potentially going to derail 2026 for ARM?
Yeah, I have a lot to worry about on my plate besides Venezuela and Greenland, but at the same time, obviously, we watch what's going on with tariffs pretty carefully. What I've seen from the administration has been ultimately, in my opinion, a willingness ultimately to get deals done. There was a lot of noise last year about Japan and Korea and Taiwan and we've solved those, right? The US has negotiated trade deals with each one of those countries. So I have no inside information. But I will just say I'm an optimist that ultimately the administration will find a way to get deals done.
Yeah. And the Trump administration has obviously been very keen to move manufacturing capacity like from places like Taiwan back to the US, in particular Intel, you know, being one of them. Has that been successful? I mean, you know, tech CEOs are fond of going to the White House and saying that they're going to invest X hundred billion, you know, which Trump likes, but are they actually genuinely following through? Are we seeing manufacturing capacity come back to the US and is that happening for you guys as well?
Slowly. And when I say slowly, Taiwan Semiconductor, TSMC, the world's leader in building chips for our industry, magnificent what they do, was a 30-year venture. And they weren't really the leaders until probably 5 to 10 years ago, maybe less depending on where the numbers line up. The bottom line is it takes a lot of time. And the US fell behind badly and investment moved its way offshore. There's definitely an appetite from the administration to accelerate this and you see that with the US investment in Intel, right? $5 billion into Intel. But it will take time. But on the flip side, I'm happy to see it starting. Not so much from a nationalistic standpoint, and ARM is a UK company, so we participate globally, but I think supply chain resiliency from a geographic standpoint is mandatory. I think single point of failure where the world's leading fabs sit in one part of the world is not healthy. And we saw this during COVID on a minor level when it took 52 weeks to get a key fob back if you lost your keys. And that was because they were built in one place and the supply chain was linked to one area. So simply for supply chain resiliency, it's quite necessary and I think it's the right strategy in the long term.
Yeah. You're doing some interesting things behind the scenes at the moment. The FT has reported that you're in talks with Meta to build your own chip. So, ARM's business model for a long time has been to provide blueprints and designs for CPUs that are then built by TSMC for other companies. And if you believe what's written in the FT, you're going to start making your own chips, thereby going into competition with some of your partners. You know, can you comment on where this Meta chip is in development and how you're going to manage the potentially, you know, are you going to manage a frictionless transition from partner to competitor with people like Nvidia?
That's why I'm here. But on that particular point, I'm not going to comment on products that we haven't announced or will not announce or may not announce. But I will say that for years we've had a lot of people asking us to do this. And you'd say, well, why would they want ARM to possibly provide an end product? Largely because the technology is so ubiquitous. The software ecosystem relies on supply. And what do I mean by that? Because we build a CPU. The CPU platform is really only as good as the software ecosystem around it, the software that's been invested in it. And there are certain markets and certain areas where we have partners who say, 'I wish we had more optionality where more software could be written.' So we know that's a high order bid and we have had customers come to us and ask, could we do this for them, and I'll leave it there for now.
Okay. But if you do go down that route, which it sounds like you are, it will change the shape of the business. You're going to go from a very high margin, you know, royalties and blueprint business to actually making your own chips, which will generate more revenue, but come with higher costs and indeed risks and put you, as I mentioned before, in competition with a lot of other very powerful companies like Nvidia, you know, sort of Nvidia, and increasingly with Qualcomm, etc. Is that how you're going to be the next phase of growth? I mean, I think your market cap, the last time I checked, was 110, 120 billion.
Yeah, I think we're green. I think we're green today. My CFO sent me a nice message saying in a sea of red, we're green today.
How do you bridge the gap with Nvidia?
That's 4.9 trillion at the moment.
