Rene Haas

This doesn't necessarily need to be chat GPT-5 running six months of training to figure out the next level of sophistication. But this could be just now you want to run a small level of inference that is helping the AI model run wherever it's at. So we are seeing AI workloads, as I said, running absolutely everywhere. So what does that mean for ARM?

This doesn't necessarily need to be chat GPT-5 running six months of training to figure out the next level of sophistication. But this could be just now you want to run a small level of inference that is helping the AI model run wherever it's at. So we are seeing AI workloads, as I said, running absolutely everywhere. So what does that mean for ARM?

So our core business is around CPUs, but we also do GPUs. We also do NPUs, neural processing engines. And what we are seeing is the need to add more and more compute capability to accelerate these AI workloads. We're seeing that kind of as table stakes. Either put a neural engine inside the GPU that can run acceleration or make the CPU more capable to run extensions that can accelerate your AI.

So our core business is around CPUs, but we also do GPUs. We also do NPUs, neural processing engines. And what we are seeing is the need to add more and more compute capability to accelerate these AI workloads. We're seeing that kind of as table stakes. Either put a neural engine inside the GPU that can run acceleration or make the CPU more capable to run extensions that can accelerate your AI.

We are seeing that everywhere. And I think that I wouldn't even say that's going to accelerate. That now is going to be the default. So what you're going to have is from the tiniest of devices at the edge to the most sophisticated data centers, an AI workload is going to be running on top of everything else that you had to do, right?

We are seeing that everywhere. And I think that I wouldn't even say that's going to accelerate. That now is going to be the default. So what you're going to have is from the tiniest of devices at the edge to the most sophisticated data centers, an AI workload is going to be running on top of everything else that you had to do, right?

So if you look at a mobile phone or a PC that has to run graphics, it has to run a game, it has to run the operating system, it has to run the apps. And oh, by the way, it now needs to run some level of co-pilot or it needs to run an agent. It's good for us because what that means is I need more and more compute capability inside a system that's already kind of constrained on cost.

So if you look at a mobile phone or a PC that has to run graphics, it has to run a game, it has to run the operating system, it has to run the apps. And oh, by the way, it now needs to run some level of co-pilot or it needs to run an agent. It's good for us because what that means is I need more and more compute capability inside a system that's already kind of constrained on cost.

It's kind of constrained on size. It's kind of constrained on area. But it's great for us because it gives us a bunch of hard problems to go off and solve. But that's clearly what we're seeing. So I would say AI is everywhere.

It's kind of constrained on size. It's kind of constrained on area. But it's great for us because it gives us a bunch of hard problems to go off and solve. But that's clearly what we're seeing. So I would say AI is everywhere.

And I think there's two reasons for that. One is the... models and the capabilities are advancing very fast. And the capability of the model is advancing how you manage the balance between what runs locally, what runs in the cloud, things around latency and security. It's moving at an incredible pace.

And I think there's two reasons for that. One is the... models and the capabilities are advancing very fast. And the capability of the model is advancing how you manage the balance between what runs locally, what runs in the cloud, things around latency and security. It's moving at an incredible pace.

I think OpenAI, and I was in a discussion with the OpenAI guys last week, they're doing the 12 days of Christmas. 12 days of shipments. 12 days of shipments, yeah. And they're doing something every day. It takes two or three years to develop a chip, right? So think about the chips that are in that new iPhone when they were conceived.

I think OpenAI, and I was in a discussion with the OpenAI guys last week, they're doing the 12 days of Christmas. 12 days of shipments. 12 days of shipments, yeah. And they're doing something every day. It takes two or three years to develop a chip, right? So think about the chips that are in that new iPhone when they were conceived.

and when they were designed and when the features that were thought about that had to go inside that phone. ChatGPT didn't even exist at that time. So I think this is going to be something that is the classic, it's going to be gradually and then it's suddenly. You're just going to see sort of a knee in the curve moment where the hardware is now sophisticated enough and then the apps rush in.

and when they were designed and when the features that were thought about that had to go inside that phone. ChatGPT didn't even exist at that time. So I think this is going to be something that is the classic, it's going to be gradually and then it's suddenly. You're just going to see sort of a knee in the curve moment where the hardware is now sophisticated enough and then the apps rush in.

Well, I think, as I said, one of the things that we're seeing is that whether it's a wearable or a PC or a phone or a car, the chips that are being designed are just being stuffed with as much compute capability to take advantage of what might be there. So it's a bit of chicken and egg relative to you load up the hardware with as much capability, hoping that the software lands on it.

Well, I think, as I said, one of the things that we're seeing is that whether it's a wearable or a PC or a phone or a car, the chips that are being designed are just being stuffed with as much compute capability to take advantage of what might be there. So it's a bit of chicken and egg relative to you load up the hardware with as much capability, hoping that the software lands on it.

And the software is innovating at a very, very rapid pace. But that intersection will come where suddenly, oh my gosh, I've shrunk the large language model down to a certain size. The chip that's going in this tiny wearable now has enough memory to take advantage of that model. And as a result, the magic takes over. And that will happen. It will be gradual and then sudden.

And the software is innovating at a very, very rapid pace. But that intersection will come where suddenly, oh my gosh, I've shrunk the large language model down to a certain size. The chip that's going in this tiny wearable now has enough memory to take advantage of that model. And as a result, the magic takes over. And that will happen. It will be gradual and then sudden.

Yeah, I do. It's interesting because many of the markets that we have been involved in, whether it's mainframes to PCs to mobile and then wearables or watches... Some new form factor drives some new level of innovation. It's hard to say what that new next form factor looks like.

Yeah, I do. It's interesting because many of the markets that we have been involved in, whether it's mainframes to PCs to mobile and then wearables or watches... Some new form factor drives some new level of innovation. It's hard to say what that new next form factor looks like.

So I think it's going to be more of a hybrid situation, whether it's around the glasses or around devices in your home that are more of a push device than a pull device instead of... asking Alexa or asking Google Assistant what to do, you may have that information pushed to you. You may not want it pushed to you, but it could get pushed to you in such a way that's looking around corners for you.

So I think it's going to be more of a hybrid situation, whether it's around the glasses or around devices in your home that are more of a push device than a pull device instead of... asking Alexa or asking Google Assistant what to do, you may have that information pushed to you. You may not want it pushed to you, but it could get pushed to you in such a way that's looking around corners for you.

And I think the form factor that that comes in, I think will be somewhat similar to what we're seeing today, but you may see some of these devices get just much more intelligent in terms of, as I said, in the push level.

And I think the form factor that that comes in, I think will be somewhat similar to what we're seeing today, but you may see some of these devices get just much more intelligent in terms of, as I said, in the push level.

Well, the amount of investment that's going on is through the roof. You just have to look at the numbers of some of the folks who are in this industry. And I think it's a very interesting time because right now we're still seeing an insatiable investment in training. Training is hugely compute intensive. It's hugely power intensive. And that's driving a lot of the growth.

Well, the amount of investment that's going on is through the roof. You just have to look at the numbers of some of the folks who are in this industry. And I think it's a very interesting time because right now we're still seeing an insatiable investment in training. Training is hugely compute intensive. It's hugely power intensive. And that's driving a lot of the growth.

But the level of compute that will be required for inference is actually going to be much larger. I think it'll be Better than half, maybe 80% over time would be inference. But the amount of inference cases that we'll need to run are far larger than what we have today.

But the level of compute that will be required for inference is actually going to be much larger. I think it'll be Better than half, maybe 80% over time would be inference. But the amount of inference cases that we'll need to run are far larger than what we have today.

So I think that's why you're seeing companies like CoreWeave and Oracle and people who are not traditionally in this space who are now running AI cloud. Well, why is that? Because... capacity. The traditional large hyperscalers, the Amazons, the Metas, the Googles, the Microsofts, there's just not enough capacity.

So I think that's why you're seeing companies like CoreWeave and Oracle and people who are not traditionally in this space who are now running AI cloud. Well, why is that? Because... capacity. The traditional large hyperscalers, the Amazons, the Metas, the Googles, the Microsofts, there's just not enough capacity.

So what I think we'll continue to see is a changing of the landscape, maybe not a changing so much, but certainly opportunities for other players in terms of enabling and accessing this growth. And for ARM, it's very, very good because we've seen a very, very large increase in growth in market share for us in the data center, AWS at reInvent,

So what I think we'll continue to see is a changing of the landscape, maybe not a changing so much, but certainly opportunities for other players in terms of enabling and accessing this growth. And for ARM, it's very, very good because we've seen a very, very large increase in growth in market share for us in the data center, AWS at reInvent,

this week who build their general purpose devices, Graviton, based on Arm. They say that 50% of all new deployments are Graviton. So 50% of anything new at AWS is Arm. And that's not going to decrease. That number is just going to go up. And I think one of the things we're seeing, whether it's devices like Grace Blackwell from NVIDIA.

this week who build their general purpose devices, Graviton, based on Arm. They say that 50% of all new deployments are Graviton. So 50% of anything new at AWS is Arm. And that's not going to decrease. That number is just going to go up. And I think one of the things we're seeing, whether it's devices like Grace Blackwell from NVIDIA.

Grace, which is the CPU, and that's ARM, using an NVIDIA GPU, that's a big benefit for us because what happens is the AI cloud is now running a host node based on ARM. And if the data center now has an AI cluster where the general purpose compute is ARM, They naturally want to have as much of the general purpose compute that's not AI running on ARM.

Grace, which is the CPU, and that's ARM, using an NVIDIA GPU, that's a big benefit for us because what happens is the AI cloud is now running a host node based on ARM. And if the data center now has an AI cluster where the general purpose compute is ARM, They naturally want to have as much of the general purpose compute that's not AI running on ARM.

So what we're seeing is just an acceleration for us in the data center, whether it's AI or inference or general purpose compute.

So what we're seeing is just an acceleration for us in the data center, whether it's AI or inference or general purpose compute.

On one hand, it would be crazy to say that growth continues unabated, right? We've seen, obviously, that that is never really the case. I think what will get very interesting in this particular growth phase is to what level does real benefit come from AI that can augment and or replace certain levels of jobs? You know, some of the AI models and chatbots today are decent, but not great.

On one hand, it would be crazy to say that growth continues unabated, right? We've seen, obviously, that that is never really the case. I think what will get very interesting in this particular growth phase is to what level does real benefit come from AI that can augment and or replace certain levels of jobs? You know, some of the AI models and chatbots today are decent, but not great.

They supplement work, but they don't necessarily replace work. But if you start to get into agents that can do real level of work that can replace what people might need to do in terms of thinking and reasoning, then that gets fairly interesting. And then you say, well, how's that going to happen? Well, We're not there yet, so we need to train more models.

They supplement work, but they don't necessarily replace work. But if you start to get into agents that can do real level of work that can replace what people might need to do in terms of thinking and reasoning, then that gets fairly interesting. And then you say, well, how's that going to happen? Well, We're not there yet, so we need to train more models.

The models need to get more sophisticated, et cetera, et cetera. So I think the training thing continues for a bit, but I can see as we get to some level of AI agent that reasons close to the way a human does, then I think it asymptotes on some level.

The models need to get more sophisticated, et cetera, et cetera. So I think the training thing continues for a bit, but I can see as we get to some level of AI agent that reasons close to the way a human does, then I think it asymptotes on some level.

I don't think training can be unabated because at some point in time, you get more now into specialized training models as opposed to general purpose models, and that requires less resources.

I don't think training can be unabated because at some point in time, you get more now into specialized training models as opposed to general purpose models, and that requires less resources.

I know he has his own definitions for AGI and he has reasons for those definitions. I don't subscribe so much to what is AGI versus ASI, artificial superintelligence, but I think more around when these AI agents start to think and reason and invent. And to me, that is a bit of a cross the Rubicon moment, right? For example, Chat GPT can do a decent job of passing the bar exam.

I know he has his own definitions for AGI and he has reasons for those definitions. I don't subscribe so much to what is AGI versus ASI, artificial superintelligence, but I think more around when these AI agents start to think and reason and invent. And to me, that is a bit of a cross the Rubicon moment, right? For example, Chat GPT can do a decent job of passing the bar exam.

But to some extent, you'd say load enough logic and load enough information into the model, and the answers are there somewhere. And to what level is the AI model a stochastic parrot and just repeats everything that it has found over the internet? Because at the end of the day, you're only as good as the model that you've trained on is only as good as the data.

But to some extent, you'd say load enough logic and load enough information into the model, and the answers are there somewhere. And to what level is the AI model a stochastic parrot and just repeats everything that it has found over the internet? Because at the end of the day, you're only as good as the model that you've trained on is only as good as the data.

But when the model gets to a point where it can think and reason and invent, create new concepts, new products, new ideas, to me, that's kind of AGI when you get to that level. And I think, I don't know if we're a year away, but I would say we are a lot closer. If you would ask me this question a year ago, I would have said it's quite a ways away.

But when the model gets to a point where it can think and reason and invent, create new concepts, new products, new ideas, to me, that's kind of AGI when you get to that level. And I think, I don't know if we're a year away, but I would say we are a lot closer. If you would ask me this question a year ago, I would have said it's quite a ways away.

You ask me that question now, I say it's much closer. What is much closer, two years, three years? Probably. And I'm probably going to be wrong on that front. You know, every time I interact with some of the partners who are working on their models, whether it's at Google or OpenAI, and they show us the demos, it's breathtaking in terms of the kind of advancements that they're making.

You ask me that question now, I say it's much closer. What is much closer, two years, three years? Probably. And I'm probably going to be wrong on that front. You know, every time I interact with some of the partners who are working on their models, whether it's at Google or OpenAI, and they show us the demos, it's breathtaking in terms of the kind of advancements that they're making.

So, yeah, I think to getting to a model that can think and reason and invent, we're not that far away.

So, yeah, I think to getting to a model that can think and reason and invent, we're not that far away.

Yeah, this is going to sound like one of those if I did it answers, right? So I got to be thinking about why would Arm consider doing something other than what it currently does? I'll go back to the first discussion we were having relative to AI workloads. What we are seeing consistently is that AI workloads are being intertwined with everything that is taking place from a software standpoint.

Yeah, this is going to sound like one of those if I did it answers, right? So I got to be thinking about why would Arm consider doing something other than what it currently does? I'll go back to the first discussion we were having relative to AI workloads. What we are seeing consistently is that AI workloads are being intertwined with everything that is taking place from a software standpoint.

So we are, at our core, we are a computer architecture. That's what we do. We have great products, our CPUs are wonderful, our GPUs are wonderful, but our products are nothing without software. The software is what makes our engine go. If you are defining a computer architecture and you're building the future of computing,

So we are, at our core, we are a computer architecture. That's what we do. We have great products, our CPUs are wonderful, our GPUs are wonderful, but our products are nothing without software. The software is what makes our engine go. If you are defining a computer architecture and you're building the future of computing,

One of the things you need to be very mindful of is that link between hardware and software. And that link in terms of really understanding where the tradeoffs are being made, where the optimizations are being made, what are the ultimate benefits to consumers from a chip that has that type of integration, that is easier to do if you're building something than if you're licensing IP.

One of the things you need to be very mindful of is that link between hardware and software. And that link in terms of really understanding where the tradeoffs are being made, where the optimizations are being made, what are the ultimate benefits to consumers from a chip that has that type of integration, that is easier to do if you're building something than if you're licensing IP.

just from the standpoint that if you're building something, you're much closer to that interlock and you have a much better perspective in terms of the design trade-offs to make. So if we were to do something, that would be one of the reasons we might.

just from the standpoint that if you're building something, you're much closer to that interlock and you have a much better perspective in terms of the design trade-offs to make. So if we were to do something, that would be one of the reasons we might.

Well, I mean, my customers are Apple. I don't plan on building a phone. My customer is Tesla. I'm not going to build a car. My customer is Amazon. I'm not going to build a data center.

Well, I mean, my customers are Apple. I don't plan on building a phone. My customer is Tesla. I'm not going to build a car. My customer is Amazon. I'm not going to build a data center.

Well, he builds boxes, right? He builds DGX boxes, and he builds all kinds of stuff.

Well, he builds boxes, right? He builds DGX boxes, and he builds all kinds of stuff.

Yeah, lock-in is a... It's either one can look at it as an offensive maneuver that you take and I'm going to do these things so I can lock people in and or you provide an environment in such that it's just so easy to use your hardware that by default you then quote locked in. Let's go back to the AI workload commentary.

Yeah, lock-in is a... It's either one can look at it as an offensive maneuver that you take and I'm going to do these things so I can lock people in and or you provide an environment in such that it's just so easy to use your hardware that by default you then quote locked in. Let's go back to the AI workload commentary.

So today, if you're doing general purpose compute, you're writing your algorithms in C or JAX or something of that nature. And now you're, let's say, wanting to write something in TensorFlow or Python. In an ideal world, what does the software developer want?

So today, if you're doing general purpose compute, you're writing your algorithms in C or JAX or something of that nature. And now you're, let's say, wanting to write something in TensorFlow or Python. In an ideal world, what does the software developer want?

Software developer wants to be able to write their application at a very high level, whether that is a general purpose workload or an AI workload, and just have it work on the underlying hardware with not really having to know what the attributes are of the underlying hardware. I don't know if there's any software people in the room. Software people are wonderful.

Software developer wants to be able to write their application at a very high level, whether that is a general purpose workload or an AI workload, and just have it work on the underlying hardware with not really having to know what the attributes are of the underlying hardware. I don't know if there's any software people in the room. Software people are wonderful.

They are inherently lazy and they want to be able to just have their application run and have it work. So as a computer architecture platform, it's incumbent upon us to make that easy. So when we think about providing a heterogeneous platform and homogeneous across the software, that is a very big initiative for us. And we're doing it today.

They are inherently lazy and they want to be able to just have their application run and have it work. So as a computer architecture platform, it's incumbent upon us to make that easy. So when we think about providing a heterogeneous platform and homogeneous across the software, that is a very big initiative for us. And we're doing it today.

We have a technology called Clyde and these are Clyde libraries for AI. And they do that for the CPU. All the goodness that we put inside our CPU products that allows for acceleration by using the libraries, and we make those available open. There's no charge. Developers, it just works.

We have a technology called Clyde and these are Clyde libraries for AI. And they do that for the CPU. All the goodness that we put inside our CPU products that allows for acceleration by using the libraries, and we make those available open. There's no charge. Developers, it just works.

So going forward, since vast majority of the platforms today are ARM-based and the vast majority are going to run AI workloads, we just want to make that really easy for folks.

So going forward, since vast majority of the platforms today are ARM-based and the vast majority are going to run AI workloads, we just want to make that really easy for folks.

Yeah, so the current update is that it plans to go to trial on December 16, which isn't very far away. And I can appreciate, because we talk to investors and partners, that what do they hate the most is uncertainty. So that I can appreciate. But on the flip side, I would say the principles as to why we filed the claim are unchanged. And that's about all I can say.

Yeah, so the current update is that it plans to go to trial on December 16, which isn't very far away. And I can appreciate, because we talk to investors and partners, that what do they hate the most is uncertainty. So that I can appreciate. But on the flip side, I would say the principles as to why we filed the claim are unchanged. And that's about all I can say.

I think what surprised me on a personal level is the amount of bandwidth that it takes away from my day that I end up having to think about things that we weren't thinking about before. But at a highest level, it's actually not a big, big change. From the perspective of Arm, Arm was public before. When SoftBank bought us, we still consolidated up through SoftBank.

I think what surprised me on a personal level is the amount of bandwidth that it takes away from my day that I end up having to think about things that we weren't thinking about before. But at a highest level, it's actually not a big, big change. From the perspective of Arm, Arm was public before. When SoftBank bought us, we still consolidated up through SoftBank.

So the muscles in terms of being able to report quarterly earnings and have them reconciled with a timeframe, all of that, we had good muscle memory on that. So I think operationally for the company, we have great teams. I have a great finance team that's really good at doing that.

So the muscles in terms of being able to report quarterly earnings and have them reconciled with a timeframe, all of that, we had good muscle memory on that. So I think operationally for the company, we have great teams. I have a great finance team that's really good at doing that.

I think for me personally, it was just maybe the appreciation that there's now a chunk of my week that's dedicated to activities that I wasn't really working on before.

I think for me personally, it was just maybe the appreciation that there's now a chunk of my week that's dedicated to activities that I wasn't really working on before.

No, I am a big believer of not doing a lot of organizational changes. To me, organizational design follows your strategy, and strategy follows your vision. And if you think about the way you've heard me talk about Arm publicly for the last couple of years, that's pretty unchanged. So as a result, we haven't done much in terms of changing the organization.

No, I am a big believer of not doing a lot of organizational changes. To me, organizational design follows your strategy, and strategy follows your vision. And if you think about the way you've heard me talk about Arm publicly for the last couple of years, that's pretty unchanged. So as a result, we haven't done much in terms of changing the organization.

I think organization changes are horrendously disruptive. We're an 8,000-person company, so we're not gigantic. But if you do a gigantic organization change, it better have followed a big strategy change. Otherwise, you've got off-sites and team meetings and Zoom calls about talking about my new leaders. And if it's not in support of a change of strategy, it's a big waste of time.

I think organization changes are horrendously disruptive. We're an 8,000-person company, so we're not gigantic. But if you do a gigantic organization change, it better have followed a big strategy change. Otherwise, you've got off-sites and team meetings and Zoom calls about talking about my new leaders. And if it's not in support of a change of strategy, it's a big waste of time.

So I really try hard not to do much of that.

So I really try hard not to do much of that.

If we were to do that.

If we were to do that.

I don't know if there was one specific trade-off. I think in this job, a CEO, which I've been now, gosh, it'll be three years in February, you're constantly doing the mental trade-off of what needs to happen in the day-to-day versus what needs to happen five years from now. My proclivity just tends to be to think five years ahead as opposed to thinking one quarter ahead.

I don't know if there was one specific trade-off. I think in this job, a CEO, which I've been now, gosh, it'll be three years in February, you're constantly doing the mental trade-off of what needs to happen in the day-to-day versus what needs to happen five years from now. My proclivity just tends to be to think five years ahead as opposed to thinking one quarter ahead.

So I don't know if there's any major trade-off that I would say that I make, but what I'm constantly wrestling with is that balance between what is necessary in the day-to-day versus what needs to happen in the next five years. Because I've got great teams. I've got the engineering team is fantastic. The finance team is fantastic. Sales and marketing, they're great.

So I don't know if there's any major trade-off that I would say that I make, but what I'm constantly wrestling with is that balance between what is necessary in the day-to-day versus what needs to happen in the next five years. Because I've got great teams. I've got the engineering team is fantastic. The finance team is fantastic. Sales and marketing, they're great.

So in the day-to-day, there isn't a lot I can do to impact what those jobs are, but the jobs that I can impact is the next five years. So where I try to do is spend areas of time for me personally, of work only I can do. And if there's work that the team can do that I'm not going to add much to it, I try to stay away from it.

So in the day-to-day, there isn't a lot I can do to impact what those jobs are, but the jobs that I can impact is the next five years. So where I try to do is spend areas of time for me personally, of work only I can do. And if there's work that the team can do that I'm not going to add much to it, I try to stay away from it.

But that's the biggest trade-off I wrestle with is the day-to-day versus the future. How different does Arm look in five years? I think we don't know that we look very different as a company, but hopefully we are continuing to be an extremely impactful company in the industry. I have big ambitions for where we can be.

But that's the biggest trade-off I wrestle with is the day-to-day versus the future. How different does Arm look in five years? I think we don't know that we look very different as a company, but hopefully we are continuing to be an extremely impactful company in the industry. I have big ambitions for where we can be.

Yes. Yeah, he is. He's a fascinating guy. One of the things I admire a lot about Masa, and I don't think he gets enough credit for this, is that he's the CEO and founder of a 40-year-old company. And he's reinvented himself a lot of times. I mean, SoftBank started out as a distributor of software. And he's reinvented himself from being an operator, whether it's SoftBank Mobile, to an investor.

Yes. Yeah, he is. He's a fascinating guy. One of the things I admire a lot about Masa, and I don't think he gets enough credit for this, is that he's the CEO and founder of a 40-year-old company. And he's reinvented himself a lot of times. I mean, SoftBank started out as a distributor of software. And he's reinvented himself from being an operator, whether it's SoftBank Mobile, to an investor.

He's a joy to work with, to be honest with you. I learn a lot from him. He is very ambitious, obviously, loves to take risks. But at the same time, he has a good handle on the things that matter. So he's, yeah, I think everything you see about him is accurate. He's a very entertaining guy.

He's a joy to work with, to be honest with you. I learn a lot from him. He is very ambitious, obviously, loves to take risks. But at the same time, he has a good handle on the things that matter. So he's, yeah, I think everything you see about him is accurate. He's a very entertaining guy.

Well, he's the chairman of the company, right? He's the chairman of the board. So from that perspective, if you think about the board's job is to evaluate the long-term strategy of the company and my proclivity towards thinking, you know, also in the long-term, he and I talk all the time about those kinds of things.

Well, he's the chairman of the company, right? He's the chairman of the board. So from that perspective, if you think about the board's job is to evaluate the long-term strategy of the company and my proclivity towards thinking, you know, also in the long-term, he and I talk all the time about those kinds of things.

That's a wonderful question. I think Jensen, Masa, Larry Ellison, Jeff Bezos, I'm sure I'm leaving out names, people who built a company and are running it 20, 30 years later and drive it with the same level of passion and innovation, carry a lot of the same traits.

That's a wonderful question. I think Jensen, Masa, Larry Ellison, Jeff Bezos, I'm sure I'm leaving out names, people who built a company and are running it 20, 30 years later and drive it with the same level of passion and innovation, carry a lot of the same traits.

And those traits are obviously very intelligent, obviously brilliant, looking around corners, work incredibly hard, but have an incredible amount of courage. And I think that is all, those ingredients are necessary for people who stay at the top that long. I'm a big basketball fan and I've always drawn analogies between, if you think about basketball

And those traits are obviously very intelligent, obviously brilliant, looking around corners, work incredibly hard, but have an incredible amount of courage. And I think that is all, those ingredients are necessary for people who stay at the top that long. I'm a big basketball fan and I've always drawn analogies between, if you think about basketball

a Michael Jordan or a Kobe Bryant when people talk about what made them great. Obviously, their talent was off the roof and they had great athleticism, but it was something in their character and their drive that cut them in a different level. And I think Jensen, Massa, Larry Ellison, the other names I mentioned, they all fall in the same group. Elon Musk, obviously.

a Michael Jordan or a Kobe Bryant when people talk about what made them great. Obviously, their talent was off the roof and they had great athleticism, but it was something in their character and their drive that cut them in a different level. And I think Jensen, Massa, Larry Ellison, the other names I mentioned, they all fall in the same group. Elon Musk, obviously.

Oh, my gosh. I guess at the highest level, as someone who's been in the industry my whole career, it is a little sad to see what's happening from the perspective of Intel as an icon. The amount of innovation that Intel has provided, whether it's around computer architecture or fabrication technology, PC platforms, servers. Intel is an innovation powerhouse.

Oh, my gosh. I guess at the highest level, as someone who's been in the industry my whole career, it is a little sad to see what's happening from the perspective of Intel as an icon. The amount of innovation that Intel has provided, whether it's around computer architecture or fabrication technology, PC platforms, servers. Intel is an innovation powerhouse.

So to see the troubles they're going through is a little sad. But at the same time, you have to innovate in our industry. There are lots of tombstones of great tech companies that don't reinvent themselves. And I think Intel's biggest dilemma is just how to disassociate being either a vertical company or a fabulous company, to oversimplify.

So to see the troubles they're going through is a little sad. But at the same time, you have to innovate in our industry. There are lots of tombstones of great tech companies that don't reinvent themselves. And I think Intel's biggest dilemma is just how to disassociate being either a vertical company or a fabulous company, to oversimplify.

And I think that is the fork in the road that they've faced for the last probably decade, to be honest with you. And Pat had a strategy that was very clear that vertical was the way to win. I will say that in my opinion, that when he took that strategy on in 2021, that was not a three-year strategy. That's a five to 10-year strategy.

And I think that is the fork in the road that they've faced for the last probably decade, to be honest with you. And Pat had a strategy that was very clear that vertical was the way to win. I will say that in my opinion, that when he took that strategy on in 2021, that was not a three-year strategy. That's a five to 10-year strategy.

So now that that's been, at least he's gone and there's a new CEO to be brought in, that's the decision that has to be made. My personal bias says that Vertical integration is a pretty powerful thing. And if they could get that right, I think they would be in an amazing position. But the cost associated with it is so high that it may be too big of a hill to climb.

So now that that's been, at least he's gone and there's a new CEO to be brought in, that's the decision that has to be made. My personal bias says that Vertical integration is a pretty powerful thing. And if they could get that right, I think they would be in an amazing position. But the cost associated with it is so high that it may be too big of a hill to climb.

Well, a couple of things with Intel. I'm not going to comment on the rumors that we're going to buy them. But I think, again, if you're a vertically integrated company and the power of your strategy is the fact that you have a product and you have fabs, inherently, you have a potential huge advantage in terms of cost. versus the competition.

Well, a couple of things with Intel. I'm not going to comment on the rumors that we're going to buy them. But I think, again, if you're a vertically integrated company and the power of your strategy is the fact that you have a product and you have fabs, inherently, you have a potential huge advantage in terms of cost. versus the competition.

And when Pat was the CEO, I did tell him more than once, you ought to license ARM, because if you've got your own fabs, fabs are all about volume, and we can provide volume. I wasn't successful in convincing him to do that, but I do think that would not be a bad move for Intel.

And when Pat was the CEO, I did tell him more than once, you ought to license ARM, because if you've got your own fabs, fabs are all about volume, and we can provide volume. I wasn't successful in convincing him to do that, but I do think that would not be a bad move for Intel.

On the flip side, I think in terms of ARM working with Intel, we work really closely with TSMC, we work really closely with Samsung, and IFS is a very, very large effort for Intel in terms of external customers. So we work very closely with them to enable them to ensure that they have access to the latest technology. We have access to their design kits.

On the flip side, I think in terms of ARM working with Intel, we work really closely with TSMC, we work really closely with Samsung, and IFS is a very, very large effort for Intel in terms of external customers. So we work very closely with them to enable them to ensure that they have access to the latest technology. We have access to their design kits.

So we want external partners who want to build at Intel to be able to use the latest and greatest ARM technology. So on that context, we work closely with them.

So we want external partners who want to build at Intel to be able to use the latest and greatest ARM technology. So on that context, we work closely with them.

I do know him a little bit. Kudos to him. I think that's a pretty good thing. I think it's quite fascinating that if you go back eight years to Trump 1.0 in terms of where we were in December as he was starting to fill out,

I do know him a little bit. Kudos to him. I think that's a pretty good thing. I think it's quite fascinating that if you go back eight years to Trump 1.0 in terms of where we were in December as he was starting to fill out,

His cabinet choices and appointees it was a bit chaotic and also at the same time There wasn't a lot of representation from the tech world and I think this time around whether it's Elon whether it's David Whether it's Vivek. I know Larry Ellison has been very involved in terms of discussions with the administration. I think it's a good thing, to be honest with you.

His cabinet choices and appointees it was a bit chaotic and also at the same time There wasn't a lot of representation from the tech world and I think this time around whether it's Elon whether it's David Whether it's Vivek. I know Larry Ellison has been very involved in terms of discussions with the administration. I think it's a good thing, to be honest with you.

Having a seat at the table and having access to policy, I think is really good.

Having a seat at the table and having access to policy, I think is really good.

I would say, not just for our business, but let's talk about China for a moment. The economies of the two countries are so inextricably tied together that a separation of supply chain, a separation of technology is a really difficult thing to architect.

I would say, not just for our business, but let's talk about China for a moment. The economies of the two countries are so inextricably tied together that a separation of supply chain, a separation of technology is a really difficult thing to architect.

So I would just say that as this administration or any administration comes into play and looks at policy around export control and such, be mindful that a hard break isn't as easy as it might look on paper. And there's just a lot of levers to consider back and forth. We are one attribute in the supply chain. But if you think about what it takes to build a semiconductor chip... There's EDA tools.

So I would just say that as this administration or any administration comes into play and looks at policy around export control and such, be mindful that a hard break isn't as easy as it might look on paper. And there's just a lot of levers to consider back and forth. We are one attribute in the supply chain. But if you think about what it takes to build a semiconductor chip... There's EDA tools.

There's the IP from ARM. There's the fabrication. There's the companies like NVIDIA, Mediatek that build chips. But then there's raw materials that go into building the wafers and the ingots and the substrates, and they come from everywhere. So it's just such a complex problem that's so inextricably linked together that I don't believe there's a one-size-fits-all policy.

There's the IP from ARM. There's the fabrication. There's the companies like NVIDIA, Mediatek that build chips. But then there's raw materials that go into building the wafers and the ingots and the substrates, and they come from everywhere. So it's just such a complex problem that's so inextricably linked together that I don't believe there's a one-size-fits-all policy.

And I think administration's being open to understanding that There just needs to be a lot of balance in terms of any solution that's put forward.

And I think administration's being open to understanding that There just needs to be a lot of balance in terms of any solution that's put forward.

No. I mean, the only thing that's probably changed for us in China, and I would say that's probably for a lot of the world, is that China used to be a very rich market for startup companies. And venture capital flew around very freely. There was a lot of innovation and things of that nature. That has absolutely slowed down, whether that is the exit

No. I mean, the only thing that's probably changed for us in China, and I would say that's probably for a lot of the world, is that China used to be a very rich market for startup companies. And venture capital flew around very freely. There was a lot of innovation and things of that nature. That has absolutely slowed down, whether that is the exit

for these companies isn't as clear, whether from a stock market standpoint, whether it's getting access to key technology isn't as well understood. We've seen that definitely slow down. On the flip side, we've seen incredible growth in segments such as automotive. If you look at companies like BYD or even Xiaomi,

for these companies isn't as clear, whether from a stock market standpoint, whether it's getting access to key technology isn't as well understood. We've seen that definitely slow down. On the flip side, we've seen incredible growth in segments such as automotive. If you look at companies like BYD or even Xiaomi,

building EVs, the technology in those vehicles is just unbelievable in terms of its capability. And selfishly for us, they all run on ARM. And the reason for that is, and China is very pragmatic in terms of how they build their systems and products. They rely very heavily on the open source global ecosystem for software. And

building EVs, the technology in those vehicles is just unbelievable in terms of its capability. And selfishly for us, they all run on ARM. And the reason for that is, and China is very pragmatic in terms of how they build their systems and products. They rely very heavily on the open source global ecosystem for software. And

All of the software libraries have been tuned for ARM, whether it's ADAS or in the powertrain or IVI, that's all ARM based. So we have our automotive business in China is really strong.

All of the software libraries have been tuned for ARM, whether it's ADAS or in the powertrain or IVI, that's all ARM based. So we have our automotive business in China is really strong.

Not really. I think my personal view on this is that the threats of tariffs are a bit of a tool to get to the negotiating table. I think President-elect Trump has proven over time that he is a businessman, and tariffs, I think, are one lever to start a negotiation. We'll see where it goes, but in my belief, I'm not too worried about that.

Not really. I think my personal view on this is that the threats of tariffs are a bit of a tool to get to the negotiating table. I think President-elect Trump has proven over time that he is a businessman, and tariffs, I think, are one lever to start a negotiation. We'll see where it goes, but in my belief, I'm not too worried about that.

I don't think we need a government OpenAI Manhattan-type project. I think the work that's being done by OpenAI or Anthropic or even the work in the open source that's being driven by Meta with Llama, we're seeing fantastic innovation on that. I think if you look and say... Is the U.S. a leader in terms of foundation models and frontier models? Absolutely, yes.

I don't think we need a government OpenAI Manhattan-type project. I think the work that's being done by OpenAI or Anthropic or even the work in the open source that's being driven by Meta with Llama, we're seeing fantastic innovation on that. I think if you look and say... Is the U.S. a leader in terms of foundation models and frontier models? Absolutely, yes.

And that's being done without a government intervention. So on the AI theme, I don't think it's necessary, personally. On the area of fabs, going back to the question that you started me on with on Intel, spending $30 to $40 billion a year in CapEx for these leading-edge nodes, that is a hard... pill to swallow for any company.

And that's being done without a government intervention. So on the AI theme, I don't think it's necessary, personally. On the area of fabs, going back to the question that you started me on with on Intel, spending $30 to $40 billion a year in CapEx for these leading-edge nodes, that is a hard... pill to swallow for any company.

And that's why Chips Act, I think, was a good and necessary thing because building semiconductors is fundamental to our economic engine. We learned that during COVID when it took 52 weeks to get a key fob replaced in terms of everything going on with supply chain. So I think having supply chain resiliency is super important. I think it's super important on a global level.

And that's why Chips Act, I think, was a good and necessary thing because building semiconductors is fundamental to our economic engine. We learned that during COVID when it took 52 weeks to get a key fob replaced in terms of everything going on with supply chain. So I think having supply chain resiliency is super important. I think it's super important on a global level.

It's definitely important on a national level. So I was and am in favor of the CHIPS Act.

It's definitely important on a national level. So I was and am in favor of the CHIPS Act.

I think one of the things that's happening is a real raise of visibility on this talent issue. And I think putting more and more money into university programs around semiconductors, semiconductor research is helping. I think that for a number of years, semiconductors and particularly manufacturing was not seen as the most attractive of degrees to go off and get.

I think one of the things that's happening is a real raise of visibility on this talent issue. And I think putting more and more money into university programs around semiconductors, semiconductor research is helping. I think that for a number of years, semiconductors and particularly manufacturing was not seen as the most attractive of degrees to go off and get.

A lot of people were looking at software as a service and things in other areas. So I think we need to get back to that on the university level. Now, one could argue if AI bots and agents can come in and do meaningful work, maybe that helps. But

A lot of people were looking at software as a service and things in other areas. So I think we need to get back to that on the university level. Now, one could argue if AI bots and agents can come in and do meaningful work, maybe that helps. But

I think in the area of building chips and semiconductor process, that is very much an art as well as a science, and particularly around improving manufacturing yields. So I don't know if we have enough talent, but I know there's a lot of effort now going to try to bolster that.

I think in the area of building chips and semiconductor process, that is very much an art as well as a science, and particularly around improving manufacturing yields. So I don't know if we have enough talent, but I know there's a lot of effort now going to try to bolster that.

Yeah, I think one of the things we were talking about earlier was we are now a public company. We were not a public company in 2022. So one of the things I've learned as a public company is break out as little as you can possibly can. So nobody can ask you questions in terms of where things are going.

Yeah, I think one of the things we were talking about earlier was we are now a public company. We were not a public company in 2022. So one of the things I've learned as a public company is break out as little as you can possibly can. So nobody can ask you questions in terms of where things are going.

Yeah, I know you are. So I would say this, no, we don't break any of that stuff out. But what we are observing is that, and I think this is only going to accelerate, is whether you're talking about an AI data center or you're talking about an AirPod or a wearable in your ear, there's an AI workload that's now running and that's very clear.

Yeah, I know you are. So I would say this, no, we don't break any of that stuff out. But what we are observing is that, and I think this is only going to accelerate, is whether you're talking about an AI data center or you're talking about an AirPod or a wearable in your ear, there's an AI workload that's now running and that's very clear.

This doesn't necessarily need to be chat GPT-5 running six months of training to figure out the next level of sophistication. But this could be just now you want to run a small level of inference that is helping the AI model run wherever it's at. So we are seeing AI workloads, as I said, running absolutely everywhere. So what does that mean for ARM?

So our core business is around CPUs, but we also do GPUs. We also do NPUs, neural processing engines. And what we are seeing is the need to add more and more compute capability to accelerate these AI workloads. We're seeing that kind of as table stakes. Either put a neural engine inside the GPU that can run acceleration or make the CPU more capable to run extensions that can accelerate your AI.

We are seeing that everywhere. And I think that I wouldn't even say that's going to accelerate. That now is going to be the default. So what you're going to have is from the tiniest of devices at the edge to the most sophisticated data centers, an AI workload is going to be running on top of everything else that you had to do, right?

So if you look at a mobile phone or a PC that has to run graphics, it has to run a game, it has to run the operating system, it has to run the apps. And oh, by the way, it now needs to run some level of co-pilot or it needs to run an agent. It's good for us because what that means is I need more and more compute capability inside a system that's already kind of constrained on cost.

It's kind of constrained on size. It's kind of constrained on area. But it's great for us because it gives us a bunch of hard problems to go off and solve. But that's clearly what we're seeing. So I would say AI is everywhere.

And I think there's two reasons for that. One is the... models and the capabilities are advancing very fast. And the capability of the model is advancing how you manage the balance between what runs locally, what runs in the cloud, things around latency and security. It's moving at an incredible pace.

I think OpenAI, and I was in a discussion with the OpenAI guys last week, they're doing the 12 days of Christmas. 12 days of shipments. 12 days of shipments, yeah. And they're doing something every day. It takes two or three years to develop a chip, right? So think about the chips that are in that new iPhone when they were conceived.

and when they were designed and when the features that were thought about that had to go inside that phone. ChatGPT didn't even exist at that time. So I think this is going to be something that is the classic, it's going to be gradually and then it's suddenly. You're just going to see sort of a knee in the curve moment where the hardware is now sophisticated enough and then the apps rush in.

Well, I think, as I said, one of the things that we're seeing is that whether it's a wearable or a PC or a phone or a car, the chips that are being designed are just being stuffed with as much compute capability to take advantage of what might be there. So it's a bit of chicken and egg relative to you load up the hardware with as much capability, hoping that the software lands on it.

And the software is innovating at a very, very rapid pace. But that intersection will come where suddenly, oh my gosh, I've shrunk the large language model down to a certain size. The chip that's going in this tiny wearable now has enough memory to take advantage of that model. And as a result, the magic takes over. And that will happen. It will be gradual and then sudden.

Yeah, I do. It's interesting because many of the markets that we have been involved in, whether it's mainframes to PCs to mobile and then wearables or watches... Some new form factor drives some new level of innovation. It's hard to say what that new next form factor looks like.

So I think it's going to be more of a hybrid situation, whether it's around the glasses or around devices in your home that are more of a push device than a pull device instead of... asking Alexa or asking Google Assistant what to do, you may have that information pushed to you. You may not want it pushed to you, but it could get pushed to you in such a way that's looking around corners for you.

And I think the form factor that that comes in, I think will be somewhat similar to what we're seeing today, but you may see some of these devices get just much more intelligent in terms of, as I said, in the push level.

Well, the amount of investment that's going on is through the roof. You just have to look at the numbers of some of the folks who are in this industry. And I think it's a very interesting time because right now we're still seeing an insatiable investment in training. Training is hugely compute intensive. It's hugely power intensive. And that's driving a lot of the growth.

But the level of compute that will be required for inference is actually going to be much larger. I think it'll be Better than half, maybe 80% over time would be inference. But the amount of inference cases that we'll need to run are far larger than what we have today.

So I think that's why you're seeing companies like CoreWeave and Oracle and people who are not traditionally in this space who are now running AI cloud. Well, why is that? Because... capacity. The traditional large hyperscalers, the Amazons, the Metas, the Googles, the Microsofts, there's just not enough capacity.

So what I think we'll continue to see is a changing of the landscape, maybe not a changing so much, but certainly opportunities for other players in terms of enabling and accessing this growth. And for ARM, it's very, very good because we've seen a very, very large increase in growth in market share for us in the data center, AWS at reInvent,

this week who build their general purpose devices, Graviton, based on Arm. They say that 50% of all new deployments are Graviton. So 50% of anything new at AWS is Arm. And that's not going to decrease. That number is just going to go up. And I think one of the things we're seeing, whether it's devices like Grace Blackwell from NVIDIA.

Grace, which is the CPU, and that's ARM, using an NVIDIA GPU, that's a big benefit for us because what happens is the AI cloud is now running a host node based on ARM. And if the data center now has an AI cluster where the general purpose compute is ARM, They naturally want to have as much of the general purpose compute that's not AI running on ARM.

So what we're seeing is just an acceleration for us in the data center, whether it's AI or inference or general purpose compute.

On one hand, it would be crazy to say that growth continues unabated, right? We've seen, obviously, that that is never really the case. I think what will get very interesting in this particular growth phase is to what level does real benefit come from AI that can augment and or replace certain levels of jobs? You know, some of the AI models and chatbots today are decent, but not great.

They supplement work, but they don't necessarily replace work. But if you start to get into agents that can do real level of work that can replace what people might need to do in terms of thinking and reasoning, then that gets fairly interesting. And then you say, well, how's that going to happen? Well, We're not there yet, so we need to train more models.

The models need to get more sophisticated, et cetera, et cetera. So I think the training thing continues for a bit, but I can see as we get to some level of AI agent that reasons close to the way a human does, then I think it asymptotes on some level.

I don't think training can be unabated because at some point in time, you get more now into specialized training models as opposed to general purpose models, and that requires less resources.

I know he has his own definitions for AGI and he has reasons for those definitions. I don't subscribe so much to what is AGI versus ASI, artificial superintelligence, but I think more around when these AI agents start to think and reason and invent. And to me, that is a bit of a cross the Rubicon moment, right? For example, Chat GPT can do a decent job of passing the bar exam.

But to some extent, you'd say load enough logic and load enough information into the model, and the answers are there somewhere. And to what level is the AI model a stochastic parrot and just repeats everything that it has found over the internet? Because at the end of the day, you're only as good as the model that you've trained on is only as good as the data.

But when the model gets to a point where it can think and reason and invent, create new concepts, new products, new ideas, to me, that's kind of AGI when you get to that level. And I think, I don't know if we're a year away, but I would say we are a lot closer. If you would ask me this question a year ago, I would have said it's quite a ways away.

You ask me that question now, I say it's much closer. What is much closer, two years, three years? Probably. And I'm probably going to be wrong on that front. You know, every time I interact with some of the partners who are working on their models, whether it's at Google or OpenAI, and they show us the demos, it's breathtaking in terms of the kind of advancements that they're making.

So, yeah, I think to getting to a model that can think and reason and invent, we're not that far away.

Yeah, this is going to sound like one of those if I did it answers, right? So I got to be thinking about why would Arm consider doing something other than what it currently does? I'll go back to the first discussion we were having relative to AI workloads. What we are seeing consistently is that AI workloads are being intertwined with everything that is taking place from a software standpoint.

So we are, at our core, we are a computer architecture. That's what we do. We have great products, our CPUs are wonderful, our GPUs are wonderful, but our products are nothing without software. The software is what makes our engine go. If you are defining a computer architecture and you're building the future of computing,

One of the things you need to be very mindful of is that link between hardware and software. And that link in terms of really understanding where the tradeoffs are being made, where the optimizations are being made, what are the ultimate benefits to consumers from a chip that has that type of integration, that is easier to do if you're building something than if you're licensing IP.

just from the standpoint that if you're building something, you're much closer to that interlock and you have a much better perspective in terms of the design trade-offs to make. So if we were to do something, that would be one of the reasons we might.

Well, I mean, my customers are Apple. I don't plan on building a phone. My customer is Tesla. I'm not going to build a car. My customer is Amazon. I'm not going to build a data center.

Well, he builds boxes, right? He builds DGX boxes, and he builds all kinds of stuff.

Yeah, lock-in is a... It's either one can look at it as an offensive maneuver that you take and I'm going to do these things so I can lock people in and or you provide an environment in such that it's just so easy to use your hardware that by default you then quote locked in. Let's go back to the AI workload commentary.

So today, if you're doing general purpose compute, you're writing your algorithms in C or JAX or something of that nature. And now you're, let's say, wanting to write something in TensorFlow or Python. In an ideal world, what does the software developer want?

Software developer wants to be able to write their application at a very high level, whether that is a general purpose workload or an AI workload, and just have it work on the underlying hardware with not really having to know what the attributes are of the underlying hardware. I don't know if there's any software people in the room. Software people are wonderful.

They are inherently lazy and they want to be able to just have their application run and have it work. So as a computer architecture platform, it's incumbent upon us to make that easy. So when we think about providing a heterogeneous platform and homogeneous across the software, that is a very big initiative for us. And we're doing it today.

We have a technology called Clyde and these are Clyde libraries for AI. And they do that for the CPU. All the goodness that we put inside our CPU products that allows for acceleration by using the libraries, and we make those available open. There's no charge. Developers, it just works.

So going forward, since vast majority of the platforms today are ARM-based and the vast majority are going to run AI workloads, we just want to make that really easy for folks.

Yeah, so the current update is that it plans to go to trial on December 16, which isn't very far away. And I can appreciate, because we talk to investors and partners, that what do they hate the most is uncertainty. So that I can appreciate. But on the flip side, I would say the principles as to why we filed the claim are unchanged. And that's about all I can say.

I think what surprised me on a personal level is the amount of bandwidth that it takes away from my day that I end up having to think about things that we weren't thinking about before. But at a highest level, it's actually not a big, big change. From the perspective of Arm, Arm was public before. When SoftBank bought us, we still consolidated up through SoftBank.

So the muscles in terms of being able to report quarterly earnings and have them reconciled with a timeframe, all of that, we had good muscle memory on that. So I think operationally for the company, we have great teams. I have a great finance team that's really good at doing that.

I think for me personally, it was just maybe the appreciation that there's now a chunk of my week that's dedicated to activities that I wasn't really working on before.

No, I am a big believer of not doing a lot of organizational changes. To me, organizational design follows your strategy, and strategy follows your vision. And if you think about the way you've heard me talk about Arm publicly for the last couple of years, that's pretty unchanged. So as a result, we haven't done much in terms of changing the organization.

I think organization changes are horrendously disruptive. We're an 8,000-person company, so we're not gigantic. But if you do a gigantic organization change, it better have followed a big strategy change. Otherwise, you've got off-sites and team meetings and Zoom calls about talking about my new leaders. And if it's not in support of a change of strategy, it's a big waste of time.

So I really try hard not to do much of that.

If we were to do that.

I don't know if there was one specific trade-off. I think in this job, a CEO, which I've been now, gosh, it'll be three years in February, you're constantly doing the mental trade-off of what needs to happen in the day-to-day versus what needs to happen five years from now. My proclivity just tends to be to think five years ahead as opposed to thinking one quarter ahead.

So I don't know if there's any major trade-off that I would say that I make, but what I'm constantly wrestling with is that balance between what is necessary in the day-to-day versus what needs to happen in the next five years. Because I've got great teams. I've got the engineering team is fantastic. The finance team is fantastic. Sales and marketing, they're great.

So in the day-to-day, there isn't a lot I can do to impact what those jobs are, but the jobs that I can impact is the next five years. So where I try to do is spend areas of time for me personally, of work only I can do. And if there's work that the team can do that I'm not going to add much to it, I try to stay away from it.

But that's the biggest trade-off I wrestle with is the day-to-day versus the future. How different does Arm look in five years? I think we don't know that we look very different as a company, but hopefully we are continuing to be an extremely impactful company in the industry. I have big ambitions for where we can be.

Yes. Yeah, he is. He's a fascinating guy. One of the things I admire a lot about Masa, and I don't think he gets enough credit for this, is that he's the CEO and founder of a 40-year-old company. And he's reinvented himself a lot of times. I mean, SoftBank started out as a distributor of software. And he's reinvented himself from being an operator, whether it's SoftBank Mobile, to an investor.

He's a joy to work with, to be honest with you. I learn a lot from him. He is very ambitious, obviously, loves to take risks. But at the same time, he has a good handle on the things that matter. So he's, yeah, I think everything you see about him is accurate. He's a very entertaining guy.

Well, he's the chairman of the company, right? He's the chairman of the board. So from that perspective, if you think about the board's job is to evaluate the long-term strategy of the company and my proclivity towards thinking, you know, also in the long-term, he and I talk all the time about those kinds of things.

That's a wonderful question. I think Jensen, Masa, Larry Ellison, Jeff Bezos, I'm sure I'm leaving out names, people who built a company and are running it 20, 30 years later and drive it with the same level of passion and innovation, carry a lot of the same traits.

And those traits are obviously very intelligent, obviously brilliant, looking around corners, work incredibly hard, but have an incredible amount of courage. And I think that is all, those ingredients are necessary for people who stay at the top that long. I'm a big basketball fan and I've always drawn analogies between, if you think about basketball

a Michael Jordan or a Kobe Bryant when people talk about what made them great. Obviously, their talent was off the roof and they had great athleticism, but it was something in their character and their drive that cut them in a different level. And I think Jensen, Massa, Larry Ellison, the other names I mentioned, they all fall in the same group. Elon Musk, obviously.

Oh, my gosh. I guess at the highest level, as someone who's been in the industry my whole career, it is a little sad to see what's happening from the perspective of Intel as an icon. The amount of innovation that Intel has provided, whether it's around computer architecture or fabrication technology, PC platforms, servers. Intel is an innovation powerhouse.

So to see the troubles they're going through is a little sad. But at the same time, you have to innovate in our industry. There are lots of tombstones of great tech companies that don't reinvent themselves. And I think Intel's biggest dilemma is just how to disassociate being either a vertical company or a fabulous company, to oversimplify.

And I think that is the fork in the road that they've faced for the last probably decade, to be honest with you. And Pat had a strategy that was very clear that vertical was the way to win. I will say that in my opinion, that when he took that strategy on in 2021, that was not a three-year strategy. That's a five to 10-year strategy.

So now that that's been, at least he's gone and there's a new CEO to be brought in, that's the decision that has to be made. My personal bias says that Vertical integration is a pretty powerful thing. And if they could get that right, I think they would be in an amazing position. But the cost associated with it is so high that it may be too big of a hill to climb.

Well, a couple of things with Intel. I'm not going to comment on the rumors that we're going to buy them. But I think, again, if you're a vertically integrated company and the power of your strategy is the fact that you have a product and you have fabs, inherently, you have a potential huge advantage in terms of cost. versus the competition.

And when Pat was the CEO, I did tell him more than once, you ought to license ARM, because if you've got your own fabs, fabs are all about volume, and we can provide volume. I wasn't successful in convincing him to do that, but I do think that would not be a bad move for Intel.

On the flip side, I think in terms of ARM working with Intel, we work really closely with TSMC, we work really closely with Samsung, and IFS is a very, very large effort for Intel in terms of external customers. So we work very closely with them to enable them to ensure that they have access to the latest technology. We have access to their design kits.

So we want external partners who want to build at Intel to be able to use the latest and greatest ARM technology. So on that context, we work closely with them.

I do know him a little bit. Kudos to him. I think that's a pretty good thing. I think it's quite fascinating that if you go back eight years to Trump 1.0 in terms of where we were in December as he was starting to fill out,

His cabinet choices and appointees it was a bit chaotic and also at the same time There wasn't a lot of representation from the tech world and I think this time around whether it's Elon whether it's David Whether it's Vivek. I know Larry Ellison has been very involved in terms of discussions with the administration. I think it's a good thing, to be honest with you.

Having a seat at the table and having access to policy, I think is really good.

I would say, not just for our business, but let's talk about China for a moment. The economies of the two countries are so inextricably tied together that a separation of supply chain, a separation of technology is a really difficult thing to architect.

So I would just say that as this administration or any administration comes into play and looks at policy around export control and such, be mindful that a hard break isn't as easy as it might look on paper. And there's just a lot of levers to consider back and forth. We are one attribute in the supply chain. But if you think about what it takes to build a semiconductor chip... There's EDA tools.

There's the IP from ARM. There's the fabrication. There's the companies like NVIDIA, Mediatek that build chips. But then there's raw materials that go into building the wafers and the ingots and the substrates, and they come from everywhere. So it's just such a complex problem that's so inextricably linked together that I don't believe there's a one-size-fits-all policy.

And I think administration's being open to understanding that There just needs to be a lot of balance in terms of any solution that's put forward.

No. I mean, the only thing that's probably changed for us in China, and I would say that's probably for a lot of the world, is that China used to be a very rich market for startup companies. And venture capital flew around very freely. There was a lot of innovation and things of that nature. That has absolutely slowed down, whether that is the exit

for these companies isn't as clear, whether from a stock market standpoint, whether it's getting access to key technology isn't as well understood. We've seen that definitely slow down. On the flip side, we've seen incredible growth in segments such as automotive. If you look at companies like BYD or even Xiaomi,

building EVs, the technology in those vehicles is just unbelievable in terms of its capability. And selfishly for us, they all run on ARM. And the reason for that is, and China is very pragmatic in terms of how they build their systems and products. They rely very heavily on the open source global ecosystem for software. And

All of the software libraries have been tuned for ARM, whether it's ADAS or in the powertrain or IVI, that's all ARM based. So we have our automotive business in China is really strong.

Not really. I think my personal view on this is that the threats of tariffs are a bit of a tool to get to the negotiating table. I think President-elect Trump has proven over time that he is a businessman, and tariffs, I think, are one lever to start a negotiation. We'll see where it goes, but in my belief, I'm not too worried about that.

I don't think we need a government OpenAI Manhattan-type project. I think the work that's being done by OpenAI or Anthropic or even the work in the open source that's being driven by Meta with Llama, we're seeing fantastic innovation on that. I think if you look and say... Is the U.S. a leader in terms of foundation models and frontier models? Absolutely, yes.

And that's being done without a government intervention. So on the AI theme, I don't think it's necessary, personally. On the area of fabs, going back to the question that you started me on with on Intel, spending $30 to $40 billion a year in CapEx for these leading-edge nodes, that is a hard... pill to swallow for any company.

And that's why Chips Act, I think, was a good and necessary thing because building semiconductors is fundamental to our economic engine. We learned that during COVID when it took 52 weeks to get a key fob replaced in terms of everything going on with supply chain. So I think having supply chain resiliency is super important. I think it's super important on a global level.

It's definitely important on a national level. So I was and am in favor of the CHIPS Act.

I think one of the things that's happening is a real raise of visibility on this talent issue. And I think putting more and more money into university programs around semiconductors, semiconductor research is helping. I think that for a number of years, semiconductors and particularly manufacturing was not seen as the most attractive of degrees to go off and get.

A lot of people were looking at software as a service and things in other areas. So I think we need to get back to that on the university level. Now, one could argue if AI bots and agents can come in and do meaningful work, maybe that helps. But

I think in the area of building chips and semiconductor process, that is very much an art as well as a science, and particularly around improving manufacturing yields. So I don't know if we have enough talent, but I know there's a lot of effort now going to try to bolster that.

Yeah, I think one of the things we were talking about earlier was we are now a public company. We were not a public company in 2022. So one of the things I've learned as a public company is break out as little as you can possibly can. So nobody can ask you questions in terms of where things are going.

Yeah, I know you are. So I would say this, no, we don't break any of that stuff out. But what we are observing is that, and I think this is only going to accelerate, is whether you're talking about an AI data center or you're talking about an AirPod or a wearable in your ear, there's an AI workload that's now running and that's very clear.