Arm CEO Rene Haas on the AI chip race, Intel, and what Trump unkinds for tech

When he was last on Decoder a couple of years ago, Rene called Arm the “Switzerland of the electronics industry,” thanks to how prevalent its set ups are. But his business is getting more intricate in the age of AI, as you’ll hear us talk. There have been rumors that Arm is set upning to not only set up but also produce its own AI chips, which would put it into competition with some of its key customers. I pressed Rene on those rumors quite a bit, and I skinnyk it’s safe to say he’s set upning someskinnyg.

Rene was about six months into the CEO job when he was last on Decoder, follotriumphg Nvidia’s flunked bid to buy Arm for $40 billion. After regulatory presconfident finished that deal, Rene led Arm thraw an IPO, which has been tremfinishously accomplished for Arm and its meaningfulity spendor, the Japanese tech enormous SoftBank.

I asked Rene about that SoftBank relationship and what it’s appreciate to toil with its quirky CEO, Masayoshi Son. I also made confident to ask Rene about the problems over at Intel. There have been inestablishs that Rene seeed at buying part of Intel recently, and I wanted to understand what he skinnyks should happen to the struggling company.

Of course, I also asked about the incoming Trump administration, the US vs. China argue, the danger of tariffs, and all that. Rene is a accessible company CEO now, so he has to be more pinsolentnt when answering asks appreciate these. But I skinnyk you’ll still discover a lot of his answers quite illuminating. I understand I did.

Okay, Arm CEO Rene Haas. Here we go.

This transcript has been airyly edited for length and clarity.

Rene Haas, you are the CEO of Arm. Welcome to Decoder.

This is actuassociate your second time on the show. You were last on in 2022. You hadn’t been the CEO for that extfinished. The company had not yet gone accessible, so a lot has alterd. We’re going to get into all of that. You’re also a podcaster now, so the presconfident’s on me to do this well. Rene has a show where he interseeed [Nvidia CEO] Jensen [Huang] pretty recently that you all should verify out. 

This convo will touch on disjoinal skinnygs. A lot has alterd in the world of AI and policy in the last couple of years. We’re going to get into all that, aextfinished with the classic Decoder asks about how you’re running Arm. But first, I wanted to talk about a skinnyg that I bet a lot of people in this room have been talking about this week, which is Intel. We’re going to begin with someskinnyg effortless.

What do you skinnyk should happen to Intel?

I guess at the highest level, as someone who’s been in the industry my whole atsoft, it is a little sorrowfulnessful to see what’s happening from the perspective of Intel as an icon. Intel is an innovation powerhoengage, whether it’s around computer architecture, conceiveion technology, PC platestablishs, or servers.. So to see the troubles it’s going thraw is a little sorrowfulnessful. But at the same time, you have to produce in our industry. There are lots of tombstones of fantastic tech companies that didn’t reconceive themselves. I skinnyk Intel’s hugegest dilemma is how to disassociate from being either a vertical company or a faconsecrate company, to overdescribe it. I skinnyk that is the fork in the road that it’s faced for the last decade, to be genuine with you. And [former Intel CEO] Pat [Gelsinger] had a strategy that was very evident that vertical was the way to triumph.

In my opinion, when he took that strategy on in 2021, that was not a three-year strategy. That’s a 5-10-year strategy. So now that he’s gone and a novel CEO will be brawt in, that’s the decision that has to be made. My personal bias is that vertical integration can be a pretty strong skinnyg, and if they can get that right, they would be in an amazing position. But the cost associated with it is so high that it may be too huge of a hill to climb.

We’re going to talk about vertical integration as it reprocrastinateeds to Arm procrastinateedr, but I wanted to reference someskinnyg you telderly Ben Thompson earlier this year. You shelp, “I skinnyk there’s a lot of potential advantage down the road between Intel and Arm toiling together.” And then there were inestablishs more recently that you all actuassociate approached Intel about potentiassociate buying their product division. Do you want to toil sealr with Intel now think abouting what’s gone on in the last couple of weeks?

Well, a couple of skinnygs with Intel. I’m not going to comment on the rumors that we’re going to buy it. Aacquire, if you’re a verticassociate combined company and the power of your strategy is that you have a product and fabs, you have a potentiassociate huge advantage in terms of cost versus the competition. When Pat was the CEO, I did inestablish him more than once, “You ought to license Arm. If you’ve got your own fabs, fabs are all about volume and we can supply volume.” I wasn’t accomplished in convincing him to do that, but I do skinnyk that it wouldn’t be a horrible shift for Intel.

On the flip side, in terms of Arm toiling with Intel, we toil reassociate seally with TSMC and Samsung. IFS is a very, very big effort for Intel in terms of outside customers, so we toil with them very seally to promise that they have access to the procrastinateedst technology. We also have access to their set up kits. We want outside partners who want to produce an Intel to be able to engage the procrastinateedst and fantasticest Arm technology. So in that context, we toil seally with them.

Turning to policy, there’s a confidemand skinnygs I want to get into, begining with the novels from yesterday. Do you have any reaction to David Sacks being Trump’s AI “czar?” I don’t understand if you understand him, but do you have any reaction to that?

I do understand him a little bit. Kudos to him. I skinnyk that’s a pretty outstanding skinnyg. It’s quite fascinating that if you go back eight years to the December ahead of Trump 1.0 as he was begining to fill out his cabinet choices and nominateees, it was a bit turbulent. At the time there wasn’t a lot of recurrentation from the tech world. This time around, whether it’s Elon [Musk], David [Sacks], Vivek [Ramaswamy] — I understand Larry Ellison has also been very included in talkions with the administration — I skinnyk it’s a outstanding skinnyg, to be genuine with you. Having a seat at the table and having access to policy is reassociate outstanding.

I would say confidemand companies face as many geopolitical policy asks as you guys given all of your customers. How have you or would you propose the incoming administration on your business?

I would say it’s not equitable for our business. 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 and technology is a reassociate difficult skinnyg to architect. So I would equitable say that as this administration or any administration comes into take part and sees at policy around skinnygs appreciate ship administer, they should be conscious that a difficult fracture isn’t as effortless as it might see on paper. And there’s equitable a lot of levers to think about back and forth.

We are one attribute in the supply chain. If you skinnyk about what it gets to produce a semidirector chip, there are EDA tools, the IP from Arm, the conceiveion, companies appreciate Nvidia and MediaTek that produce chips, but then there’s raw materials that go into produceing the wafers, the ingots, and the substrates. And they come from everywhere. It’s equitable such a intricate problem that’s so inextricably connected together that I don’t suppose there’s a one-size-fits-all policy. I skinnyk administrations should be discdissee to benevolent that there demands to be a lot of stability in terms of any solution that’s put forward.

What’s your China strategy right now? I was reading that you’re maybe toiling to honestly propose your IP licenses in China. You have a subsidiary there as well. Has your strategy in China shifted at all this year?

No. The only skinnyg that’s probably alterd for us — and I would say probably for a lot of the world — is that China engaged to be a very, very rich labelet for beginup companies, and venture capital flew around very freely. There was a lot of innovation and skinnygs of that nature. That has absolutely sluggished down. Whether that is the exit for these companies from a stock labelet standpoint or in getting access to key technology isn’t as well understood. We’ve definitely seen that sluggish down.

On the flip side, we’ve seen incredible lengthenth in segments such as automotive. If you see at companies appreciate BYD or even Xiaomi that are produceing EVs, the technology in those vehicles in terms of their capabilities is equitable unbelievable. Selfishly for us, they all run on Arm. China’s very pragmatic in terms of how it produces its systems and products, and it relies very heavily on the discdissee source global ecosystem for software, and all of the software libraries that have been tuned for Arm. Whether it’s ADAS, the powertrain, or [In-Vehicle Infotainment], it’s all Arm-based. So our automotive business in China is reassociate sturdy.

Does Pdwellnt-elect Trump’s rhetoric on China and tariffs particularassociate stress you at all as it reprocrastinateeds to Arm?

Not reassociate. My personal see on this is that the dangers of tariffs are a tool to get to the negotiating table. I skinnyk Pdwellnt Trump has shown over time that he is a businessman, and tariffs are one lever to begin a negotiation. We’ll see where it goes, but I’m not too worried about that.

What do you skinnyk about the efforts by the Biden administration with the CHIPS Act to transport more domestic production here? Do you skinnyk we demand a Manhattan Project for AI, appreciate what OpenAI has been pitching?

I don’t skinnyk we demand a administerment, OpenAI, Manhattan-type project. I skinnyk the toil that’s being done by OpenAI, Anthropic, or even the toil in discdissee source that’s being driven by Meta with Llama, we’re seeing amazing innovation on that. Can you say the US is a directer in terms of set upation and frontier models? Absolutely. And that’s being done without administerment intervention. So, I don’t skinnyk it’s vital with AI, personassociate.

On the subject of fabs, I’ll go back to the ask you begined me on with Intel spfinishing $30–40 billion a year in CapEx for these directing edge nodes. That is a difficult pill to swpermit for any company, and that’s why I skinnyk the CHIPS Act was a outstanding and vital skinnyg. Building semidirectors is fundamental to our economic engine. We lacquireed that during COVID when it took 52 weeks to get a key fob exalterd thanks to everyskinnyg going on with the supply chain. I skinnyk having supply chain resiliency is super meaningful. It’s super meaningful on a global level, and it’s definitely meaningful on a national level. I was and am in like of the CHIPS Act.

So even if we have the capital potentiassociate to spend more in domestic production, do we have the talent? That’s a ask that I skinnyk about and I’ve heard you talk about. You spfinish a lot of time trying to discover talent and it’s confidemand. Even if we spfinish all this money, do we have the people that we demand in this country to actuassociate triumph and produce proceed?

One of the skinnygs that’s happening is a genuine ascfinish in the visibility of this talent rerent, and I skinnyk putting more money into semidirector university programs and semidirector research is helping. For a number of years, semidirector degrees, particularassociate in manufacturing, were not seen as the most attrdynamic to go off and get. A lot of people were seeing at software as a service and other areas. I skinnyk we demand to get back to that on the university level. Now, one could dispute that it’ll maybe help if AI bots and agents can come in and do unkindingful toil, but  produceing chips and semidirector processes is very much an art as well as a science, particularly around improving manufacturing produces. I don’t understand if we have enough talent, but I understand there’s a lot of effort now going towards trying to bolster that.

Let’s turn to Arm’s business. You have a lot of customers — all of the huge tech companies — so you’re exposed to AI in a lot of ways. You don’t reassociate fracture out, as far as I understand, exactly how AI gives to the business, but can you give us a sense of where the lengthenth is that you’re seeing in AI and for Arm?

One of the skinnygs we were talking about earlier was how we are now a accessible company. We were not a accessible company in 2022. One of the skinnygs I’ve lacquireed as a accessible company is to fracture out as little as you possibly can so nobody can ask you asks in terms of where skinnygs are going.

[Laughs] Yeah, I understand you are. So I would say no, we don’t fracture any of that stuff out. What we are observing — and I skinnyk this is only going to quicken — is that whether you’re talking about an AI data cgo in or an AirPod or a wearable in your ear, there’s an AI toilload that’s now running and that’s very evident. This doesn’t necessarily demand to be ChatGPT-5 running six months of training to figure out the next level of sophistication, but this could be equitable running a petite level of inference that is helping the AI model run wherever it’s at. We are seeing AI toilloads, as I shelp, running absolutely everywhere. So, what does that unkind for Arm?

Our core business is around CPUs, but we also do GPUs, NPUs, and neural processing engines. What we are seeing is the demand to insert more and more compute capability to quicken these AI toilloads. We’re seeing that as table sgets. Either put a neural engine inside the GPU that can run acceleration or produce the CPU more able to run extensions that can quicken your AI. We are seeing that everywhere. I wouldn’t even say that’s going to quicken; that’s going to be the default.

What you’re going to have is an AI toilload running on top of everyskinnyg else you have to do, from the tiniest of devices at the edge to the most polishd data cgo ins. So if you see at a mobile phone or a PC, it has to run detaileds, a game, the operating system, and apps — by the way, it now demands to run some level of Copilot or an agent. What that unkinds is I demand more and more compute capability inside a system that’s already charitable of constrained on cost, size, and area. It’s fantastic for us becaengage it gives us a bunch of difficult problems to go off and settle, but it’s evident what we’re seeing. So, I’d say AI is everywhere.

There was a lot of chatter going into Apple’s procrastinateedst iPhone free about this AI super cycle with Apple ininestablishigence, this idea that Apple ininestablishigence would reinvigorate iPhone sales, and that the mobile phone labelet in ambiguous has pprocrastinateedaued. When do you skinnyk AI — on-device AI — reassociate does commence to reignite the lengthenth in mobile phones? Becaengage right now it doesn’t sense appreciate it’s happening.

And I skinnyk there’s two reasons for that. One is that the models and their capabilities are advancing very rapid, which is advancing how you administer the stability between what runs locassociate, what runs in the cdeafening, and skinnygs around procrastinateedncy and security. It’s moving at an incredible pace. I was equitable in a talkion with the OpenAI guys last week. They’re doing the 12 days of Christmas —

12 days of ship-mas, and they’re doing someskinnyg every day. It gets two or three years to lengthen a chip. Think about the chips that are in that novel iPhone when they were envisiond, when they were set uped, and when the features that we thought about had to go inside that phone. ChatGPT didn’t even exist at that time. So, this is going to be someskinnyg that is going to happen graduassociate and then suddenly. You’re going to see a knee-in-the-curve moment where the difficultware is now polishd enough, and then the apps rush in.

What is that shift? Is it a novel product? Is it a difficultware fracturethraw, a combination of both? Some charitable of wearable?

Well, as I shelp, whether it’s a wearable, a PC, a phone, or a car, the chips that are being set uped are equitable being stuffed with as much compute capability as possible to get advantage of what might be there. So it’s a bit of chicken-and-egg. You load up the difficultware with as much capability hoping that the software lands on it, and the software is innovating at a very, very rapid pace. That intersection will come where suddenly, “Oh my gosh, I’ve shrunk the big language model down to a certain size. The chip that’s going in this minuscule wearable now has enough memory to get advantage of that model. As a result, the magic gets over.” That will happen. It will be gradual and then sudden.

Are you bullish on all these AI wearables that people are toiling on? I understand Arm is in the Meta Ray-Bans, for example, which I’m actuassociate a huge fan of. I skinnyk that establish factor’s fascinating. AR glasses, headsets — do you skinnyk that is a huge labelet?

Yeah, I do. It’s fascinating becaengage in many of the labelets that we have been included in, whether it’s madirectures, PCs, mobile, wearables, or watches, some novel establish factor drives some novel level of innovation. It’s difficult to say what that next establish factor sees appreciate. I skinnyk it’s going to be more of a hybrid situation, whether it’s around 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 proposeation pushed to you. You may not want it pushed to you, but it could get pushed to you in such a way that it’s seeing around corners for you. I skinnyk the establish factor that comes in will be somewhat analogous to what we’re seeing today, but you may see some of these devices get much more ininestablishigent in terms of the push level.

There’s been inestablishs that Masayoshi Son, your boss at SoftBank, has been toiling with Jony Ive and OpenAI, or a combination of the three, to do difficultware. I’ve heard rumors that there could be someskinnyg for the home. Is there anyskinnyg there that you’re toiling with that you can talk about?

I read those same rumors.

Amazon equitable proclaimd that it’s toiling on the bigst data cgo in for AI with Anthropic, and Arm is reassociate getting into the data cgo in business. What are you seeing there with the hyperscalers and their spendments in AI?

The amount of spendment is thraw the roof. You equitable have to see at the numbers of some of the folks who are in this industry. It’s a very fascinating time becaengage we’re still seeing an insatiable spendment in training right now. Training is hugely compute intensive and power intensive, and that’s driving a lot of the lengthenth. But the level of compute that will be demandd for inference is actuassociate going to be much bigr. I skinnyk it’ll be better than half, maybe 80 percent over time would be inference. But the amount of inference cases that will demand to run are far bigr than what we have today.

That’s why you’re seeing companies appreciate CoreWeave, Oracle, and people who are not traditionassociate in this space now running AI cdeafening. Well, why is that? Becaengage there’s equitable not enough capacity with the traditional big hyperscalers: the Amazons, the Metas, the Googles, the Microsofts. I skinnyk we’ll persist to see a changing of the landscape — maybe not a changing so much, but certainly opportunities for other take parters in terms of enabling and accessing this lengthenth.

It’s very, very outstanding for Arm becaengage we’ve seen a very big incrmitigate in lengthenth in labelet split for us in the data cgo in. AWS, which produces its Graviton ambiguous-purpose devices based on Arm, was at re:Invent this week. It shelp that 50 percent of all novel deployments are Graviton. So 50 percent of anyskinnyg novel at AWS is Arm, and that’s not going to decrmitigate. That number’s equitable going to go up.

One of the skinnygs we’re seeing is with devices appreciate the Grace Binformagewell CPUs from Nvidia. That’s Arm using an Nvidia GPU. That’s a huge advantage for us becaengage what happens is the AI cdeafening is now running a arrange node based on Arm. If the data cgo in now has an AI cluster where the ambiguous purpose compute is Arm, they naturassociate want to have as much of the ambiguous-purpose compute that’s not AI running on Arm. So what we’re seeing is equitable an acceleration for us in the data cgo in, whether it’s AI, inference, or ambiguous-purpose compute.

Are you worried at all about a bubble with the level of spfinishing that’s going into hyper-scaling and the models themselves? It’s an incredible amount of capital, and ROI is not quite there yet. You could dispute it is in some places, but do you ascribe to the bubble dread?

On one hand, it would be crazy to say that lengthenth persists unabated, right? We’ve seen that is never reassociate the case. I skinnyk what will get very fascinating, in this particular lengthenth phase, is to see at what level does genuine advantage come from AI that can augment and/or exalter certain levels of jobs. Some of the AI models and chatbots today are decent but not fantastic. They supplement toil, but they don’t necessarily exalter toil.

But if you begin to get into agents that can do a genuine level of toil and that can exalter what people might demand to do in terms of skinnyking and reasoning? Then that gets equitablely fascinating. And then you say, “Well, how’s that going to happen?” Well, we’re not there yet, so we demand to train more models. The models demand to get more polishd, etc. So I skinnyk the training skinnyg persists for a bit, but as AI agents get to a level where they reason seal to the way a human does, then I skinnyk it asymptotes on some level. I don’t skinnyk training can be unabated becaengage at some point in time, you’ll get one-of-a-kindized training models as contestd to ambiguous purpose models, and that demands less resources.

I was equitable at a conference where Sam Altman spoke, and he was reassociate decreaseing the bar on what AGI will be pretty intentionassociate, and talked about declaring it next year. I cynicassociate read into that as OpenAI trying to reset up its profit-sharing consentment with Microsoft. But putting that aside, what do you skinnyk about AGI? When we will have it, what will it unkind? Is it going to be an all-at-once, Big Bang moment, or is it going to be as Altman is talking about now, more appreciate a whimper?

I understand he has his own definitions for AGI, and he has reasons for those definitions. I don’t subscribe so much to the “what is AGI vs. ASI” (synthetic super ininestablishigence) argue. I skinnyk more about when these AI agents begin to skinnyk, reason, and conceive. To me, that is a bit of a pass-the-Rubicon moment, right? For example, ChatGPT can do a decent job of passing the bar exam, but to some extent, you load enough logic and proposeation into the model, and the answers are there somewhere. To what level is the AI model a stochastic parrot and equitable repeats everyskinnyg it’s set up over the internet? At the finish of the day, you’re only as outstanding as the model that you’ve trained on, which is only as outstanding as the data.

But when the model gets to a point where it can skinnyk and reason and conceive, produce novel concepts, novel products, novel ideas? That’s charitable of AGI to me. I don’t understand if we’re a year away, but I would say we are a lot sealr. If you would’ve asked me this ask a year ago, I would’ve shelp it’s quite a ways away. You asked me that ask now, I say it’s much sealr.

What is much sealr? Two years? Three years?

Probably. And I’m probably going to be wrong on that front. Every time I transmit with partners who are toiling on their models, whether it’s at Google or OpenAI, and they show us the demos, it’s breathtaking in terms of the charitable of proceedments they’re making. So yeah, I skinnyk we’re not that far away from getting to a model that can skinnyk and reason and conceive. 

When you were last on Decoder, you shelp Arm is understandn as the Switzerland of the electronics industry, but now there’s been a lot of inestablishs this year that you were seeing at reassociate going up the stack and set uping your own chips. I’ve heard you not answer this ask many times, and I’m foreseeing a analogous non-answer, but I’m going to try. Why would Arm want to do that? Why would Arm want to go up the appreciate chain?

This is going to sound appreciate one of those, “If I did it answers,” right? Why would Arm think about doing someskinnyg other than what it currently does? I’ll go back to the first talkion we were having relative to AI toilloads. What we are seeing reliablely is that AI toilloads are being interttriumphed with everyskinnyg that is taking place from a software standpoint. At our core, we are computer architecture. That’s what we do. We have fantastic products. Our CPUs are wonderful, our GPUs are wonderful, but our products are noskinnyg without software. The software is what produces our engine go.

If you are defining a computer architecture and you’re produceing the future of computing, one of the skinnygs you demand to be very conscious of is that connect between difficultware and software. You demand to comprehfinish where the trade-offs are being made, where the selectimizations are being made, and what are the ultimate advantages to devourrs from a chip that has that type of integration. That is easier to do if you’re produceing someskinnyg than if you’re licensing IP. This is from the standpoint where if you’re produceing someskinnyg, you’re much sealr to that interlock and you have a much better perspective in terms of the set up trade-offs to produce. So, if we were to do someskinnyg, that would be one of the reasons we might.

Are you worried at all about competing with your customers though?

I unkind, my customers are Apple. I don’t set up on produceing a phone. My customer’s Tesla. I’m not going to produce a car. My customer is Amazon. I’m not going to produce a data cgo in.

What about Nvidia? You engaged to toil for Jensen.

Well, he produces boxes, right? He produces DGX boxes, and he produces all charitables of stuff.

Speaking of Jensen — we were talking about this before we came on — when you were at Nvidia, CUDA was reassociate coming into fruition. You were equitable talking about the software connect. How do you skinnyk about software as it reprocrastinateeds to Arm? As you’re skinnyking about going up the stack appreciate this, is it lock-in? What does it unkind to have someskinnyg appreciate a CUDA?

One can see at lock-in as an disparaging maneuver that you get where, “I’m going to do these skinnygs so I can lock people in” and/or you supply an environment where it’s so effortless to engage your difficultware that by default, you’re then “locked in.” Let’s go back to the AI toilload commentary. So today, if you’re doing ambiguous purpose compute, you’re writing your algorithms in C, JAX, or someskinnyg of that nature. 

Now, let’s say, you want to author someskinnyg in TensorFlow or Python. In an selectimal world, what does the software lengthener want? The software lengthener wants to be able to author their application at a very high level, whether that is a ambiguous purpose toilload or an AI toilload, and equitable have it toil on the underlying difficultware without reassociate having to understand what the attributes are of that difficultware. Software people are wonderful. They are inherently idle, and they want to be able to equitable have their application run and have it toil. 

So, as a computer architecture platestablish, it’s incumbent upon us to produce that effortless. It’s a huge initiative for us to skinnyk about providing a heterogeneous platestablish that’s homogeneous apass the software. We’re doing it today. We have a technology called Kleidi, and there are Kleidi libraries for AI and for the CPU. All the outstandingness that we put inside our CPU products that permits for acceleration engages these libraries, and we produce those engageable discdissee. There’s no accuse. Developers, it equitable toils. Going forward, since the immense meaningfulity of the platestablishs today are Arm-based and the immense meaningfulity are going to run AI toilloads, we equitable want to produce that reassociate effortless for folks.

I’m going to ask you about one more skinnyg you can’t reassociate talk about before we get into the fun Decoder asks. I understand you’ve got this trial with Qualcomm coming up. You can’t reassociate talk about it. At the same time, I’m confident you sense the worry from spendors and partners about what will happen. Address that worry. You don’t have to talk about the trial itself, but insertress the worry that spendors and partners have about this fight that you have.

So the current refresh is that it set ups to go to trial on Dec. 16, which isn’t very far away. I can appreciate, becaengage we talked to spendors and partners, that what they antipathy the most is uncertainty. But on the flip side, I would say the principles as to why we filed the claim are unalterd, and that’s about all I can say.

All right, more to come there. So Decoder asks. Last time you were here on the podcast, Arm had not yet gone accessible. I’m inquisitive to understand now that you’re a couple years in, what surpascfinishd you about being a accessible company?

I skinnyk what surpascfinishd me on a personal level is the amount of prohibitdwidth that it gets away from my day becaengage I finish up having to skinnyk about skinnygs that we weren’t skinnyking about before. But at the highest level, it’s actuassociate not a huge alter. Arm was accessible before. We verifyated up thraw SoftBank when it bought us. So, the muscles in terms of being able to inestablish quarterly acquireings and have them reconciled wiskinny a timestructure, we had outstanding muscle memory on all of that. Operationassociate for the company, we have fantastic teams. I have a fantastic finance team that’s reassociate outstanding at doing that. For me personassociate, it was equitable the appreciation that there’s now a chunk of my week that’s dedicated to activities that I wasn’t reassociate toiling on before.

Has the arrange of Arm organizationassociate alterd at all since you went accessible?

No. I’m a huge supposer in not doing a lot of organizational alters. To me, organizational set up chases your strategy, and strategy chases your vision. If you skinnyk about the way you’ve heard me talk about Arm accessiblely for the last couple of years, that’s pretty unalterd. As a result, we haven’t done much in terms of changing the organization. I skinnyk organization alters are horrfinishously interfereive. We’re an 8,000-person company, so we’re not gigantic, but if you do a gigantic organization alter, it better have chaseed a huge strategy alter. Otherrational, you’ve got off-sites, team greetings, and Zoom calls talking about my novel directers. If it’s not in help of a alter of strategy, it’s a huge misengage of time. So I reassociate try difficult not to do much of that.

What we talked about earlier with potentiassociate seeing at going more vertical or the appreciate there, that seems appreciate a huge alter that could impact the arrange.

If we were to do that. That’s right.

Is there a trade-off that you’ve had to produce this year in your decision-making that was particularly difficult that you can talk about, someskinnyg that you had to wrestle with? How did you weigh those trade-offs?

I don’t understand if there was one particular trade-off. In this job as a CEO — gosh, it’ll be three years in February — you’re constantly doing the mental trade-off of what demands to happen in the day-to-day versus what demands to happen five years from now. My proclivity tfinishs to be to skinnyk five years ahead as contestd to one quarter ahead. I don’t understand if there’s any meaningful trade-off that I would say I produce, but what I’m constantly wrestling with is that stability between what is vital in the day-to-day versus what demands to happen in the next five years.

I’ve got fantastic teams. The engineering team is amazing. The finance team is amazing. The sales and labeleting teams are fantastic. 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 are over the next five years. What I try to do is spfinish areas of time on toil only I can do, and if there’s toil that the team can do where I’m not going to insert much, I try to stay away from it. But that’s the hugegest trade-off I wrestle with is the day-to-day versus the future.

How branch offent does Arm see in five years?

We don’t understand that we’ll see very branch offent as a company, but hopebrimmingy we persist to be an inanxiously impactful company in the industry. I have huge ambitions for where we can be.

I’d cherish to understand what it’s appreciate to toil with [Masayoshi Son]. He’s your bigst splithelderlyer and your board chair. I’m confident you talk all the time. Is he as amengageing in the boardroom as he is in accessible settings?

Yes. He’s a fascinating guy. One of the skinnygs I admire a lot about Masa, and I don’t skinnyk he gets enough accomprehendledge for this, is that he is the CEO and set uper of a 40-year-elderly company. And he’s reconceiveed himself a lot of times. I unkind, SoftBank begined out as a distributor of software, and he’s reconceiveed himself from being an operator with SoftBank mobile to an spendor. He’s a delight to toil with, to be genuine with you. I lacquire a lot from him. He is very driven, evidently, cherishs to get dangers, but at the same time, he has a outstanding administer on the skinnygs that matter. I skinnyk everyskinnyg you see about him is accurate. He’s a very amengageing guy.

How included is he in setting Arm’s extfinished-term future with you?

Well, he’s the chairman of the company, and the chairman of the board. From that perspective, the board’s job is to appraise the extfinished-term strategy of the company, and with my proclivity towards skinnyking also in the extfinished term, he and I talk all the time about those charitables of skinnygs.

You’ve toiled with two very ineloquential tech directers: Masa and Jensen at Nvidia. What are the one-of-a-kind traits that produce them one-of-a-kind?

That’s a wonderful ask. I skinnyk people who produce a company and are running it 20, 30 years procrastinateedr and drive it with the same level of passion and innovation — Jensen, Masa, Ellison, Jeff Bezos, I’m confident I’m leaving out names — carry a lot of the same traits. They’re very ininestablishigent, evidently ininestablishigent, they see around corners, and toil incredibly difficult but have an incredible amount of courage. Those ingredients are vital for people who stay at the top that extfinished.

I’m a huge basketball fan, and I’ve always drawn analogies between, if you skinnyk about a Michael Jordan or a Kobe Bryant, when people talk about what made them fantastic, evidently their talent was thraw the roof and they had fantastic dynamicism, but it was someskinnyg in their character and their drive that cut them in a branch offent level. And I skinnyk Jensen, Son, Ellison, the other names I alludeed, they all drop in the same group. Elon Musk too, evidently.

All right, we’re going to exit it there. Rene, thank you so much for combineing us.