Intel Corporation (INTC) Earnings Call Transcript & Summary

Thank you, everybody, for joining us today, the first inaugural Mizuho Tech Conference. Good morning. I'm Vijay Rakesh, senior analyst for -- covering U.S. technology for semiconductor, semi-cap equipment and auto technology. Today, I have the pleasure to kick off the conference here with the breakfast keynote from Saf Yeboah-Amankwah, Senior Vice President and Chief Strategy Officer at Intel Corporation. Saf leads Intel's Corporate Strategy and Venture Group, focused on driving growth-oriented strategies, including strategic partnerships, Intel Capital, equity investments and incubation initiatives. Saf joined Intel in 2020 from McKinsey & Company, where he was the Senior Partner and Global Head of the Transformation Practice for Telecom Media and Technology, TMT. Prior to that, he served as Managing Partner for South Africa and Head of McKinsey's TMT and digital practice for Africa, among other roles. Saf, thank you for joining us. It's a pleasure, so let's get started. I think there is a -- you want -- you have your presentation? I'll do a quick safe harbor here guys. Just take a read through this, but I'll do the abbreviated version here. Before we begin, please note that today's discussion may contain forward-looking statements that are subject to various risks and uncertainties and may reference non-GAAP financial measures. Please refer to Intel's most recent earnings release and annual report on Form 10-K and other filings with the SEC for more information on the risk factors that could cause actual results to differ materially and additional information on non-GAAP financial measures, including reconciliations, where appropriate, with the corresponding GAAP financial measures. Thank you. You can feel free to read that, too. Thanks. Is Saf's microphone working? Is it working now? Okay. Awesome.

Good morning to everybody, and obviously, safe harbor rules. I'm not going to make any news today. I'm going to be spending time giving sort of behind-the-curtain view, a bit of kind of probably thinking about our strategy, where we're going, which is going to drive, a bunch of the announcements and a bunch of the acquisitions and deals that we've done as well as some of the investments and how we're thinking about capital allocation. I'm going to go through that and include some detail as we talk about -- we do our Q&A, and I'm really looking forward to the engagement. I joined Intel about 3.5 years ago from McKinsey. Intel was a client of mine, so I knew the company quite well and came in and worked with -- as I said, I was the forward deployment Pat. So of course, Pat came 8 weeks after I joined. And when Pat came, we spent a lot of time as a management team doing 2 very important things. One is really developing sort of a turnaround strategy for the company. Our view and Pat's view, the Board's view was this is a 10-year journey. It's not a 2-year turnaround story because the challenges we faced did not -- have occurred over quite a long period of time. And it's -- the industry takes a long time to turn around because of the cycles. And so we had a 10 year -- 10- and 20-year view of the world. And we knew that this was going to be challenging in multiple places. The second thing that we did is we took on a pretty comprehensive assessment of our portfolio. I spent, with our CFO at the time and Pat, going through every single component of our portfolio and made a decision -- try to make a decision on whether or not it would fit with our strategy or did not fit with our strategy, and whether or not it was creating shareholder value or whether or not those businesses on assets belonged with somebody else or needed to be exposed to the market. So there's a lot of work that we did. And some of the things that we've since done come out of that exercise. But the most important part of the exercise was really thinking about what we call the IDM 2.0 strategy. And the IDM 2.0 strategy essentially thinks about the company as 2 companies in 1, 2 excellent companies in 1. There's the first half of the company, which is really our products, operations, and we needed that to be excellent. And we felt that it was very clear that our product leadership had waned and that we needed a -- and in some ways, had been -- sometimes has been held by our process leadership, our technology leadership in the manufacturing side and vice versa. And it's also very clear to us that our manufacturing leadership had waned. What everybody talked about from our manufacturing leadership was obviously our process technology, 10-nanometer, 7-nanometer. But what I think was underappreciated was that our capacity leadership had also waned. For almost 8, 9 years, we have not added any capacity into the -- into our footprint. And this is a fixed cost business and obviously having scale and capacity is a strategic asset. And so hence, leading to our IDM 2.0 strategy. The other critical part that we spent time on was trying to get our point of view on what it meant to be a great asset to be invested in, and hence, the real focus around financial discipline, financial leadership. And I think we spend a lot of time as a management team and communicating all the work that we're doing to get to that. And then obviously, the foundation of all of this is our people. One of the things that we found 3-odd years ago was that the -- Intel had losses [indiscernible] a little bit. Internally, the north star of innovation driving leadership, really listening to our customers, setting the pace from the open source ecosystem, driving both software and hardware co-optimization, driving process and product co-optimization had really waned. And we needed a cultural transformation to sustain that. So those are the -- essentially the core principles or core drivers of what led to the IDM 2.0 strategy. Not for the faint of heart in terms of transformation because both sides required a huge amount of rethink and re-vectoring, which we've been doing. Takes time, right? It takes 2-odd years to develop new products, and we kind of redid the whole product road map. And it takes -- it's going to take 4 years, 5 nodes in 4 years to get back to process leadership. The good news is that we feel -- there's been bumps as in all markets. The market has -- there's -- the market around us changed. Obviously, there's much more competitive intensity in our business. There is geopolitical dynamics that we have to navigate. But at the end of the day, the core of the core of our strategy, which is we have -- having product leadership, having manufacturing leadership both in terms of technology and capacity and continuing to generate the magic of when you have co-optimization of our semiconductors with our manufacturing creating 1 plus 1 equals 3 remains the core of the strategy. And importantly, to -- sorry, to really accelerate this and make this both internally clear, but also externally clear, we've gone down the path of really creating 3 entities inside Intel. The first, obviously, is the Intel Products team. And what we've done is what we've asked ourselves, what we focused on is saying, what would it take to have the most excellent product company on Earth at the scale of $55-odd billion of revenue. What does it mean in terms of technology, in terms of cost structure, in terms of road map? What does it mean in terms of where we focus, both in terms of CPUs, GPUs, accelerators and all of the other technologies that we have? And how do we drive the go-to-market engine that is fit for purpose? And on the other side, we've also created an Intel Foundry that serves both Intel Products, but as part of the critical part of the Intel strategy in terms of Intel Foundry strategy also serve third parties. We know that it took TSMC 20, 30 years to create TSMC. It's not lost on us. We know that we need to play differently than just simply going down the path that they've gone to. And hence, we are playing an attacker role in the foundry business. We're being very focused on leading node capacity. We're being very focused on providing a value proposition around resilience, i.e. having geographic and company resilience when it comes to manufacturing. And we also know that we have some critical leadership capabilities, that we know the market is going to need and want, including some technical leadership in manufacturing, like backside power, but also some really compelling capabilities around packaging, around photonics that we know the market will need us as accelerated computing becomes more and more important. And that's the focus that we have on Intel Foundry. But we also know that the Intel Foundry business, when it was engaged, intertwined with the Intel Products business, both sides have become, I wouldn't say lazy, but there was a lot of inefficiencies, catching each other that creates a lot of cost, but frankly, missed opportunities. And so creating this separation both gives the Intel Foundry third-party customers' confidence, but more importantly, creates a much better, cleaner, thoughtful, efficient ways for the product organization to work with the foundry organization. And then the third is that we have Intel Corporate. Intel Corporate basically is the -- basically the G&A functions that run the company, manage the company. But in Intel Corporate also are a number of critical capabilities that we have that both Intel Foundry and Intel Products leverage. Intel Federal, Intel Labs, which is driving a lot of innovation in the industry and then obviously, Intel Capital, which I run, which does a lot of investments in the ecosystem to both enhance the ecosystem that we're working, but also to drive eyes and ears for the company. And these are really both for both sides. And so we're running these 3 things very separately, and we go into a process to make sure we're doing this as efficient as possible. And hence, all of the discussions and the announcements we've been making and trying to create much more clarity, both internally, frankly, and has led to a lot of cost reduction opportunities, but also externally, both to customers, both on the foundry side and the products side, but also to our employees. So that is the core of kind of the journey we're on, as I said, and we'll talk about some of that now. We're starting to feel -- we knew this is going to be tough, right? We didn't get the -- this is not an easy journey, but we are confident about the assets, capabilities, technologies and brand that we have. And we feel very confident that the secular trend around compute growth, right, getting to $1 trillion, if you will, in semiconductors, foundry growth and the need for leading node capacity for most of the fabless companies and also the need for geopolitical -- geo-diverse capacity for resilience is a trend that creates -- has huge amount of headroom for shareholder value creation that if we are -- we execute, we can be -- we can reward our investors with. So that's essentially the journey that we're on. With that, I think I've come to the end.

Great. Thanks. Fantastic, Saf. That's very helpful. I just want to kick off with some quick questions here. As Head of Strategy, you also head up Intel Capital, as you mentioned, and you're also in charge of equity investments and incubation opportunities. Where do you see opportunities in Intel, especially on the AI, data center, compute side, where Intel can invest or do partnerships? So where do you see opportunities to further that road map and maybe catch up, as you mentioned?

Great. And I see this from many different sites as an investor in start-ups, but also shepherding our corporate strategy and working with our business units. I think I'll say 2 things, and then I'll talk about Intel specifics. First is that Intel has been investing in AI for 15-odd years, right? This has been a huge focus of the company in different parts of AI. And on the data center side, our view, especially around gen AI, is that we are in the first innings of a 9-inning game, and I can describe that in multiple different dimensions. If you think about the models that are being used, we're starting to get to places where we need fundamentally different ways of modeling and using these foundational models because the scalability and the power consumption is becoming unwieldy. If you think about making security and accuracy, there's huge amount of headroom left. The way that I think about it is that these models give you -- it's like a high school kid that gets a B+. It's fantastic for high school, but for enterprises and for doing health research and so on and so forth, you can't afford B+. You have to have an A+, or at least consistently an A. And so there's a lot of headroom in terms of performance. And then even -- and then in the infrastructure side, we think there's a tremendous amount of innovation and work to be done, not just on the accelerator or the CPU side, but also on the networking, which actually today is the biggest bottleneck on memory and then also on all the other technologies, like cooling. So on the data center side, we are obviously investing, and we feel pretty excited about the progress we're making on the accelerator side. But I think what maybe we don't talk a lot about is that we are also investing, and we're very excited about all the innovations and work we're doing on the other bottlenecks, including memory, including networking and a whole bunch of technologies that we have inside Intel Labs and otherwise around how do you actually run a fleet of CPUs and GPUs. And how do you think about cooling? And how do you think about providing fleet management? And so these -- in this space, we're super excited about that. And we feel that -- and by the way, photonics is also another example of innovation. And we feel that given where the beginning at the ground level of this massive explosion, that the -- there are quite a number of opportunities for intersection as we think about our -- on the client side, AIPC, we are -- we continue to believe that there's 2 problems we're trying to solve. The first problem we're trying to solve is refresh. The refresh rates in the PC industry has gone from 2 to 3 years to 5 to 7 years. And obviously from a time perspective, that's a pretty important issue for us to solve. And so we believe that the AIPC is going to drive another wave of refreshes in the industry, and we're going to drive that using the benefit of AIPC and the need for everybody to have an AIPC to drive that. And I think we think that, that's a pretty important trend over the next 5, 7 years as the kind of the refresh cycle happens again. The second is that we believe that, while there is a lot of excitement and a lot of discussion right now, but the current versions of the AIPC, we think this is just the beginning of a journey. The beautiful thing about Gen AI is that it's kind of working everybody up to the possibilities for both large models, but also small models that are running on your computer. But it is just text, right? What we all know is that innovation at the edge or the consumption level has to be more than just text. It has to be video, has to be audio, it has to be -- it has to work with human beings. And there lies, in our minds, a pretty big headroom in terms of innovation and growth over the next 4, 5 years, if the industry, with us obviously being the largest PC player, driving that. And so lots of strategic discussions and thoughts going on, on what that could look like, how that works with the ISV ecosystem, and then also what that means from a platform perspective and what we produce.

Got it. I think that's very helpful, especially on the AI data center side and the AIPC side how that's going to evolve. But talking about competition, obviously, you guys dominate the x86 side. And what you're seeing is ARM is trying to come in both on the data center side and the PC side as well. But then you have a partnership with ARM as well on the foundry side. So how do you balance that? On the one hand, they're probably taking -- trying to eat your lunch. On the other side, you're partnering with them on the foundry side, maybe helping them further in their road map, I guess. How do you balance that? What do you see as -- down the road, what do you see as the movement there?

So we're a very paranoid organization, right? I think one of our founders famously talked about [indiscernible] that's how you survive, number one. Number two, you said dominate, I didn't. We don't use that word at Intel. Number three, what I'd say is we deeply believe that x86 and the innovations around x86 has been a huge benefit to the world in driving compute capacity and cost down using obviously Moore's Law and other innovations. And we believe there's still a lot of headroom in that innovation. Having said that, we think there's a place for ARM-based architectures. We think there is a place for ARM-based architectures in different parts of the compute ecosystem. And we're very excited to, from a product perspective, compete with it. We believe that, especially on the CPU side, we know how to make CEVA-class CPUs, and we know so many things about that space that once our road map that we have comes to bear, that we are confident competing for there's all kinds of technologies around fleet management, around memory management, around power management, around that we've learned painfully over the last 10 years, 15, 20 years, that we think gives us an advantage as long as, obviously, we meet the threshold around the product being compelling. And so we're excited to have that battle, and we think it's good for the industry. We also think that part of the reason that the ARM-based server chips have become popular is frankly because of us, right? We did not -- we were not as responsive to the market as we probably should have been around the CSPs needing different types of offerings because we felt pretty confident that what we were going to offer is going to solve all their problems. And so I think we are excited about the opportunity to compete. And if we don't win all the sockets, we're totally excited about them coming to Intel Foundry to fab it. On the client side, I think this is actually a multiyear decade story around ARM coming into the PC environment. Again, we think the question is not ARM versus -- it's not an [ RTL ] issue. It's a product leadership issue. And we feel pretty confident that our product road map on the client side, again, as we start seeing Lunar Lake and Panther Lake, that the combination of the product leadership with our go-to-market engine and our marketing engine and our channel engine and frankly, our scale, combined with our ability to really deliver enterprise-grade systems, right, again, that allows for manageability allows for remote. I mean, there's a whole bunch of things that enterprises require before they buy. We believe that, obviously, with the bar being great products, gets us back into competing.

How do you balance that with being a foundry partner for ARM?

We're totally comfortable with that. We think that we will not win every socket. And we really also believe that there's a whole bunch of edge applications, phone applications, a lot of form factors, which x86 is now going to [ interact ]. And we want to make sure that our foundry business is ready to intercept all of those. And in many ways, why the separation is incredibly important because we don't want to constrain the foundry business with the product business.

Got it. That's very helpful. Just to go on the unlocking value path, I guess. One of the things that Pat -- when Pat came in, he said was going to unlock value, and one of the things that he started off with was Mobileye. I think it's -- you guys still have it as a subsidiary now. You've announced Altera as well, I guess. What's the path forward on that? And where else do you see opportunities where you can -- there is a much bigger value of that entity as a stand-alone or the market can appreciate that a lot better the growth there? What are the opportunities ahead? If you can start with Altera, there might be...

Sure. The Mobileye transaction was, I think, came out of the portfolio review that I talked about, where we decided that there were quite a number of assets that would be more valuable or would unlock a lot of shareholder value by having externally. We don't talk a lot about the IMS, for example, which is another asset that we believed -- that we had that we believed was more of an industry asset than just a captive access that we sold an equity stake in and which is another example. And then Altera is another example where we feel very passionately that it makes sense as a stand-alone entity. And there's a lot of -- we announced it, a lot of work going on. And obviously depending -- it all depends on timing, what I've realized. I used to be a consultant. I didn't really realize that timing was so important in the equity market. But it's all depending on timing and pricing. And obviously, the industry is going through a bit of a correction or resetting post pandemic. But we are pretty excited about either doing a stake sale or an IPO, and we're weighing all those options. In terms of beyond that, I'll say 3 or 4 things. One, I think Pat mentioned this, we've sold 5 or 6, maybe 7 assets in the portfolio already. We shut down because we think that's also value creation and a number of businesses inside the company that we've felt shareholders were not going to get a return on. And we continue to be -- to look at multiple different assets and options where we believe we spun out a company called Articul8 into -- which was an internal organization, that is now our business that Intel Capital owns a large portion of, that is doing incredibly well. We're looking -- we are very focused on looking at every opportunity. We think there's 5 or 6 more that we're going to do -- to consider.

That's very helpful. And just -- I know you started off with the foundry strategy. And I think you guys have something called spot capital or I believe the SCIP, Semiconductor Co-investment Programs, I guess. How do you see that evolving? Because I think you announced a pretty remarkable Apollo partnership with Ireland fab, pretty major $11 billion investment from them. But then you pulled back on the Israeli fab in the last couple of days. But just wondering, is that the road map forward that you want to have a co-investment as you put out the next round of fabs? Or how do you see that evolving, I guess, the whole SCIP program? And I'm sure it's part of Intel Capital as well. So...

Yes. So I think let me just broadly say that the Smart Capital, which Dave, our CFO, owns is very focused on leveraging fit for purpose on capital, has 3 components to it. The first is, we don't -- we build shells ahead of demand. We don't build factories ahead of demand. A shell -- factory is $20 billion. Shell is about 15% to 20% of that, and it buys you essentially an option. The rest is the equipment and the power and all the other stuff that you actually put in. And so #1 part of the Smart Capital is that we don't want to be in a place where we can't get capacity, but we also don't want to spend $20 billion buying it. So that's incredibly important. The second part of this Smart Capital strategy is we want to leverage capital to do these build-outs in the most efficient way possible. The first, obviously, is we'd love to get grants and support from governments, especially where we think it makes sense. And that's one critical part of the Smart Capital part. I think we made an announcement around what we think that project -- that will project for the company over the next 5, 7 years as we build out. That's obviously incredibly important. The second is we have been approached by multiple -- I cut my teeth in the telecom industry. The telecom industry has been very successful in partnering with capital markets to fund core infrastructure. And so SCIP basically came out of that, where we felt that quite a lot of the infrastructure that we fund is things that are not even telecom -- I mean, tech-related, right? The construction of the roads, the construction of the buildings, the constructions, the power, and these are all assets that Intel, we should not be using our balance [ sheet ]. And so part of the whole SCIP structure is to essentially have Intel pay the equipment and our partner pay for "the infrastructure" and ask to in a joint venture way drive the business, pay for the investment out of the profits of the factory. That's the simple way of thinking about it. And that's what we did with Brookfield. That's what the structure strategically is trying to do with Apollo. Will we do more of these? I think we will always look for alternative capital. We're very, very jealous of our balance sheet. And so we will look for multiple different ways to get more leverage out of the balance sheet, if we can, obviously in a prudent way, in a way where it's not too expensive and in a way in which we are -- where investors who are interested in taking equity risk with us or taking business risk with us because they believe in the sort of the secular investment cycle that we're in, that we'd consider.

So SCIP, as I see it, is your own private initiative. So that plus CHIPS Act would be a help for you, that'll be completely separate. On the SCIP side, do you see partnering not just with private equity, like Apollo, but also with customers, like hyperscalers or big IDMs...

Short answer is, yes.

Got it, okay. And then that's more like in the pipeline kind of the...

Yes. These are all things that we've had incomings. We also thought about this because some -- there are some clear fabless companies who will want customization and certainty of supply. And if that -- how do you do that if that means that they co-invest with us in the factory, if they give us a prepayment, large prepayment in order to secure [indiscernible] we would consider -- we consider all of those things.

That's the whole onshoring strategic supply kind of...

That's all part of Smart Capital.

Yes. I want to quickly go back and look at, and if you look at Intel itself, you are pushing on a couple of 3 or 4 different verticals simultaneously. You have the whole data center going on with Genoa and not Genoa, sorry, Sapphire Rapids and Granite Rapids and all that. You have the AIPC side going on. You have AI going on with Gaudi 2, Gaudi 3. And then you have the whole foundry strategy, which is a whole different universe. How do you balance this? How do you prioritize it? Do you see any risk when you are running multiple strategies at the same time? Obviously, investors want to see you succeed, but how are you planning on prioritizing all the different strategies?

It's a great question, something that we think about a lot. And I think the way that we have 2 very important notions. One is simplify the company and the structure of kind of having 2 companies in 1, say 3 operations in 1, I think, helps a lot. So we have 2 management teams. One management team that is focused 100% on making Intel Foundry successful, and they don't think 1 minute about the product side of the house. And we have a management team that is 100% focused on making the product side successful, and they don't think about -- they think about the foundry side as their supplier, a very close supplier but their supplier. And that separation helps create clarity. The second one is that we've been maniacally focused on in the foundry side on getting 18A and 14A working and getting 1 or 2 foundry -- large foundry customers. And so that's -- it's incredibly simple in that way. And on the product side, we have been maniacally focused on getting the next-generation products on the data center side, on the CPU side and on the client side, which is very, very clear, right, Lunar Lake and Diamond Rapids with obviously Clearwater Forest also being a very important product done right. And -- because we feel that those are the things that we monitor. And if we can get those things done right and on time with what we think the performance will be. And so that's how we -- we have 6 things we monitor in the company all the time. We have the top 6 jobs, and that's how we're driving it.

Got it. Fantastic. I'd like to take some questions from the audience. There's a couple of microphones around. So maybe if you want to ask some questions, here's your opportunity. He's the head of strategy. There you go.

Why don't you talk a little bit about your next-generation product, Granite Rapids? Can you talk about your confidence about where that's going to be relative to AMD or competition? How do you want to describe that? And when do we actually kind of see you, to your point, get the products back where they need to be, to get maybe just competitive, like, average servers and then AI [ customers ]?

Maybe just quickly introduce yourself...

Yes. Steve Smigie, I'm with Norris Research Group.

I didn't hear the second half of that question. I heard about the product leadership relative to -- I think you said Genoa or whoever...

Granite Rapids versus.

Okay.

[indiscernible] and then what are the products that you're like going to test in the data center with [indiscernible].

Yes, it's a great question. And in the spirit of not -- I'm not here to give -- make news. Here's what I would say. Granite Rapids and Diamond Rapids are 2 products that we were able to modify, change, and improve post Pat showing up. And they are products that are designed with a different mindset and a different set of vectors around product leadership. And so we feel that the -- they get to -- both gets to parity on certain critical things, but also real leadership on certain critical things. And obviously, we focused on our performance, we focused on cost, we focused on throughput, right? Given this a scalar-type workloads that in general, that we're focused on. And we focused on what I'd call manageability, all the things that are -- that our customers care about to be able to take on our product and run it as a fleet. We feel very confident on Granite. We feel even more confident on Granite. I think that's a fair great answer. I couldn't hear your second part of your question.

Steve?

Yes. Okay, it's working now. So it's -- I'm not trying to put you on the spot, right? Is it like, hey, we're pretty -- enterprise class is not the same as what the folks at the hyperscalers need. Do you think you're pretty close at parity in the next generation for server class, and that's kind of in the bag, you're done, you're ready to grow, not to lose share anymore? And then you're like maybe a generation behind, call it, a year behind on what has to go into the data center to kind of stop losing share there?

And that's a Granite Rapids specific point?

It doesn't have to be Granite Rapids. That was kind of the product that had -- like a couple of years ago, it seems like, hey, this one has a reasonable chance of kind of hitting what we need to get to. And I feel like it fell behind again, but...

I don't know if I will make a point around share. I think that there are many other things I'm going to share. What I would say is that Granite Rapids and Diamond Rapids make us very confident that we are -- we can compete very effectively with both the -- our traditional competitors and our new competitors.

Right. Thanks, Steve. Any other questions? There you go. On the back...

And I think you threw in photonics/optical. Do we need a materials transition, you think, to sort of deal with these issues, kind of the materials issues that the industry is coming up against?

The short answer is yes. The longer answer is, and we're working on it, as I said, right? So for my sense, the photonics business actually for Intel works for me, is in my shop because we're going through a transformation of it. What I would say is what we're finding for the work -- all the work that we're doing is that the big bottleneck is on network and access to memory. And there are different paths to try and solve that. One path, obviously, is to fix the connectivity problem, and photonics is a fantastic way of doing that. And to -- which means you have to get into co-packaging and a whole bunch of things that we are working on, that the industry is working on. There's other ways to solve the problem, which is more on the wafer side, right? So you could do it -- you could use glass as an example, to -- as a material to solve that. Again, this is something that we've -- I think we've made public they'll be spending a lot of time on, and we have capabilities through labs and throughout. And then there's another way of solving it, which is more at a system level of a card or a cluster, which is kind of where NVIDIA has traditionally, given all of the huge demand they've had on their clusters, is solving it, which is networking and better types of networking and better algorithms for networking. So it's a long-winded way of saying, yes, we need new materials. But I think there is lots of different vectors trying to address this issue. And it's probably going to be all of the above because the huge growth and the reality is that in these AI clusters, the GPUs are sitting around waiting for quite a lot. And it's the -- and so how do you get the GPUs closer to memory? Well, how do you get the GPUs really physically virtualized using photonics? Or how do you actually architect the system so that you can use the GPUs and the CPUs in a very different way? These are -- as I -- what I was saying at the beginning, which is we are going to very early on in how the system level or the chip level and at the networking level, how AI systems and AI platforms are going to be built. And these are all the things that we are kind of trying to -- we're working on that we believe we have, based upon our history, quite a lot of goodies to bring to the discussion.

Right. Any other questions? Saf, I think you had talked a lot about the hardware side, chip design, foundry, et cetera, et cetera, not much on the software side. What's happening on the software side? And where do you see opportunities there? Where do you see the build-out partnerships? I think we all would like to know.

Yes. So we think about software in 2 very different ways. We think about the software at the, call it, the orchestration layer -- orchestration of hardware layer. And there, we think there continues to be quite a lot of opportunity, both on the PC side as well as on the kind of the data center side to better orchestrate infrastructure that is deployed. And this is kind of our bread and butter. It's upstreaming capabilities of the hardware into the higher layers of the stack. It is driving better security. So we're investing a lot in security in terms of protecting at the chip level or the system level security. So we -- that whole area, we think, is a pretty big focus. And there's a guy called Greg Lavender who is focused on that. We think that's a lot of opportunity. If you go higher in the stack "above the frameworks," we also believe that and we're, as Intel Capital, investing a lot in 2 different places. The first place is on the enterprise side, where we believe that the enterprises are still -- are really struggling with an easy button for deploying AI systems for onprem deployment and use in a secure way. And we think that there's a whole ecosystem that needs to be put in place to drive, and we are actively working with the ecosystem to try and create that as a real alternative as a real sort of easy button to drive. And then on the client side, we think there's -- obviously, in partnership with our best friends, Microsoft, we think there is quite still a lot of innovation and software work to be done to really make the AIPC hum. And...

Obviously on the data center side, too. So...

And also on the data center side. But on the data center side, I think the CSPs are making a huge amount of investment.

Something like a CUDA or a ROCm from AMD. Yes. Got it. Fantastic. Saf, I think that brings us to the end of our slot, so really appreciate the time.

Thank you.