Davos 2024: A conversation with Pat Gelsinger

This transcript has been generated using speech recognition software and may contain errors. Please check its accuracy against the audio.

Linda Lacina, World Economic Forum: Welcome to Meet the Leader, the podcast where top leaders share how they're tackling the world's toughest challenges. In today's special episode, recorded at the World Economic Forum's annual meeting in Davos. Pat Gelsinger, the CEO of Intel, is interviewed by our founder, Klaus Schwab, all from the hallowed conference halls here at Davos, Switzerland. It's a conversation on trends and technology. You won't want to miss. Let's let them get started.

Klaus Schwab, World Economic Forum: Good afternoon and it's one of the settings I prefer during this meeting, instead of being on the big stage to be really in the discussion and having a conversation with people who really matter, and, Pat Gelsinger, of course, I don't have to introduce you Pat. You are the CEO of Intel, and you are here not only as a representative of Intel, but as a representative of a whole industry. I think your insights into an industry, which is so much discussed at the moment and the many aspects I think would be very valuable for all of us. We know after some kind of sluggish growth last year, I think the prospects for 2024 more robust and there are many issues to be discussed and Pat — I would like maybe as a first question to ask you, how do you see the evolution of — it’s a very general question — how do you see the evolution of such a market?

Pat Gelsinger, Intel: Yeah. And obviously a lot to talk about. And first, it's just an honour to be here with you, Klaus. And you know what you've done in creating the World Economic Forum. I just say for everybody who is looking for continued progress and continued peace, we owe you a debt of thanks.

So, thank you Klaus. In the semiconductor industry, the chip industry obviously coming out of Covid, you know, we went from potato chips to semiconductor chips. You know, the world knew what we were about. And we achieved what I believe were the two most consequential pieces of industrial policy legislation since World War two: the EU Chips Act and the US Chips Act. You know, to fundamentally say that this idea of semiconductors.

And I always like to ask people, what aspect of your life is not becoming more digital? Everything. Everything digital runs on semiconductors. Where oil reserves have defined geopolitics for 50 years. Where the technology supply chains is is more important for the next 50 years. And that's why we've seen this whole idea of the industrial policy around semiconductors to be so absolutely seminal to economies, to industries, to national security around the world.

And since then, you know, the world has woken up to the criticality of this industry. And we're well underway. You know, we've announced projects, others have announced projects to start rebuilding. And if we were here in 1990, 80% of the world's semiconductors would have been in Europe, in US, and today, 20%. You know, when we said by 2030 could we achieve a moonshot goal of 50/50, you know, where we truly have rebuilt the industries in US and Europe, which, by the way, almost all of the technologies and semiconductors emanate from US and Europe. This is our industry. Let's rebuild it. Let's create a balanced, resilient supply chain. And in the almost three years since we've been on this journey, and February 15th is three years for me at Intel, that we've made tremendous progress in beginning to rebuild those supply chains, but this takes a long time Klaus. You know, to build these factories is about five years here for them to come up in place. So three years later, guess what? They’re not up in place yet. You know, we're still got a lot of work to do, but all of the things are moving in the direction. And I'm super proud of the things that we've accomplished in that time.

Klaus Schwab: Pat, when we when we speak about semiconductors or chips, we have sometimes a feeling it's more or less the identical product. But actually, we should much more distinguish between different categories. Could you, could you explain a little bit and, and maybe also when you said 80/20% and now 50/50% if you relate it to the different categories.

Pat Gelsinger: I put semiconductors into four buckets: memory chips, you know, DRAMs, flash memories into speciality chips. You know, these would be sensors for cameras or RF chips for analogue radios, mature chips. You know, these would be things like for microcontrollers. And older cars and then leading edge. And today you know the leading edge market is hot — largely driven by AI. Because you know imagine if you have an old node chip. Maybe you do ten tops. If you have a modern chip you do 200 tops. You know, it's just dramatically different. You know, the performance capability. So those are the four major categories of semiconductors as I like to view it. And obviously when you're making factory commitments that take five, six years to come into manufacturing, you know, tell me what the demand signal is next quarter. You can't even answer that question, much less what the demand signal is five to eight years in the future.

And these are long lead time investments. So you end up with crazy cyclicality in the industry because, you know, markets go up and down, but you build factories that are five years into the future. And the only way to run a semiconductor factory, you know, huge capital asset. If you run it empty, it cost X. If you run it full, it cost X. So, you always run it full. And that creates some of these, you know, crazy economics around semiconductors. And today we're in a reasonable position of supply-demand balance. Obviously, we've had an oversupply of memory. We're probably going to go into an undersupply of memory because nobody's been investing in it.

We have lots of investment going into leading edge. You know, I expect there's going to be a lot of pressure on mature, because China in particular has been building a lot of capacity in that area. But now, as we think about the decade in front of us, AI is this insatiable demand for compute insatiable. We're going to be building the biggest machines that have ever been built, and we're going to be driving AI to the edge, that everything that we touch will be fuelled by AI capabilities. And it's going to need more capabilities as a result. So we're going to see this being, you know, I think we have ten great years in front of us for advanced logic requirements driving, you know, the industry and many of these factories that we're building right now.

Klaus Schwab: You mentioned cutting edge chips. May I ask you to come back to this 80/20, 50/50. How does it apply to those cutting edge chips. But maybe first, I want to ask you why did the west or the US particularly and Europe lose out? All those cutting edge chip factories — I think 95% are now located in Taiwan. If I'm not mistaken.

Pat Gelsinger: Yeah, Taiwan in particular, a couple of other, you know, Korea as well. But fundamentally, as I like to describe it, never was a vote taken in Brussels or DC to get rid of this industry. But there were votes taken in the Asian countries, Korea, Taiwan, China to get this industry. You know, they put in place long-term industrial policy, tax investment, etc., to attract this industry. And 30 years later they attracted the industry. And all of a sudden in Covid, we all woke up and we realized, oh my gosh, what happened? 30 years of industrial policy and we lost something so critical to our future. The world needs balanced, resilient supply chains and that's what created the environment for the US and the EU Chips act.

Klaus Schwab: Now when you look at, I mean already 2016, if I’m not mistaken China put out several development plans. But even today, China has not yet come up to the level of Taiwan. Why is it so difficult, could you tell us, to produce those cutting edge?

Pat Gelsinger: As I like to say, you know, if you want to be a cutting edge semiconductor supplier, you just need to invest 20, $30 billion of capital per year. You need to invest, you know, 8 to $10 billion of R&D per year. And you need to do that for 30 years.

Klaus Schwab: But how much do you invest?

Pat Gelsinger: In technology. You know, we do $18 billion of R&D per year. You know we, you know, we'll do over 25 billion of capital last year. That's going up next year. It will go up the year after that.

You know, this is extraordinarily long-term R&D and capital intensive. It's a consolidating industry. There's only three companies in the world that are capable of doing these cutting edge chips: TSMC, Samsung and Intel. You have key technologies like the EUV technology from ASML, key research IMEC on Belgium. Frankly Europe has many of the core technologies. And as I like to say, between US, Europe and Japan are the source of all of these core technologies. It's expensive, it's R&D, it's extraordinary innovation. And sometime Klaus, I'm going to get you I'm into a bunny suit on one of our fabs because you these fabs you know, they're almost $30 billion to build one of these complexes. These are the largest construction projects on Earth happening today. To build the smallest things that have ever been built on Earth. This is extraordinary, what's being done. And they're done then at scale over and over again, we're producing these extraordinary chips. You know, today our most advanced chip is about 100 billion transistors, you know, on a single package. And as we continue the progression of Moore's Law, and as I like to say, until the periodic table is exhausted, Moore's law is alive and well, and we're going to hit a trillion transistors by the end of the decade in a single package device. You know, this is just extraordinary.

Klaus Schwab: But if I understand correctly, it's not so much the development, which is costly. It's what you call the foundries. And can you explain why there are quite a number of manufacturers, but so few foundries, if I'm not mistaken there are five foundries or five to six, I don't know.

Pat Gelsinger: Lots of people can design chips and this idea of, you know, fabs of semiconductors, you know, it's R&D centric, but it's not capital and manufacturing centric. There's very few people that manufacture, you know, semiconductors. And only, as I said, three that can do leading edge, you know, today. And you know this is very expensive, very R&D intensive, extraordinarily intellectual property heavy. You know, we have, you know, 100,000 plus patents are associated with the leading edge of semiconductor manufacturing.

Klaus Schwab: 100,000?

Pat Gelsinger: Yeah, you know, associated with it. You know, it's just quite stunning what's been built up over the 50 year history of the semiconductor industry. And every aspect of your life needs more of them going forward.

Klaus Schwab: So if you build, for example, now a factory, I think in Magdeburg in Germany, is it a foundry? Or for cutting edge, or is it just a normal chip factory?

Pat Gelsinger: No, this is cutting edge.

Klaus Schwab: So it's cutting edge.

Pat Gelsinger: And this will be, you know, we expect when it comes online, you know, our most advanced process technology, you know, that we're just soon to bring into manufacturing is what we call 18A sub-two nanometres. You know, this will be beyond that. So this will be on the order of one and a half nanometre devices that we will build in Magdeburg. So, this will be not only the most advanced manufacturing in Germany, it will be the most advanced manufacturing in the world will occur, you know, in the Magdeburg site. You know, we're quite excited about that underway.

Klaus Schwab: I know that Chancellor Schulz is very excited indeed. I know how much he referred to you and probably saw you.

Pat Gelsinger: Indeed, put the most advanced manufacturing today in Europe, you know, is what we do in Ireland at four nanometres, you know, outside of us, you know, it's probably in the 10 to 15 nanometre range. So, this is a huge leap forward for all of Europe.

Klaus Schwab: So, if I understand you correctly, you may design chips and you may outsource the production to someone else to the foundry.

Pat Gelsinger: We design chips and we're a foundry for chips as well.

Klaus Schwab: The intellectual property is you?

Pat Gelsinger: Correct, for the foundry. But the intellectual property for the design is with the design company.

Klaus Schwab: Yeah, which you are both now. Yeah.

Pat Gelsinger: Essentially I'm rebuilding Intel. We're taking one broken company and we're creating two distinct companies inside of Intel, a manufacturing and a foundry company and a product and fab company inside of one.

Klaus Schwab: Thank you. Now, you referred to Artificial Intelligence, and this is one chip factory. I don't want to mention the name, which is very much talked about because, its chips are particularly related to AI. Could you explain? And are you also in this area? What is special about AI?

Pat Gelsinger: Yeah. And you know, this idea of high performance accelerated computing. And I give credit to my friend Jensen, they have pursued that domain pretty consistently for 20 years, you know, high performance, throughput computing. And then he got lucky with AI. Really lucky. But he had steadfastly pursued that architectural pursuit of high-performance computing, largely in graphics, you know, for a period of time. And then AI happened then, remember, AI is an overnight 50-year success.

The foundations of AI, you know, were in the late 60s. And then for over 40 years, nothing happened. When I was the architect of the 46 in the 80s, I was going to make it a great AI chip. What happened in the 80s and AI? Nothing. And then all of a sudden, the compute got big enough, the data got big enough, the algorithms got good enough. And we've seen this explosion of AI. And right now we're seeing these very large systems for training. And we think about, you know, GPT and OpenAI and so on. But the next several years of AI won't be around big model training, you know, for hundreds of billions and trillions of parameters. The next several years of AI, in my estimation, is how do we use those models. How do we deploy them, how do we inference them, and how do we move them out of the big data centres into every device that we use?

We've just launched what we call the AI PC. How do we make, you know, instead of having 100 billion parameters in the cloud, how can I put 10 billion parameters of your data on your PC that you're operating locally? And I call it the three laws of edge and AI. You know, one is the laws of economics. If it's on your device locally, much more cost effective 10, 100 times cheaper. Your second is the laws of physics. If I have to round trip to the cloud, the speed of light is still the speed of light, versus doing it locally. And the third is the loss of the land. The regulatory requirements. Am I going to take my real time factory data to somebody else's cloud environment for local inferencing of my manufacturing line? Absolutely not. 80% of the data is still on premise and privately held. So, because of that, we see this idea of going from training, which still has exciting science in front of it, to much more inferencing, deployment and edge applications. And that's why I'm quite excited, because there's going to be both. We're going to continue to see the science of AI evolve, you know, but very much the deployment and utilization of it was where we believe, you know, the primary action will be for the next couple of years.

Klaus Schwab: So the research comes mainly from insights of companies, not from university and from the scientific world today, because at the origin it was mainly the scientific world which drove the semiconductor development, you know, and of course you were a big beneficiary. In the first stage of career, you were [inaudible] one of those trials. But today can you tell us how much the research is driven by industry and how much by the scientific world?

Pat Gelsinger: We've seen this shift of more and more of the research and development to industry over a sustained period of time. You know, one piece of the Chips act was you know, the establishment of NSTC in the US. You know, we had hollowed out that leading edge research. And one of the things I believe, you know, needs to for, you know, Western democratic-oriented countries, we must rebuild that long-term research. You know, long term research is 30 plus years. AI took over 40 years to maturity. The transistor took over 30 years to maturity. You know, we must keep that long-term investment. And that's primarily, you know, from governments public sector, start-up or like things. We have lost that because businesses like mine we're sort of like 15 years or less in the R&D. And we've seen that shift over time. I think that's very unfortunate.

And now as we think about the science of AI, you know, in particular, you know, today you know, you remember Daniel Kahneman's book Thinking Fast, Thinking Slow. You know, all of our AI systems today are thinking fast. We haven't brought reasoning into AI. It's a huge area of research. And most of the foundations of that, you know, were seeds planted 20, 30 years ago about how we can prove correctness of systems. Today our systems hallucinate. Tomorrow, if we're going to use them broadly, they have to be right. So how can they be fast and right? Just like, you know, we've seen Thinking Fast and Thinking Slow, huge areas of research in front of us. As I like to say until the periodic table’s exhausted, we're not done. Well, we have about two thirds of the periodic table to go. Still lots of new science, new materials to be discovered. You know into the future. This is you know, I think for a technologist the next couple of decades, I just wish I was still in my 20s, the next couple of decades are going to be absolutely.

Klaus Schwab: It’s in the exponential phase now.

Pat Gelsinger: Oh, my gosh, it's going to be so good Klaus. You know, how old are you? You and I need to find a way to get younger. Because it’s going to be a thrilling few decades.

Klaus Schwab: We keep young, you see. Coming come back to this 80/20, semiconductors are essential element for national defence and national security. Why did the US particularly, but also Europe, have it slipping out.

Pat Gelsinger: I think people were somewhat asleep at the wheel from the policy perspective. There were, you know, a few people raising the concerns. But remember, we were in an era, where it was all about lowest cost supply chains, you know, hey, we don't care if we manufacture, you know, we're going to create the intellectual property. We'll invent the algorithms, we'll write the software. You know, we're sort of happy to bring other countries in the world into that manufacturing. That's hard, that's low-end, we really don't want to work, right? You know, we want to invent.

And I think that was the overall general views, quite literally, of the Western world for a couple of decades. Then all of a sudden we realized — and critical industries, and I think the automotive industry is facing this now — we could lose the entire automotive industry because we've lost the supply chain. And the semiconductor industry, you know, literally, we were on the hair's edge of never being able to recover the manufacturing of this industry.

You know, I've said to some that, you know, had I started the rebuilding of Intel a year later, I don't think I could have accomplished it. It was that much on the edge that we would have outsourced too many generations of technologies and never been able to rebuild it in the future because, particularly for semiconductors, R&D and manufacturing are co-mingled. A factory without R&D in the semiconductor industry, you know, essentially becomes a boat anchor within six months.

You know, what is that? You know because you're constantly doing research in the factory, in the manufacturing process itself. R&D and manufacturing are inextricable in this industry.

Klaus Schwab: What is dependence of the industry on rare materials? Is this a limiting factor?

Pat Gelsinger: It's actually surprisingly not that terribly constrained, in that sense, you know, we don't use that many materials. Remember, God's gift to mankind was silicon. Sand, the most prevalent material on Earth, is our core material. So, he gave us lots of it. You know, and the refining of it's in, you know, Japan and Asia, in Europe. So the core materials are actually not that severely limited, other portions of the supply chain, but silicon itself isn't that bad. But we are taking steps to make sure we have built resilience and sustainability, you know, the environmental aspects of the chemicals that we use is a super important factor to us. You know, Intel is decades ahead of anybody else in terms of the sustainability. You know, we've committed to by 2030, 100% renewable in our factories in our type 1 and type 2, you know, full sustainability of our water use, 2040 net zero, you know, so it's a huge and important topic for us over time. And any place we find critical material limitations is how do we go, you know, expand that. Because remember, most of the critical minerals issues are not minerals issues, they're refining issues. You know, where are the minerals? You know, they're usually fairly available. It's where are they refined are the key points of constraint.

Klaus Schwab: But could energy be a limiting factor? Because some people are concerned about the high consumption of electricity.

Pat Gelsinger: Today when you think about the IT industry as a whole, different studies, but probably somewhere in the 3 to 5% range of energy consumption, you know, goes for IT infrastructure today, different studies would indicate, and particularly with the surge of AI, that that could triple or quadruple over the next decade. So that could become quite constraining, you know, to build these big, massive data centres, which, by the way, puts pressure on us. How do we keep reducing the power requirements of our chips as well as, you know, things like water cooling, so that we can increase the efficiency of them as well. Another area of significant research.

Klaus Schwab: I come back to China. When do you think China will have caught up with really the frontiers of chip technologies?

Pat Gelsinger: The export policies that have been put in place, you know, recently we've seen the Dutch in particular, you know, in place the US policies, Japanese policies, etc. it sort of puts a floor in the 10 to seven nanometre range. And you know, we're racing to go below two nanometres and then one and a half, you know, we see no end to that in sight, you know. And I see the policies in place — and it's not like China's not going to keep innovating. But this is a highly interconnected industry. You know the staging of Trumpf. The mirrors of ZEISS, you know, the equipment assembly of ASML, the chemicals in resist in Japan, the mass making of Intel, all of those together, you know I think this is a ten year gap. And I think it's a sustainable ten-year gap with the export policies that have been put in place today. And I do believe that that portends well for the policies that have been established in the world, you know for export and competitiveness in this environment. And, you know, we're leaning hard into making sure that's the case now.

Klaus Schwab: So, so many investments going on, like in Magdeburg and so on, I think there are plans in France, plans in Ireland, additional ones. Aren't you afraid, because every country wants to have now its own asset, aren't you afraid about over-capacity?

Pat Gelsinger: Well, if you go back to the beginning of the discussion. We're making eight, you know, five and eight year investments, and we have cyclicality quarters. I mean, you can't even tell me what the demand is next quarter, you know, much less five years from now. That said I think fundamentally compute and AI in particular are going to be like the gas law. They will expand to consume every amount of compute capacity that is economically available. You know, so I believe that becomes such a driver of demand that I think advanced logic capabilities — I have no concern in this at all. Now, of course, when I explain to my board my crazy capital request, I, of course have to say that. Not only say that, but say it with conviction.

Klaus Schwab: Now we have, I think, five minutes to go. Let me see is there is any question here in the room, sir? Yeah, please.

John Chipman, International Institute for Strategic Studies: John Chipman, executive chairman of the International Institute for Strategic Studies. [inaudible] — is that true? And the second question I have is I am equally told that when Morris Chang establishes a foundry in Arizona or elsewhere, he does not have the confidence that that foundry can reach the same level of expertise as his own in Taiwan for reasons of human capital. So, can you explain when you're producing something 80,000 times bigger than a human hair, why is the human skill so important?

Pat Gelsinger: Yeah. So, you know, on the first topic, the way I like to describe it is Moore's law is like driving down the road on a foggy night. You maybe see 100m in front of you when you've gone 20 or 30m down the road, you see about 100m in front of you. And Moore's Law, we see about a decade in front of us. And I've been in the industry now for 42 years. I’m almost embarrassed to say it's been that long, but you know, with it I've heard about the death of Moore's Law, you know, for about three decades. And I sit here today telling you I know how to build a trillion transistor chip by the end of the decade. Moore's law is alive and well, and we're going to keep making it true because it's not a law of physics. As you suggest, when you get to one nanometre, you know, it's a law of innovation, of commitment, of economic pouring into the criticality of this technology. And we as a company, Intel, are going to keep making it true. Second: TSMC is an extraordinary company. You know, they have refined a manufacturing capability and a customer orientation that is really unmatched in the industry. However, they do R&D and manufacturing in one place: Taiwan. Intel has operated with a distributed manufacturing footprint for over four decades. We know how to operate in different segments of the world. We know how to do this. I see my best friend, Senator Portman here from Ohio. We're about to start the Silicon heartland in the centre of the United States, bringing manufacturing back to the heartland of America. You know, with it we know how to operate distributed manufacturing. They don't yet. And this idea of tethering, of R&D and manufacturing. He's right. They only have one place of R&D. You know, I have a distributed R&D model. I have a centre of R&D in Oregon and Arizona that are extraordinary in the world. We're absolutely convinced we know how to do this. Let the race continue. They're a great company, but we're committed to being back to unquestioned technology leadership in the world. And I think by 2025, we will have proven that to the world with manufacturing of sub 2 nanometre chips that are simply the best in the world, you know, based on Western technology. And we're pretty proud to be in that race.

Klaus Schwab: How many people do you have in research and development?

Pat Gelsinger: We're on the order of 50,000 engineers total. In the TD space, you know, in the core technology space on the order of 20,000 years.

Klaus Schwab: Impressive.

Pat Gelsinger: Yeah, this is at scale R&D.

Klaus Schwab: We have time for one last question.

Pat Gelsinger: You know, Senator Portman, their.

Rob Portman, US Senate: [inaudible] the difference the Chips act makes. And as a co-sponsor of that bill and having heard from a lot of my Republican colleagues about industrial policy and how that's not our direction, some of them thought we were overreaching and doing too much. The reality is there will still be considerable dependency even when you combine the EU and the US and you touch on that a little bit. But if you could talk a little more about that, I think going 37% of chips 30 years ago to about 12% today, and even after $52 billion and about $39 billion on the manufacturing side, my understanding is we will only be at what number in terms of the percentage?

Pat Gelsinger: Will be crawling our way back into the 30s. You know, with chips one, you know, I do believe there'll need to be a chips two at some point to continue building those policy decisions. And remember, these were bad policy decisions in the US and tax policy and capital investment policy, long term R&D investments, all of those aggregately caused us to go from 80 to 20. To rebuild that, we're not going to do that in one Act as good as it was — thank you Senator — you know, in a few years. This will take decades to rebuild this industry. I really believe that chips one, chips two fundamental industrial policy, you know, improvements in R&D, improvements in manufacturing, capital policies, you know rethinking of some of the financial requirements long term, human capital development. I firmly believe that we can rebuild this industry. And it's the thing for our economies, it is the thing for industries and it's the thing for our national security. And we're deeply committed to help drive that. You know, simply put, it's the thing for the heritage of Intel. It's the right thing for the technology industry and it is the thing for the world. We're going to make it happen.

Klaus Schwab: Thank you Pat. We have come to an end of this discussion. And I think what we have seen as a leader with something which we so much need today when we talk about leadership, it's conceptual thinking. And you are in one of the most complex areas: High tech, at the edge. And I think we all leave this with a better knowledge about the industry and very impressed by your by your leadership. Thank you very much.

Linda Lacina: That was Pat Gelsinger and Klaus Schwab. Find a transcript of this episode, as well as transcripts from my colleagues’ episodes of Radio Davos podcasts. Me and my colleagues are covering the Annual Meeting all week, so make sure to follow us on social media with the hashtag #wef24 or online at weforum.org for the latest insights from the world's top leaders.

This episode of Meet the Leader was produced and presented by me, with Taz Kelleher as editor and Gareth Nolan driving studio production. That's it for now. I'm Linda Lacina from the World Economic Forum reporting from beautiful Davos. Have a great day.