Intel CEO Pat Gelsinger on Intel's place in the semiconductor industry

Good afternoon, everyone. Thanks so much for coming to our MIT Manufacturing Distinguished Speaker Series, where we have the great honor of hosting the CEO of Intel Corporation, Pat Gelsinger. In the format of the hour will be a conversation on the future of semiconductor manufacturing. So to make formal introductions, please welcome Professor Anantha Chandrakasan, Dean of the MIT School of Engineering. Thank you, John, and good afternoon, everyone. I'm delighted to be here today to welcome Pat Gelsinger, the CEO of Intel to MIT. He started at Intel in 1979 as a teenager and eventually became the company's first chief technology officer. During this time at Intel, he was behind key technologies we're familiar with today, including the original 8486 processor. His work helped Intel become the world's preeminent microprocessor manufacturer. In 2009, he became President and Chief Operating Officer of EMC, where he was responsible for storage, backup, and data computing products. In 2012, he became CEO of VMware, a leader in cloud infrastructure, enterprise mobility, and cybersecurity. He rejoined Intel in 2021 as CEO and almost one year after the start of the pandemic. In his role as CEO, he leads more than 130,000 employees across 46 countries. He's committed to significantly expanding semiconductor manufacturing in the US. Intel is adding two fabrication facilities to its Phoenix complex and opening a new facility in Columbus, Ohio. He's focused on making the company the leader in chip manufacturing and packaging and revitalizing the semiconductor industry in America. He's been in close touch with Washington, D.C. and has strongly supported and driven the CHIPS Act. He holds engineering degrees from both Santa Clara University and Stanford University. In 2019, he was ranked the top CEO in America in Glassdoor's annual employee survey. And just last month, he was elected to the National Academy of Engineering. Please join me in giving Pat a warm welcome. I would also like to welcome our other distinguished panelists, Professor Daniela Rus, who is the Andrew and Erna Viterbi Professor of Electrical Engineering and Computer Science and the Director of the Computer Science and Artificial Intelligence Laboratory. Professor Vladimir Bulovic, the Fariborz Masi Professor in Emerging Technology and the Director of MIT.Nano. And Professor Jesus DeLalamo, who is the Donner Professor and is Professor of Electrical Engineering in the Department of Electrical Engineering in the Department of Electrical Engineering and Computer Science. So please join me in welcoming our panelists, and now I pass it over to Jesus. Welcome. All right. Thank you, Anantha. Welcome, everybody. Pat, welcome. It's a real treat to have you here today for this conversation. So we have a few questions for you. Maybe the first one should really be about a much commented issue, the end of Moore's Law, which has been in the media for far too long, and in spite of all this, the semiconductor industry keeps making tremendous progress for years. So still the question is, how far can we go? At some point, integrating more transistors might not be economically advantageous. So how will this industry continue to increase the power of electronics beyond that point when making more transistors is not worthwhile? Yeah. And a few thoughts on this. Number one is, I think we've been declaring the death of Moore's Law for about three to four decades. Right? And it's one of those, it's like, all right, rumors of my death are highly exaggerated. And, you know, to some degree, you know, we have about a decade of visibility in front of us. And when we've gone three or four years into the future, we have about a decade of visibility in front of us. Right? So, you know, it's like, you know, we keep solving problems that allow us to keep rolling forward about a decade in front of us. Right? And, you know, that's the way it's been. Now, fundamentally, you know, we had what we called the Denard era of Moore's Law, where essentially you were just able to shrink in the XY, right? And being able to do 0.7X shrinks in the X and the Y, right? You're able to get this doubling every two years, approximately. And that was like the golden era of Moore's Law. You know, we're no longer the golden era of Moore's Law. It's much, much harder now. So we're probably doubling effectively, you know, closer to every three years now. So we've definitely seen a slowing. We've seen a shift in the economics associated with Moore's Law. You know, a modern fab, you know, seven, eight years ago would have cost about $10 billion. Now it costs about $20 billion, right? So you've seen a definite shift in the economics. But here we are today. And the most advanced chip that we build today is about 100 billion transistors, right, on a single package. Well, that's a lot of transistors. Now, where we are today, we believe that four major things are allowing us to get to a trillion transistors by the end of the decade, which is approximately doubling every two years. And with that, you know, one is we have a new transistor type, right? We're going to this gate all around. So we've solved the transistor problem, right? We have a new power delivery mechanism. We call it ribbon-fed, essentially creating a wafer sandwich, putting power delivery from the backside as opposed to the top side. So you solve a lot of power, droop, you know, delivery, density issues by doing that. So number two. Number three, lithography. Lithography has been the heart of semiconductors ever, ever, ever, right, you know, for it. And with EUV, right, extreme ultraviolet, essentially using X-ray, you know, we went from 193 nanometer light down to 13 and a half nanometers, right? So you saw a dramatic, right, decrease, right? It took 20 plus years and was a decade late, you know, but that is now solved and we have the next generation of that called high numerical aperture lithography, right, you know, which will have the first machine in our Oregon facility by the end of this year. Yeah, first one on the planet, right, outside of Europe. So lithography is solved. But number four is that chips themselves are going from being 2D to 3D, right, where now you're using, right, more, I'll say, legacy wafer techniques becoming advanced packaging techniques. And as a result, we're stacking and being able to essentially innovate in three dimensions, right, as opposed to two dimensions. And as a result of this new use of two and a half and 3D packaging, you know, we see this ability to bring a trillion transistors into a single package by the end of the decade. So against that, you know, for all of the critics that declare we're dead, right, and we think of Intel as the steward of Moore's law, right, you know, until the periodic table is exhausted, we ain't finished, right, you know, we're going to stay on that journey. Moore's law is alive and well, and if you want to help it be true to the future, come on down, we're happy to have you, right, if you're good, right, right, you know, if you're really good, we'll hire you. And sorry, you know, I'm a Stanford grad. So right, I'm not sure I would have made it if I was at MIT. But anyway, you all are pretty smart, too. So we are the stewards, we're going to keep it going and a trillion transistors by the end of the decade. Wonderful. That's a great, great future ahead of us. So I recently learned something interesting. I learned that the energy consumed in the fabrication of the chips that Apple uses is the single biggest item in Apple's carbon emissions. In fact, the chips that they use every year to build their devices consumes, making those chips consumes more energy than all the Apple devices worldwide being used on that year. And in fact, also, more advanced generations of semiconductor technology consume more energy in their making. So what's the path for greening this industry, which is a big concern at this time? Yeah, and you know, it's super interesting. And, you know, Intel is unquestionably been the environmental leader, right, in semiconductors for decades. And this was a huge passion. You know, our three founders, I call them the trinity of Silicon Valley and the silicon industry, you know, Gordon Moore, Robert Noyce, Andy Grove. And Gordon Moore was particularly passionate on this topic. Right. You know, and, you know, climates and environmental and so on. So it's deep in the ethos of Intel from our very founding. So today, if you look at, you know, for instance, energy use, we're about 90 percent green energy today in our fabrication facilities, which is an extraordinary statement. When I was being, right, you know, giving my Senate testimony to support the CHIPS Act, right, passage, I would view it as more of an assault by senators on you if you're doing Senate testimony, you know, all deference to our legal ridden, you know, Senate leaders. But they, you know, one of the senators pointedly asked me, he says, so if we give you CHIPS money, do you commit to being net zero by 2050? And I said, Senator, no, we're going to do it by 2040. Right. Now, of course, most of the other people that were testifying for the Senate had no specific plans yet. You know, but we are well advanced in this area. But to us, it's not just energy and green energy. In that sense, it's also water, right? You know, and like our Arizona fab, you know, we have a fab in the middle of the desert of Arizona. We're about 92 percent water recycling. And if we combine that with all the water reclamation projects we do, we're over 100 percent water, right, for, you know, it's also things like chlorophyllides and other, you know, noxious materials and how we take those out of the manufacturing process as well. And it's just one of these things. You just got to get on it. Right. It's good engineers, good scientists, good environmentally conscious one by one. But you know, we stand in stark contrast to the rest of the industry. You know, we are decades ahead of, you know, Samsung's, TSMC, you know, in this regard is something we're quite proud of. So if you want an environmentally conscious chip, you need Intel inside. So Ann Kelleher, who is the Intel CTO, has said recently that Moore's law is and always has been driven by innovation. And so we know also that there is a deep connection between innovation and manufacturing prowess through the learning curve. The more you make, the better you get at it, the cheaper your products are going to be, the more competitive you will be. And so as manufacturing volume and ability, therefore, manufacturing volume and innovation go hand in hand. And perhaps this is why the industry has consolidated in the last few years to essentially three companies that are sitting at the leading edge. So if we look forward, what does this mean for the future of semiconductor manufacturing at the leading edge? And what does it mean for Intel? How many of you believe we're going to have a more digital future? Okay, duh. And everything digital runs on semiconductors. There is no digital without semiconductors today. And as you think about that digital future, and one of the things that has driven my passions and coming back to the company, the CHIPS Act, et cetera, is where do you want that digital future to be built? In 1990, 80% of semiconductors were built in the US and Europe. Today, 80% of semiconductors are built in Asia, with a huge concentration in Taiwan, one of the most geopolitically unstable areas of the earth. And let me just give you three factoids about Taiwan. Other than not having any natural resources, not having energy, no energy reserves, sitting on a major fault line, and now being essentially tapped out of its talent capacity, they don't have an MIT locally, they have good schools, and so on. So essentially, you have this extraordinary world dependency on a very small area of the planet. This is not good for the geopolitical stability, the resilience of our supply chains, and this became acutely visible as we went through the COVID crisis. So what do we want to do? It's not like we, you know, de-globalization, that's not the right way to view it. We want balanced, resilient supply chains spread across the world, and that's what we're out to accomplish with the CHIPS Act and what Intel is driving quite aggressively. Let's build the fabs where we want them. Geopolitics has been defined by oil over the last 50 years. Technology supply chains is more important for a digital future than oil for the next 50 years. Build them where we want them, and we can build them right here, right at that point. So we've taken a huge agenda to establishing this balanced, resilient view of our technology supply chains. You know, it's not just that we want to, you know, I have nothing against software engineers, but at the end of the day, it runs on hardware. Our digital future runs on semiconductors, and semiconductors is a volume manufacturing game, right? The engineering, the innovation is never done, right? A fab, you know, is not something you turn on and just pull wafers out, right? You are innovating in that fab. Every single lot, wafer, telemetry, readout, et cetera, is learnings on defect densities, new, you know, material understanding, you know, new defect models, you know, new equipment efficiencies. This is a living, breathing, manufacturing organism if you're going to be in semiconductors, right? Go for it. And this is one of the things that the founders of Intel deeply believed. You can't innovate and not manufacture. Those are inextricable, right, in our industry. And if we're going to be an innovator at the heart and soul of the digital future, we must be a manufacturer at scale. So this, you know, deep in it, you know, what I'm driving as a company, you know, some might think, you know, there's a thin line in, you know, being a CEO between being bold and crazy. And right now, Wall Street's not sure which of that line I am on, right? You know, for, you know, before I became CEO, Intel gave $70 billion to shareholders and put $50 billion into capital. The five-year plan I have laid out for the company is $150 billion of capital and $20 billion to shareholders. This is very expensive. This is very hard. But these are also great industries for us, you know, great jobs, you know, the most innovative PhDs in material science, all the way down to advanced construction jobs for these facilities. Because these construction sites, you know, a $20 billion fab is an extraordinary statement. It is one of the biggest construction projects on the planet, but it is also one of the smallest, most refined, you know, materials are being produced on Earth. So it's not only the biggest, you know, but also the most precise as well. You know, and that's what we're doing with these projects is reshoring, rebalancing our manufacturing, leading in the core technology for the digital future and doing it, right, in the U.S. and Europe, you know, what I believe is essentially the future, right, of the planet rests on what we're doing in building our technology future. If that doesn't inspire you, why are you here? I'm inspired. I'm inspired by your vision for sustainable manufacturing and for connecting innovation and manufacturing. And I think we're going to need increasingly more of that because AI is here, AI is revolutionizing the marketplace. With Chad GPT, AI has become democratized. It's in everyone's pockets. But in order for AI to do its magic, it needs the platforms, it needs the computers. So my question is, what is Intel doing for the development of AI hardware and how do you see that as a competitive advantage? Yeah. You know, one of the things, a little bit of my own personal story, you know, I was, you know, I started at Intel so young, right, you know, I went through puberty there, right. You know, I feel, you know, I grew up at the feet of Grosvenor, Noyce and Moore, 30 years with the company. Then I took an 11-year vacation, right, you know, to EMC and VMware, which, by the way, I was outside of Intel almost to the day, the same time Steve Jobs was out of Apple, right. It's like, okay, you know, it's like the death of the vision. I wanted to be the CEO of Intel. But when I was pushed out of the company 13 years ago, they killed the project that would have changed the shape of AI, right. I had a project underway, it was called Larrabee at the time, to do high throughput computing in the x86 architecture. They killed it after I left. NVIDIA had unmerited potential in the space of AI hardware, essentially throughput computing versus scalar computing. Nobody put pressure on them. And Jensen, by the way, you know, I mean, he's a very, I know him super well, right. He worked super hard at owning throughput computing, primarily for graphics initially. And then he got extraordinarily lucky, right. I mean, they didn't even want to support their first AI project, right. And it's just, hey, big flop machines was needed, you know, for these big data sets that they were looking at. So Intel basically did nothing in the space for 15 years. And now as I come back, I have a passion, okay, we're going to start showing up in that space. And, you know, our strategy here is number one is democratize AI. Because the day you see the collision of high performance computing AI, you know, for these hundred billion, soon going to be trillion dollar, you know, parameter models that we're training on that take, you know, 30, 60, 90 days for training. How many of those can you do? Right. I mean, you know, what is the ultimate of, you know, like high performance computing, weather modeling, nuclear simulation, high end training, you can't afford it, and so on. So what are we going to do? We want to democratize AI. We want to make it available to every application developer, you know, that you can put it under your desk, it's in your laptop, our next generation client products will have 20 tops, right, in a standard mediocre, right, you know, PC, okay, you know, it's 20 tops, right, you know, that was more than, you know, it was possible on a high end supercomputer about 15 years ago. So, you know, democratize and also open up the software stack, right, for industry standards, get away from the prior proprietary technologies, and we're driving a technology called SYCL, an open standardized parallel C, so that we can eliminate proprietary technologies like CUDA. And, right, we begin to offering our high end computing trips to compete in that high end training space as well, build it into our standard Xeon processors to truly make AI available for any, for any and all, because it is the most important active workload today. And, you know, hey, it's evolving like crazy, right, you know, it's like, you know, I helped to create IEEE P854, right, which was the floating point standard at 64 and 80 bits. Okay, now we're talking about FP8, you know, I mean, it's like, what, right, you know, such trivial low end data types, and are, you know, suitable, right, you know, because instead of big, big vectors, you know, let's create little things that are guessing right at the right answer. And AI is just sort of a probabilistic guess for most things. So, you know, it's changing rapidly, sparse matrices, different algorithmic models, breakthroughs happening, you know, rapidly. And if we think about AI, you know, essentially, you look at something like chat GPT, I mean, incredible technology, but it's on the simplest data set, text, right, we're going to start getting to some hard data sets, right, going forward, which are going to be orders of magnitude more challenging than the ones that we're solving today with AI. So, you know, I think we have at least a decade, probably two decades of just sheer innovation in front of us. And all of it's going to need extraordinary amounts of data, extraordinary amounts of networking, and big honking compute. So we're going to build lots of fabs that we can build lots of compute. So just following up a bit on that, I'm wondering, so is your, do you see AI evolving in terms of building increasingly bigger models? Or do you see other aspects of AI taking flight? And what do you think is the role of hardware versus software, so that we can achieve your democratization vision? Yeah, and you know, I mean, there's huge, huge activities right now. How do you take these enormous models that are produced by the training and be able to crush them down into something that you can put on to a single node in the server, right, or even into a desktop, right, you know, and be able to compress those models. And I think that's going to continue in this idea of democratizing AI, right, you know, we can't have these super high-end, high-performance computing machines as the only way that you can do model training. You know, we have to bring it down. There's also dramatic innovations going on in, you know, I'll say, you know, just the algorithmic domain of AI as well, as we're going to larger memory spaces with sparser representations as well, you know, if you view computing as a world of scalar computing versus throughput computing, you know, well, when you get to sparser matrices, it's sort of those worlds are coming more together, right, and you're combining vectors and high-throughput matrix structures as well, you know, it's not clear that the algorithmic domains that are dominating AI today will continue to dominate, you know, in the next couple of years of innovation that are going on. And in that, I'll say, you know, if anybody tells me that you can predict what the next major AI breakthrough is, you know, I doubt that you have figured it out, because a lot of this is we're doing, you know, heuristic understanding of different algorithms that have been proved effective, you know, combining them together against certain training data set, data set processes, and coming up with tremendous insights, you know, that result from it. And I expect that that's going to be the case for the next decade, you know, for it. And, you know, hey, some of them, hey, maybe you're going to find, you know, the Xeon CPU is the only way that you do that problem, and we're going to be lucky as well. So I would hope so. But, you know, and also just one more point on this is, as these models, you know, they're also now going to start touching on different aspects of computing as well. How do you secure these models? Right. You know, as you know, if you're going to spend, you know, essentially, you know, a billion compute hours creating a model, right, are you going to give it away or is it going to be protected? Also, how are you going to do distributed retraining of models? Right. You know, because you may or may not want your medical data, you know, being exposed in a model and the biases associated with that. How are you going to have competing models to make sure that the models themselves weren't trained with biased data in the first place? You know, right. You know, there's going to be lots of problems that emerge as we start really understanding some of these next algorithmic domains. It's a thrilling time to be a computer scientist, but it's even more thrilling to be a semiconductor manufacturing engineer. But the two are married. You cannot have one without the other. So let's see. How do you think of the other big players like Google, Microsoft? How do you see your relationship with them and do you compete with them? Do you collaborate with them? In many ways, you know, these are big customers of ours, right? And Google, Amazon, Microsoft, Alibaba, you know, these are some of my biggest customers, so I have to like them, right? And most of them have some level of programs that compete with us as well, where they say, huh, you know, I want to go do a TPU because I want to do a TPU. I have engineering, I have innovation going on. And even for those programs, as we become a foundry for companies for the world, I still want to capture their designs on my foundry. So even in that case, hmm, you know, I would prefer they use my chips, my designs, but even not, I want them to use my wafers as well. So in many cases, it's like, and you know, this bubbling cauldron of innovation, right, that's going on. It's super hard to decide who's a competitor and who's a collaborator, right, because there is so much going on in this space. You know, and the technology industry has, you know, one is we've thrived on open, right, and we deeply believe in open standard innovations where, you know, we get to stand on the shoulders of the work that has gone before us, you know, something we deeply believe in at that level. But it's also thrived on coopetition, right, because it's not clear some days who's going to be your competitor or who's going to be your customer, right, the next day. So this combination of open innovation, you know, combined with coopetition, right, it's like I like them all because they might be my biggest customer tomorrow, right, even though I'm competing with them in certain areas today. So that's amazing. May I ask another question? Okay, one more. Two more. One more. Okay, so I wanted to ask you about your workforce, about your employers, employees, how, what does Intel do to ensure that everyone in some sense understands AI, understands the power, the transformational power of AI? How do you prepare your workforce for this future where your hardware will be increasingly targeted to AI? Yeah. And, you know, as we think about it, it's, you know, it's always an even hard times, good times. And right now it's a hard time as we've done some reductions in the company. We never stop hiring new college grads, right, you know, it's like the lifeblood of innovation is, you know, continue to bring in, you know, next generation talent into the company. And we would say that's something super important, you know, because, you know, right, these kids don't realize what's possible and not, right. That's sort of the source of innovation, right. You know, when you get old, you start limiting yourself and you're thinking, oh, we've always done it that way. Right. You know, so you keep that source of innovation. A lot of that is driven by these next generation technologies like AI coming into the company. Well, why have you do it that way? Have you tried this? You know, right. You know, you keep that activity. You know, also we make sure that we keep a very vibrant research community, you know, and Intel Labs is a very active, you know, research community where we're, you know, participating in this bubbling cauldron of innovation, very actively research papers, open source projects, you know, engaging, you know, with research grants. And, you know, by the way, I was actually pretty startled in my briefing coming here. MIT is my number two research funded university. Yeah. Number one. That's an MIT thing to say. I was pretty startled you were number two. So now you're number one, you know, but we don't hire enough of you. So that's a different problem. Right. But we'll, but, you know, part of that is, you know, you're always in the technology industry. You know, what was good yesterday, right, is standard today is legacy tomorrow. Right. You just got to be constantly going, you know, through that innovation cycle. How do you approach re-skilling and training to make sure that your workforce is ready for the transformation of tomorrow? Painfully hard. You know, and I just, you know, re-skilling, re-training, you know, workforce development, et cetera. And particularly a lot of our jobs, you know, like in our fabs and so on, there aren't good programs anymore, right, in that area. So a lot of it is bring in good raw talent and immediately put them into the settings, hands-on, you know, developing them, you know, but we're also constantly investing in, you know, workforce development for, you know, internally as well as for the future. And when, you know, one of the proudest days of, you know, being the CEO of a turnaround, you have good days. You know, every CEO has good days and bad days. The CEO of a turnaround, you have more bad days than good days, right, for, but last year I had five great days. Not good days. I had five great days. And one of my great days last year was the Ohio groundbreaking, right. And you know, when I got to stand on stage next to POTUS and I said, today the Rust Belt ends. Today the Silicon Heartland begins. You could feel the national pride just welling up inside of you. And when we did that, you know, in addition, huge capital investments and everything else that goes into it, but we also announced a hundred million dollar program, 50 million from us, 50 million from NSF, specifically on workforce development, you know, for that manufacturing, you know, location that we were, you know, beginning to get underway. So it's, you know, it's deeply in the care abouts of our company. You know, what are we doing in training workforce development, you know, and skilling of the future. Something we deeply, deeply believed in, that's been one of the areas that Intel has always differentiated itself. Maybe I can add to that and ask you, look five years from now and imagine the workforce you truly need, because those are the people we need to start training right now. Who are they? How are they different? What is unique about them that the previous generations didn't have? Yeah. You know, it's one of the things that, I mean, we've touched on already, you know, clearly, I mean, they're going to view, you know, AI tools and skills, you know, like, like I would have used the spreadsheet right early in my, oh, of course I do that. Right. You know, let's go run an AI model against that. You know, here's my data sets. You know, that's just going to be like standard, you know, tools of the trade walking in. But, you know, another aspect that we desperately, and I think we will uniquely want is, you know, today in our factories, you know, if you look at the workforce man hours required to produce a thousand wafer starts a week, okay. I am not at the level of automation and efficiency as TSMC is today. So I need a class of workers that are going to be able to create the most efficient factories that, you know, with the highest automation, metrology, tool efficiency, you know, et cetera. And a lot of the AI tools are going to be useful for that as well, because my cost of labor is higher than Taiwan or Asia as well. So not only do I need to automate to get to the same number of hours per wafer, but if I'm going to actually be cost effective in the world, I have to be more because our cost of labor is higher as well. But hey, that's what America has always done. Right. You know, every hard, you know, problem in efficiency, we've come up with new automation and innovation solutions to solve it. So, you know, that's a class of things where if we're going to bring manufacturing back to the US, we can't just have it be, you know, equal jobs running, you know, in the US. We have to have better jobs using higher automation, more robotics, more, you know, innovation and AI efficiency associated. So another class of the kind of talent that we're going to need, and I need people who are, you know, like, you know, like loving in the nano, right, MIT location. Those are the kind of folks I need. Right. You know, people who would love to say, you know, that, you know, it isn't just some ether of software. No, I built something today. Let me show you what came out of our factory today. Let me show you my quarter billion dollar lithography machine, you know, and I've how I've innovated that, how I've printed the finest line that's ever been produced on planet Earth. How I've, you know, how I've, you know, cantilevered molecules into a, you know, structure that has never been produced before on Earth. That's what I want to do. Right. You know, this is like the woodshops of the past. Right. You know, this is the microelectronics woodshops of the future. Right. You know, and people that are passionate about hardware and doing things that fundamental to science, creating right structures and physics and leveraging the periodic table in ways that it's never been leveraged before. Those are the people I need. Well, we have a lot of those. So we very much look forward to indeed enticing them to join Intel. But if you look at the gap between what's happening in academia and what the industry truly needs, at sometimes it appears to be a very large gap between us where we can extrapolate and new science and rethink the way the transistors can be made with no semiconductors. We might be able to make a switch with a 12 volt sub-threshold slope, a 12 millivolts per decade sub-threshold slope, right? That is unbelievably good performing and of course will not be scaled because it is not fitting necessarily into the model of what has been scaled up to that point. So I'm wondering, is there a way to bridge that gap? Is there a way to bridge the gap between tremendous power and complexity? The most incredibly innovative work that humanity has done in the form of present chips and the new innovations. We are trying to redesign that old innovation. After all, we tell all our students their job is to change the world, not follow the yesterday's guide. One of the things I'm quite excited about coming out as part of the CHIPS Act was also NSTC, which is the National Semiconductor Technology Center that's also being formed as part of that. And the related, what I'll call the secure facility sidecar, defense innovation, so on. These fabs, truly they cost 20, 30 billion dollars. What's the entire budget of the nano center? 60 million tools, 10 million a year to run it. So 60 million versus 30 billion if you're going to build one of these modern super facilities. The scale, the cost, literally I have single pieces of lithography equipment that are a quarter billion dollars for one piece of equipment. And to run a fab, I need about a dozen of those to run a fab at scale. So with that, this gross economic gap between what you can do in academia, even with all the power of the defense and intelligence community, they can't afford their own fab. Can't be done at that level. So how do we bridge those two worlds forward? And I just call it the sidecar. And I want to take my super highway that I am building, and I want to create a research sidecar and a secure sidecar that meets all of the unique requirements of defense, intelligence, et cetera. So the secret handshakes and the special skiff facilities and all of those kind of things that you need to meet their needs, but also create this domain for research innovation as well. Because you cannot possibly afford to put these equipments into a research domain. So how do we create an environment where we can sort of work together? And with NSTC helped to form that halfway house, if you would, between these massive investments for the commercial thoroughfare and being able to open up pathways for fundamental innovative research work that we're able to do together. We run the university shuttle program that we started, which is allowing people to design. And it's on Intel 16, so a 16 nanometer facility. I'm anxious to get that to Intel 3, so a 3 nanometer process technology. In 2021, we ran about 60 shuttles. In 2022, we ran about 400 shuttles. And my goal would be about an order of magnitude each year so that every university, right, every university class and researcher can go on to a modern technology node for their research work, for their design work, for their skilling work as well. So that's the vision that I hope that we bring about. And NSTC and the CHIPS Act are key components of building, I'll say, the sidecars, right, and that scaling of the shuttle programs. That's fantastic. It does require us, though, to design for those specific tool sets, as opposed to using our present tools and trying to reinvent the next set of technologies just with them. So imagine we do reinvent it. What other materials, besides silicon, might find themselves being able to be used in this material, in this very advanced system? So a little, you know, silicon, right, is like God's gift for humanity, right? You know, this 4-4 structure, very stable, very manufacturable, and you can shove all sorts of other atoms and molecules into it, right? You can stress it, right? You can bend it, you can shove things into it, you know, and it's just amazing. And if you look at a silicon chip today, there's amazingly little silicon in it when you look at the transistors. I mean, you're using, you know, high-K metal gates, you're using different materials on it, right? You know, it's essentially the scaffolding that we're building microelectronics on top of, right? And that's why it's so amazing. And that's why I say we're not finished until we've exhausted the periodic table, right? You know, we're going to find ways to add materials and use the scaffolding that God gave us in the silicon, right, and just keep shoving more materials, you know, on top of it, in it, and bend it. And right now, there's like a lot of research going on on ferroelectrics, okay? How do we layer those on top of silicon? Because if I can make it somehow work on top of silicon, I got massive factories I can run it into. I got all these equipments, you know, I got etchers and depositors and lithographics and so on like that. So to some degree, the direct answer to your question is what kind of materials? All of them. Right, you know, because we want to just keep opening that envelope, right, of the periodic table of, you know, innovation and new molecules and discovering new effects. And then if I can put it on top of silicon, and by the way, you know, it's proven to be this extraordinary material, right, I can now make it at scale for the planet. You know, like we have our quantum program, you know, our quantum dot program, you know, ours is different than, you know, some are doing them, you know, with super cooling effects and some are doing them with lasers and so on like that. We're doing ours on top of silicon, right? And part of the reason I believe we will win in producing the first useful quantum computing is because you probably need, you know, let's estimate maybe a million qubit machine, right, to create a useful couple thousand qubits, logical programmable qubits that are stable, error corrected, etc. So I don't think any of the other approaches you're ever going to get close to being able to build a million qubits. I am confident I can build a million qubits because that's what we do, you know, with manufacturing at scale. So I believe we're going to win the quantum race as well. Maybe you don't. Maybe you love one of those other techniques. That's okay. Let the race be gone. But, you know, I think by about 2030, we will deliver, you know, sustainable quantum computing, right, and deliver quantum supremacy on meaningful algorithms in a substantive way. And I think that's the new, right, if you talked about the AI collision with HPC, you know, the next supercomputers will be quantum based and, you know, my thumb in the air is we'll hit that right around 2030, 2031. So I'll ask my last question, which is, I've completely subscribed to everything you just said. Very excited. What do I need to do to help you out? What does, what do universities need to do to assist Intel in reaching this vision? Well, you know, let me give the flip answer, you know, all of your students that are good need to come to Intel and work for us and all of your research projects should be aligned with Intel's agenda and everything, right? So right. Excellent. Along that line, what advice, what other advice do you have for our students? Yeah. So now being a little bit more serious in that, in that regard, you know, I do believe that we are in this period of time where we get to decide if we manufacture the future. I am confident that we will invent the future. The question in my mind is, will we manufacture the future? You know, are we going to be owning the supply chain of manufacturing and delivering this, you know, at scale? Are we going to, you know, are Western countries going to do that or are we going to allow that to drift away? And I think this is a moment in time, right? Where if we don't act decisively, you know, and that's why, you know, when I had the chance to be on the White House lawn for the Chips Act signing this last year, that was another one of my five great days, right? You know, at that point, because that was the most significant piece of industrial policy legislation in the United States since World War II, right? You know, and if you just think about that, you know, we ain't done industrial policy for 70 years. So one of the things as I would look at MIT is, are you going to help make the vision that the president signed, you know, that Congress passed, are you going to make that vision of not just inventing, but also manufacturing the future truer, right? And I would say, you know, and that to me, that is why I rejoined Intel as CEO. And that to me is a cause, a purpose, a vision worth us pursuing together. And obviously, you know, there's, you know, MIT attracts the best of the best. Maybe not as good as Stanford, but anyway, I mean, you know, incredible students, incredible minds, you know, and I would hope that every one of them gets into the, you know, into the nano lab, you know, and falls in love again with building hardware, building silicon at scale, because I want the digital future to be based on the things that we invent, that we manufacture, that we deliver to humanity at scale. And that's a very different trajectory than the one that we have been on as a nation for the last 30 years. You know, Congress never voted to get rid of the semiconductor industry, but the parliaments and congresses of China, Taiwan, and Korea voted to get this industry, right? We just voted to get it back. And I need a whole lot of help to make that happen, and I hope to have your participation in that. But we have minutes for questions from the audience, if you don't mind. Sure. There's a microphone here, if you could please. I think we have a question right over there, John. Yeah. Hey. I wanted to thank you for the University Shuttle Program that supported my entire PhD here at MIT, including about a half-dozen ships. We've now spun out into a startup company here in Cambridge, and the company is now deciding whether our foundry will be Intel or otherwise, and so... Is there a choice? Yeah. It's a really important choice and a difficult one, but, you know, you talked about the roadmap with University Shuttle and supporting universities, and I was hoping you could comment on, you know, your plans for the growing domestic shuttle community. Yeah. And, you know, obviously, my passion for the Shuttle Program is so that it becomes a passion for the Venture Program, right? So if you've done your research, right, on Intel Shuttle Program, it's just, you know, I mean, you just, you know, go two steps to the right, and boom, you're running on us as a foundry for a startup program, and then you go, right, as an early, you know, as an early user at scale. I mean, that is the pipeline that I'm off building. Now, hey, we are a early foundry provider, right, and I'll just say our technologies aren't as mature as they need to be yet, you know, I don't have as rich of a EDA, right, you know, IP, right, ecosystem as yet, but we are extraordinarily committed to build that out over the next several years together, and, hey, if we're not good enough yet, you know, help us get there. More questions. I have to stand up. Do you see a role for biological materials in your industry and in this manufacturing of the future? You know, hey, you know, I think there's been lots of crazy ideas for different computing structures over time, and, you know, lots of different memory structures, and, you know, and when you look at all of those approaches, very few of them have mattered at the end of the day, right, for it. And, you know, you've essentially ended up with three, four memory types. You've ended up with a couple of transistor types, you know, that have just dominated everything, right, you know, for it. So, you know, there's a lot of the different approaches that people have talked about, and it's great and wonderful research. I'm skeptical that any of them have the kind of market impact that your basic digital transistor has had, right, and silicon as the core material, you know, that we've built on, and, you know, the fundamentals of CMOS, NMP devices, you know, have for the future. It's just proven to be so extraordinarily scalable in that regard, and the kind of benefits that something like a biological computer has been able to demonstrate are comparably modest. Now, obviously, by the comments I, you know, gave as well, hey, quantum computing is an entirely different domain, right, in that sense, where, you know, you move past the digital domain, right, you know, you have, you know, such, you know, a radical different effect that you're leveraging, potentially a super interesting, so that's one that we are excited about. You know, we're also doing things like neuromorphic computing is another research area for us, but in all of those cases, I'm trying to do it on top of a silicon infrastructure so that I know how to scale it over time. We may not be right, but we're not in doubt. Okay, a couple more questions. Thank you very much for being here. Was Sematech the right concept? I'm sorry, what's? Sematech in Austin, Texas, and what happened to it? The, you know, Sematech is an interesting history, because most people would look at Sematech and call it a failure, right, at that point, but I would also argue that it catalyzed the industry, you know, when we were getting our heads handed to us in the U.S. to Japan, right, and fundamentally, we won, right, you know, and that, you know, resurgence of industry. So in some senses, I would call it a huge success, because it catalyzed the industry, even though that particular research and collaboration model didn't have the kind of effectiveness of flowing things into the industry as it was intended, right, and in some regards, it ended up almost on a model that it was competing with the industry that it was trying to serve. So, right, as we ponder NSTC, the National Semiconductor Technology Center, you know, I think the Sematech experiences need to be in our minds as we're shaping what this next phase of a structured research model needs to look like for the industry, because I will argue that one was not successful, or certainly not as successful as it would have hoped to be at the time. All right, final question. I believe in the old days, all the great VCs made their reputations investing in semiconductors, and nowadays, you ask a deep tech VC, hey, what are you doing, and they say blockchain. And they say, you know, would any of you invest in semiconductors? No, no, we would never invest in semiconductors anymore. So how do you get the venture community, if you think they're relevant, to share your brilliant vision? Yeah. And, you know, they're actually, you know, if you look at what venture has done, right, you know, with the, I'll say the maturity of the foundry ecosystem that has largely been around TSMC, right, and, you know, obviously, I hope to change that. But, you know, largely, there's actually been a somewhat resurgence, right, of venture investing around semiconductors over, you know, the last, you know, eight, nine years in that regard. And you're certainly seeing that, you know, results with some of the networking components, but now some of the AI chips that have emerged, some of the RF sensors, other things like that, you know, that have emerged. So I'll say it's a bit more, you know, but not a vibrant, you know, community around it. So I do think that certainly that the NSTC, if I go back to that again, you know, if you get, you know, good, solid research underway, you know, venture, you know, venture doesn't go do something, they respond to things, right. And I think if NSTC starts creating things, the venture community will respond to those things appropriately, right, you know, in that sense. And that's why, you know, I have viewed the chips and NSTC, ITC, what we're doing, I view it like this domino that starts a set of other dominoes underway, you know, and when I go, for instance, meet with the ASML head, you know, the lithography, EUV, equipment head, your next R&D location, going to be in Ohio, right, you know, okay, you know, we're ready to start moving that, you know, and, you know, your, where was my friend up here, your venture company is going to be on the Intel foundry, right, you know, and that will be, you know, a startup that gets funded, you know, you're going to be well funded, I am certain of it, right, right, you know, as you go for, you know, to me, it's this, you know, we have to start this period of momentum, and then momentum builds on itself going forward, so, and maybe I'm just crazy optimistic on this, but I think this domino has started, you know, since the CHIPS Act got underway, we've had five major manufacturing facilities announced in the U.S., Samsung, TSMC, Micron, and two by Intel that we've done, wow, you know, it isn't going to rebalance the world, but still, that's more manufacturing investment announcements in the U.S. in the last two years than was in the ten years prior, so I will claim, okay, the domino is starting to move forward, a long, long way to go, none of us should be confused, you know, I've said the moonshot is, is that we've gone from 8020 to 2080 to 5050 in a decade, and hey, if, even if we get to 6040 in a decade, that is extraordinary, you know, we've doubled the manufacturing capacity on the most critical ingredient, right, of the economy, and the digital future, and every human on the planet, right, man, what an extraordinary result, so that would be, you know, if we're able to even get to that level over this period of time, and I think we can. Well, thank you, Pat, for coming to share with us your optimistic outlook for the semiconductor industry, thank you again, please. Thank you so much, so inspired, so inspired.