Intel CEO Pat Gelsinger! (CNET's FULL Interview)

It's been a difficult half decade or so for Intel. When you were looking at taking this Intel CEO job, what convinced you to go for it? What told you that this is a turnaround that you can accomplish? I was drawn to the legacy of the founders. You know, Grove, Moore, Noyce. You know, I sort of joke that I joined Intel so young I went through puberty here. And, you know, I mean, I was trained at the feet of the founders and it's largely in their legacy. You know, hey, I want to restore this icon. Secondly, it's critical for the industry. You know, this is the company that put silicon in Silicon Valley. Right? You know, I just say, you know, a healthy Intel means a healthy industry. But also, you know, this idea that, boy, in a global semiconductor shortage, you know, rebalancing a more resilient, globally balanced supply chain. This is critical to the world, to our nation, to our economy, to our national security. So, you know, it's very much sort of driven by those three perspectives. And, you know, as I was interviewing for the job, I wrote a strategy paper for the board. I said, this is what we should do. And when they called me for the assignment, I demanded unanimity. Right? So every one of them had to say they were hiring me and agreeing with the strategy. Right? You know, so part of the reason we've been able to move rapidly is I already had the board support before I agreed to the job. And it was really a strategic alignment with board was already done. And with that, I had increased confidence that we could go and execute rapidly and successfully. What were the manufacturing and business aspects that you needed to reassure yourself were solid enough? I always felt like there was enough technology littered inside of the company. Right? You know, I mean, just stumble over fellows and breakthroughs and innovate, you know, just all over this place. Clearly, some of that was wandering away, you know, having lost the vision a little bit. But I felt pretty confident that there was plenty of things to build on. You know, I was concerned about how we get the process roadmap back in shape. We underinvested in capital. You know, I went to the board and said, we're done with buybacks. We're investing in factories. And, you know, that's going to be our, you know, the use of our cash as we go forward. And they aggressively supported that perspective that we needed to just start investing. And those investments would start creating a cycle of momentum that would get our factory teams executing better. If I'm going to do a process technology every year, right, which is what we've laid out to do, you know, Steven, well, I needed to give them twice the wafers for the TD experiments because we got to go twice as fast. The support for those gave me the confidence that, yeah, we can get this machine back to where it was before. There's a credibility issue. You've talked a lot about coming back to the Andy Grove ethos where you make a promise, you keep that commitment. There have been some big stumbles here in the last few years. So why is it that you're comfortable making those kinds of commitments about a very long, very ambitious roadmap? Right. We're measuring these things essentially every week. I'm getting updates on these different aspects. You know, I felt we had the talent, we had the, you know, capabilities in these areas, but they had to be re-fired up. We had to restore some of the Grovian aspects of tough, direct, data-driven conversations. And we're doing it. You know, we've laid out product milestones of, you know, how we're going to execute and deliver upon them. And we're now opening up our fabs and our foundry strategy to scrutiny by very tough engineering teams who are making these decisions. You know, trust me, the Qualcomm team, they know what a foundry looks like. We have a lot of work to do, right? You know, a decade of bad decisions. You know, this doesn't get fixed overnight. But, you know, the bottom is behind us and the slope is starting to feel increasingly strong. When you're talking about that foundry business, I've run into some skepticism from some players. I've talked to some people who make their chips at TSMC in Taiwan, and it hadn't really sort of even occurred to them to even give you guys a call. So, I guess, why is it different this time? You've stumbled with the foundry business in the past. What's going to be different this time when you're talking about manufacturing chips for other companies? We are taking this extremely serious. This is a core strategy of the company. Now, before it was sort of like, oh, you know, go try to sell a few wafers to somebody else. Now, this is central. You know, we're not selling you last year's technology. I mean, we're putting customers like Qualcomm on the leading edge of what we do and, you know, putting our best packaging and process technologies forward. I've set it up as an independent business unit, right? And that unit is going to be measured on one thing, right? The success of our foundry. Arguably, one of the problems going from 14 nanometer to 10 nanometer was you bit off more than you could chew. You tried to make too many advancements in one step. And I look at the roadmap in the future for you guys, and it looks potentially considerably more ambitious when you go to Intel 7, Intel 4, Intel 3, 20A, 18A. You're talking about rapid improvements to all your manufacturing processes, you know, after a time when you really struggled. So can you really deliver the goods here? Some of that will be, hmm, I got to prove it to you. And a little bit of skepticism is appropriate. But we've also been very, very measured in how we're taking risk. And, you know, as we think about that cadence that we've laid out, 7, 4, 3, 20A, 18A, and beyond. And, you know, as an example to that, Tyche, the backside power, as we've called it, Power Via, for that. Hey, we're going to put that on a variant of Intel 3 before we introduce it into 20A. You know, we're going to get a lot of that backend processing done, you know, debugged and manufacturing volume so that we've sequenced the risk profile for every one of those process nodes. So each one of them, you know, I've sort of doubled my capacity for innovation, right? We've given them a lot more equipment, a lot more wafer starts. So essentially, we can be running 2x the number of experiments, short loops, more TD engineers that are running this. So all of those things, we're building confidence. And, you know, 7, 4, 3, 20A and 18A, they're all looking good. Our yields on 10 and 14 are looking very good. The factories are ramping and, you know, shovels are going into the ground ahead of schedule in Arizona. It seems like it's a lot more work to improve each new technology manufacturing node. It's upstream work. And I wonder, as that gets harder and harder with each new step, progress slows, what new skills actually become more important that weren't before? Clearly, you know, designing transistors, still very important, but what new skills come into play? Some of the new skill areas, as well, as we keep innovating, I'll call it the 3D designs. Because as you look at our advanced packaging now, you know, I can cut and slice the chips, the tiles, as we call them, in almost any way that you like. I can stack them. And almost any way that you like, right? We can create different composites of, you know, which process technologies we use, where, where we put memory, you know, and it will be a whole new class of system design in the package. And I call it, you know, it's system on package now. And with the power of our advanced packaging technologies, you know, it's a whole new domain of optimization and innovation. There's been a big chip shortage, obviously. It's been problems. People can't buy the cars they want. They can't buy the PCs they want. How has that changed your business at Intel? Every day, we're getting inbound calls because our customers are struggling with the shortage. And, you know, part of our incredible zeal that we have to accelerate our manufacturing build out and do more of that here in U.S. and in Europe is exactly this reason. You know, we just got to build capacity quicker. We have to ramp it faster. We have to get better yields more rapidly. We have to get what we call match sets. We have to work with our industry partners because if I give you a laptop CPU, but you don't have an LCD to go with it, you still can't ship the laptop, right? We got to work with the industry more closely. So it's driving us to re-imagine and innovate our supply chain and the transparency of our customers and their suppliers' supply chain with our own. So causing a lot of shift. We're looking at the potential $52 billion coming in U.S. government subsidies. How is that going to move the needle for U.S. chip manufacturing? A huge amount of the electronics industry is already overseas. Will this actually make a significant difference? The expectation that I've said is that over the decade in front of us that we should be striving to bring the U.S. to 30% of worldwide semiconductor manufacturing. Today, it's 12%. In 1990, it was 37%. Europe in 1990 was 44%. Today, it's 9%. And the European moonshot, as I've been meeting with the European officials, is to get them to 20%. So 30%, 20%, 50% in the decade. Wow. That's a much more balanced and globally resilient supply chain. Now, obviously, and as your question suggests, there's a lot of other aspects to the supply chain. And I believe those need to be more balanced as well. System assembly, package tests, other elements of the supply chain. And again, it's not that we're moving out of Asia. What we're saying is let's build up capabilities so that economic balance, that we have dependence. No, we have interdependence, where each of these regions can operate more independently. And we've built more natural resilience so that we're not as susceptible to individual locations, national disasters. We get to decide where the fabs are, right? And we should take advantage and put them where we want them. You're talking about two huge new greenfield sites, one in the U.S. somewhere and one in Europe. How much are those going to cost and when are they going to arrive? My objective on both of those is to announce those new greenfield locations before the end of the year. We would expect as we launch both of those, that they would be what we call mega fab locations. That we'd be building them such that we could put up to eight fabs in each of those locations. Each fab, ten-ish billion dollars, so you can quickly do the math. And obviously we're anxious for support from U.S. governments and European as well to make sure that they're economically competitive with the Asian alternatives. You are investing a fair amount right now in Arizona, as you mentioned, the 20 billion dollars on Fab 5 and Fab 6 there. Is that going to help increase the capacity in time for this chip shortage? You know, these new fabs that we're starting now, it takes about three to three and a half years to bring them on production. Clearly we're hopeful that, and I've indicated that, I think that this crisis peaks about now. Right, the worst of it is Q3, Q4. Each quarter next year gets incrementally better, right, as we go through it and, you know, we can expect to get some level of supply-demand balance in 23. We're also making investments in other areas that will have more immediate feedback. And we've been able to pull things that, you know, improve our supply. Your investments look big until you look at the scale of TSMC's investments. And I wonder, do you need to spend even more to keep up with them? There's the club, if I would call it, where you need to be, you know, able to be beyond 10 nanometers, you know, where you're investing well over 10 billion per year. And clearly Intel's in that club at that level. And then it's really driven by, are you on the leading edge? Right, which we've already talked about, boy, you know, 7, 4, 3, 20A, 18A. So we're comfortable that we're on the leading edge. And we're scaling our investments to meet the increasing demands of the industry, right. Some of it is to catch up, you know, we should have made some of these investments earlier. And then third is to get ahead, right, where, you know, launching our new businesses, expanding, supporting the foundry build-out. What is your forecast for continued progress? It seems like each step is harder and harder and delivers less and less. Are we just going to see sort of a gradual, slow progression to, you know, no more progress? We've definitely seen a bit of the slowing. You know, Moore's Law used to be a doubling every two years. And we've definitely seen a two-and-a-half to three-year kind of shift, Stephen. The pace that we're now putting Intel on, I think we're actually pushing back, you know, closing that gap. We see a pretty clean path on lithography. With our new RibbonFET technology, we see a transistor structure, you know, that we believe is going to be very healthy through the end of the decade. So we're actually seeing ourself that we've overcome some significant issues in Moore's Law. And I actually see a pretty healthy cycle in front of us. But third is this move to 3D, right, with advanced packaging technologies where I can redefine the partitioning of chips, I can stack them together. So I believe that you're going to see from 2025 to 2035 a very healthy period for Moore's Law-like behavior.