Former Intel CEO: Why This is the Best Time to Build Hardware

1. 0:00 – 1:00

   Intro

   1. PGPat Gelsinger0:00

      In a AI digital age, energy capacity equals economic capacity. Why build the new data center and buy the million GPUs if I can't power them? You're gonna see more and more defaults happening on many of those data center projects because the energy won't be there.

   2. GAGuido Appenzeller0:16

      Whenever you have the technology to make something easy, that means the bottleneck moves somewhere else.

   3. PGPat Gelsinger0:21

      Nothing's a chip anymore, it's a rack. It took me three months to design it, but it's nine months until I can actually start to use it. Exactly how many major new memories have we had over the last 30 years?

   4. GAGuido Appenzeller0:31

      Zero.

   5. RRRaghu Raghuram0:31

      Zero.

   6. PGPat Gelsinger0:32

      Memory innovation for the first time in 30 years is nigh upon us. I declared the death of copper about 25 years ago. Eventually I'll be right for all of us hardware guys. This, this is like renaissance in front of us.

   7. RRRaghu Raghuram0:43

      [laughs]

   8. PGPat Gelsinger0:44

      AI inference accelerator chips-

   9. RRRaghu Raghuram0:46

      Yeah

   10. PGPat Gelsinger0:46

       ... and I'm sure I don't even know them all. Why are you guys funding so many of those?

   11. RRRaghu Raghuram0:49

       You funded your fair share too.

   12. PGPat Gelsinger0:51

       [laughs]

   13. GAGuido Appenzeller0:52

       Historically, there have not been 100 competing processor vendors in any industry ever.

   14. PGPat Gelsinger0:56

       Yeah.

   15. GAGuido Appenzeller0:56

       Is this a temporary thing?

   16. PGPat Gelsinger0:57

       Yeah.

   17. GAGuido Appenzeller0:57

       It'll converge back to a few?

   18. PGPat Gelsinger0:59

       I see

2. 1:00 – 5:44

   From tech school to Intel at 18

   1. PGPat Gelsinger1:00

      it as a-

   2. RRRaghu Raghuram1:02

      I'm here with, uh, our esteemed guest and dear friend and former boss, Mr. Pat Gelsinger. Welcome, Pat. Pat is currently the general partner at, uh, at, uh, Playground Global. But, uh, as, uh, you're very, very well known in the industry for leading Intel, for being the CTO of Intel, and of course, leading VMware and many other things. So welcome.

   3. PGPat Gelsinger1:28

      Hey, thank you Raghu. Great to be with you and Guido. Right, and to me this feels just a little bit like old home, right?

   4. RRRaghu Raghuram1:34

      Yes. Yes.

   5. PGPat Gelsinger1:34

      You know, it's like, you know... You know, it's like, uh, we superimposed a year of change, right? But, uh, you both look the same. I feel, uh, energetic, so let's dive in.

   6. RRRaghu Raghuram1:43

      There has never been a day when you stop being energetic-

   7. PGPat Gelsinger1:45

      [laughs]

   8. RRRaghu Raghuram1:45

      ... so that's, uh, no news there. Uh, but yeah, no, let's start actually from, uh, from your Intel career, right? I mean, recently, Andreessen Horowitz, uh, int- created this Horowitz Andreessen Academy-

   9. PGPat Gelsinger2:00

      Mm-hmm

   10. RRRaghu Raghuram2:00

       ... which was to find talented people between 16 and 22, and then put them into a, a modern educational setting so they'd be well-prepared for either doing things on their own or joining the companies of today. You had a similar experience. Like, you went to a regular trade school and then jumped to Intel-

   11. PGPat Gelsinger2:21

       Yeah, yeah

   12. RRRaghu Raghuram2:21

       ... when you were what, 18 or 19 or something?

   13. PGPat Gelsinger2:23

       Yeah. Yeah. Uh, you know, it was really a sort of magical, uh, period. You know, I'm 16 years old. I accidentally win a scholarship. I go to tech school, right? Skipped my last year and a half of high school, and Intel comes recruiting when I'm 18 years old.

   14. RRRaghu Raghuram2:37

       [laughs]

   15. PGPat Gelsinger2:37

       And, uh, you know, I'd never been on an airplane. Uh, had already fallen in love with computers at tech school. And, uh, the interviewer, right, Ron Smith was his name, he writes on his page, and I was number 12 that he interviewed-

   16. RRRaghu Raghuram2:50

       Uh-huh

   17. PGPat Gelsinger2:50

       ... uh, that day. And if you've interviewed 12 people in a row, you can't tell male from female-

   18. RRRaghu Raghuram2:54

       Yeah

   19. PGPat Gelsinger2:54

       ... you know, by the end of it, right? You know, I'm number 12. He says, "Smart, aggressive, arrogant. He'll fit right in."

   20. RRRaghu Raghuram3:00

       [laughs]

   21. PGPat Gelsinger3:01

       So I got invited to-

   22. RRRaghu Raghuram3:02

       There you go

   23. PGPat Gelsinger3:02

       ... uh, you know, come to, uh, Intel and, you know, it just became, uh, you know, really, uh, uh, glorious. Uh, you know, starting as a technician, moved into the design team at the end of the 286. You know, engineer number four on the 386. You know, uh, architect and design manager for the 486. All of that while doing my master, my bachelor's, uh, my master's and PhD work. So it was like-

   24. RRRaghu Raghuram3:27

       Ah

   25. PGPat Gelsinger3:27

       ... the, the, the best career that you could possibly have. You're learning by, you know, day and you're putting it to practice at night, and-

   26. RRRaghu Raghuram3:35

       That's right. That's right.

   27. PGPat Gelsinger3:36

       Yeah.

   28. RRRaghu Raghuram3:36

       And you're creating new, completely new era of chips for... I bet most of what you learned and what you built, there was a big gap-

   29. PGPat Gelsinger3:45

       [laughs]

   30. RRRaghu Raghuram3:45

       ... because you're breaking new ground.
3. 5:44 – 7:23

   The 486 and the birth of modern EDA

   1. RRRaghu Raghuram5:44

      [laughs]

   2. GAGuido Appenzeller5:44

      I think it's pretty, yeah.

   3. RRRaghu Raghuram5:46

      Yeah. No, uh, 486 was your big-

   4. PGPat Gelsinger5:50

      Yeah

   5. RRRaghu Raghuram5:50

      ... accomplishment, if you will. Obviously, lots of people worked on it. Um, and, uh, you guys broke some new ground in chip design there, right?

   6. PGPat Gelsinger5:59

      Mm-hmm. Mm-hmm.

   7. RRRaghu Raghuram6:00

      And, um- Increase what, uh, functions the chip was supposed to do and so on and so forth.

   8. PGPat Gelsinger6:05

      Yeah, that was a pretty magic period, uh, as well because it was before what you would think of as the EDA industry.

   9. RRRaghu Raghuram6:12

      Oh, wow. Yes, that's right.

   10. PGPat Gelsinger6:13

       Right. And, uh, you know, the, the end of the, a little bit of the 386, but the 486 was really the first chip to implement what, what you would think of as modern EDA techniques. You know, we did a high-level design, right, a description, an RTL description-

   11. RRRaghu Raghuram6:31

       Yeah

   12. PGPat Gelsinger6:31

       ... but there was no Verilog.

   13. RRRaghu Raghuram6:33

       Mm-hmm.

   14. PGPat Gelsinger6:33

       So we invented HDL, right, the Intel hardware description language.

   15. RRRaghu Raghuram6:38

       Oh, wow.

   16. PGPat Gelsinger6:38

       So, you know, I wrote my own language. Well, you had to build a compiler for that language, right? You know, so we created a compiler for the language. There was no automatic place and route, you know, so we had to invent that, and we worked with, uh, Alberto Sangiovanni-Vincentelli at Berkeley, and some of his students for the first, you know, placement, the first routing, the first automated-

   17. RRRaghu Raghuram6:58

       Yeah, yeah

   18. PGPat Gelsinger6:58

       ... timing, you know, management. And-

   19. GAGuido Appenzeller7:00

       So, so they created EDA to some degree, yeah.

   20. PGPat Gelsinger7:02

       Yeah.

   21. RRRaghu Raghuram7:02

       Yeah.

   22. PGPat Gelsinger7:02

       You know, to a, to a great degree that, uh, you know, and that was really one of the hallmarks, that not only was the 486, you know, this compatible, you know, pipeline microprocessor, but we ushered in, you know, many of the foundations of what became the modern EDA, uh, industry. And, you know, it was really, you know, a pretty magical, uh, period of the industry.

   23. RRRaghu Raghuram7:22

       Yeah, yeah. No.

4. 7:23 – 9:18

   How AI changes chip design

   1. GAGuido Appenzeller7:23

      So, so if you look at today-

   2. RRRaghu Raghuram7:25

      Yeah

   3. GAGuido Appenzeller7:25

      ... will people look back as Blackwell and saying, "You know, this was of the last generation that was, was done before AI took over the design process for, for micro..." Are we, are we in a similar transition right now?

   4. PGPat Gelsinger7:37

      I think there are certainly aspects of that. And, uh, you know, I think, you know, w- when you look at, you know, things like Jalapeno today, right, you sort of say, you know, the, you know, that's, that's sort of a first principle's use of AI-

   5. RRRaghu Raghuram7:49

      Mm-hmm

   6. PGPat Gelsinger7:50

      ... in the chip design process, right, where you throw out a lot of those things. Now, you know, of, you know, there are pieces of the design that are still really hard, right, uh, in that sense. And a lot of the analog, right, aspect, SerDes is probably the best example-

   7. RRRaghu Raghuram8:04

      Yeah

   8. PGPat Gelsinger8:05

      ... of that, you know, are still not AI-able, right? You just need lots of silicon data, right, to get those. So you either get so conservative in your analog design, you know, that you're able to, I'll say, AI it, uh, or right, you gotta do those the hard way, uh, you know, for it. But to, now many of the logic functions can really be done with, uh, AI tools-

   9. RRRaghu Raghuram8:27

      Mm-hmm, yep

   10. PGPat Gelsinger8:27

       ... and techniques in pretty incredible ways, right? Your transistor budgets are big enough. The design tools are getting smart enough. You know, a few experts guiding the tools and how to apply it, and I really think it will be somewhat like the 486 that way, right? We can sort of look back and say, "Yep, that was the beginning of a new era of chip design."

   11. GAGuido Appenzeller8:46

       W- whenever you have the technology to make something easy, that means the bottleneck moves somewhere else.

   12. PGPat Gelsinger8:51

       Yep.

   13. RRRaghu Raghuram8:52

       Yeah.

   14. GAGuido Appenzeller8:52

       Can we guess at this point where the bottleneck will be in the future?

   15. RRRaghu Raghuram8:56

       Well-

   16. GAGuido Appenzeller8:56

       You know, there, there's some things which it seems like AI is incredibly good at, right?

   17. RRRaghu Raghuram8:59

       Mm-hmm.

   18. GAGuido Appenzeller8:59

       Like creating the software layers-

   19. RRRaghu Raghuram9:00

       Mm-hmm

   20. GAGuido Appenzeller9:00

       ... you know, pr- writing kernels. Um, you know, a lot of the... I, I think a lot of the sort of being able to specify the, the, uh, the objectives at a higher level, which then gets sort of translated into Verilog, right?

   21. PGPat Gelsinger9:13

       Mm-hmm.

   22. GAGuido Appenzeller9:13

       And so what, what, what is the new frontier? What is the thing that, that will be difficult going forward?

5. 9:18 – 14:49

   Why silicon still takes nine months

   1. PGPat Gelsinger9:18

      Well, uh, you know, when you look at today, you know, so let's say, you know, Guido and Pat, we're off to do a great new chip together. We're gonna-

   2. GAGuido Appenzeller9:24

      We're building the X accelerator.

   3. PGPat Gelsinger9:25

      Right. You know, right. You know, and it's the one that's gonna leap ahead because we have understanding of the-

   4. RRRaghu Raghuram9:30

      This is the 2086 instead of the 486. [laughs]

   5. PGPat Gelsinger9:32

      [laughs] You know, we have understanding of the AI workloads. We have understanding then of how to compose that, and then the right multiply, accumulate, you know, register structures.

   6. GAGuido Appenzeller9:41

      The, the perfect visibility of the model workloads of tomorrow as well. [laughs]

   7. PGPat Gelsinger9:43

      Yeah. You know, and all of that kind of stuff. And, you know, let's say, you know, we turn our AI, you know, tools loose. We are great. Within three months, we have an awesome design, right? You know, you know, f- for that-

   8. RRRaghu Raghuram9:56

      Yeah

   9. PGPat Gelsinger9:56

      ... we've been a little bit conservative on all the analog components of it. You know, now we still have, like, a couple of really major bottlenecks today. You know, one is, right, we still have nine months of silicon processing time.

   10. RRRaghu Raghuram10:09

       Mm-hmm. Yeah.

   11. PGPat Gelsinger10:10

       Right? You know, something like that, right? So I can design the thing in three months, but I can't actually get it into real silicon at scale for nine months. Okay, that sucks, right? So we have to really see that bottleneck improve if we're gonna have this innovation because we know-

   12. RRRaghu Raghuram10:23

       Are you talking about the va- validation and verification stages-

   13. PGPat Gelsinger10:26

       Well-

   14. RRRaghu Raghuram10:26

       ... or even beyond that?

   15. PGPat Gelsinger10:26

       Well, you know, first, I can't get anything out of fab-

   16. RRRaghu Raghuram10:29

       Yeah

   17. PGPat Gelsinger10:29

       ... you know, in less than three months.

   18. RRRaghu Raghuram10:31

       Yeah.

   19. PGPat Gelsinger10:31

       And that's if I have, you know, like, supercharged design flows, right? So let's say that, you know, you know, how do I compress that, right, uh, you know, for it? And then I gotta get into advanced packages. 3D packages are getting complex, so that's like another three months.

   20. GAGuido Appenzeller10:45

       Very complex.

   21. PGPat Gelsinger10:45

       Right? You know, and then I have to put it into a rack scale solution 'cause nothing's a chip anymore. It's a rack.

   22. RRRaghu Raghuram10:50

       Yeah.

   23. PGPat Gelsinger10:50

       Yeah, so we're nine months. You know, it took me three months to design it, but it's nine months until I can actually start to use it. You know, so how do we start to compress, you know, those aspects of design? And I think we need new forms of lithography, you know, to enable that. We need, you know, design flows that don't require, you know, $50 million of mask cost until I can get things into prototyping. So, you know, to me, how do you compress that to a month or two, right? 'Cause I can now do the design in three months. How can I have that done in a month or two? Because if it takes me a year and a half until I actually get scale and software on it, okay, my understanding-

   24. RRRaghu Raghuram11:26

       The world is yours

   25. PGPat Gelsinger11:27

       ... of the AI workloads is no longer applicable-

   26. RRRaghu Raghuram11:29

       Yeah

   27. PGPat Gelsinger11:29

       ... to the chip that I designed, right? You know, so, you know, all of those-

   28. GAGuido Appenzeller11:33

       We seem to play out live right now-

   29. PGPat Gelsinger11:35

       Yeah, right

   30. GAGuido Appenzeller11:35

       ... I think with the AI chips, yeah.
6. 14:49 – 22:08

   Will 100 AI chips converge to a few?

   1. GAGuido Appenzeller14:49

      back to a few?

   2. PGPat Gelsinger14:50

      Yeah. And, you know, I see it as a more temporary thing, and it will converge, is, is what I expect. And there's probably three different reasons that I see that to be the, uh, case. You know, one of them is, you know, this emerging heterogeneity that you see in the compute, uh, environment, where, okay, now I have specialized prefill versus decode. Well, you know, and now people are saying, "Well, we really need a specialized tier for mid-fill, not just pre-fill," right? You know, so now I have early pre-fill, and now I have mid-fill, which is, you know, different and-

   3. GAGuido Appenzeller15:22

      Well, maybe speculate and verify-

   4. RRRaghu Raghuram15:24

      Yeah

   5. GAGuido Appenzeller15:24

      ... if we, uh, listen to OpenAI here, right?

   6. PGPat Gelsinger15:26

      Yeah. You know, and, you know, all of a sudden my compute fleet becomes more and more granular across the workload, and I think any time that you've seen that in history is not sustainable. And then all of a sudden, right, when, you know, as, as you look at that and now people are saying, "Well, as I go to reasoning models, I want something that looks more like a CPU again."

   7. RRRaghu Raghuram15:46

      Yeah.

   8. PGPat Gelsinger15:47

      So, you know, all of a sudden I didn't want that level. You know, that didn't become the predominant portion of my compute capacity. I needed more of it over here, right, in model and model workflows and, you know, so on. You know, so I really don't like, I'll say, at-scale specialization, right? Because I think the workloads are gonna continue-

   9. RRRaghu Raghuram16:08

      Yeah

   10. PGPat Gelsinger16:08

       ... to moderate, you know, and migrate so significantly, first. You know, second is, I think today's models are about to go through some evolutionary breakthroughs as well.

   11. RRRaghu Raghuram16:17

       Yes.

   12. PGPat Gelsinger16:17

       Right? You know, I think the LLM is sort of reaching the limits, and now as people look to how do I really model 3D, right? You know, where, you know, a flat-ish LLM doesn't work well when you go to m- molecules and chemicals and, you know, right, imaging w- you know. So I think we'll see limits of that that'll cause different shifts in the algorithmic domain. I also think that, uh, some of the most interesting workloads become where you start, you know, I'll say, bringing HPC-like things back into AI-like things, where all of a sudden, you know, 64-bit provision, uh, precision counts again, right? So I see-

   13. RRRaghu Raghuram16:52

       What would be an example of that?

   14. PGPat Gelsinger16:54

       Well, you know, imagine that my AI model is now inducing the five most interesting domains for a chemical analysis, right? Okay, those chemical algorithms-

   15. RRRaghu Raghuram17:04

       Of course

   16. PGPat Gelsinger17:04

       ... are gonna run with high precision, right? You know, when, uh, you know, which is gonna be the dominant piece of the workload, right? Getting to the five algorithms I need to run or running the five algorithms, right, on different chemical or biological systems. So, you know, I do think that, uh, you know, or as we optimize our AI systems, they will lead us back to some of the things that have been more traditional high-performance computing. So I see that aspect of workload, uh, as well. So, you know, all of that said is, you know, this extreme specialization to me in the compute architect, I s- I sort of don't like it because I don't think it's gonna be what the workloads will look like-

   17. RRRaghu Raghuram17:42

       Mm-hmm

   18. PGPat Gelsinger17:42

       ... two, three, four years from now. So that's one reason I don't see all of these AI chips as they get more and more specialized for-

   19. RRRaghu Raghuram17:50

       Yeah

   20. PGPat Gelsinger17:50

       ... portions of the compute workload to necessarily be right. You know, s- second is, right, there's like 100 of them now. Right? You know, there's-

   21. RRRaghu Raghuram17:57

       Yeah

   22. PGPat Gelsinger17:57

       ... multiple optical ones. There's probabilistic ones and so on. And you know they're not all gonna win, and they're not all gonna win because at the end of the day, you have to get scale-

   23. RRRaghu Raghuram18:06

       Yeah

   24. PGPat Gelsinger18:06

       ... on these things. And scale requires, you know, you have to win, you have to get capital, you have to get workloads onto it. So I think it sort of defies logic that you're gonna have 100 of these things. So I see them, you know, coming back, uh, as a result of, you know, simple, you know, capital market share, et cetera. You know, which are the winning teams, you know, winning designs, winning architecture. So, you know, I see the workload driving that. You know, I see the natural, you know, industry consolidations, you know, that will occur. And then I also expect that the winners will pick some winners, right? You know, where-

   25. RRRaghu Raghuram18:40

       Yeah

   26. PGPat Gelsinger18:40

       ... you know, an OpenAI, an Nvidia, an Anthropic will say, "I like this one," because it isn't just hardware, it's how do the hardware and software co-evolve.

   27. RRRaghu Raghuram18:51

       Come together. Yeah.

   28. PGPat Gelsinger18:53

       Right? You know, for it. And I do expect that there's gonna be, "Hmm, you know, here's this 10 I could pick for this. I like that one." And it does require investment in the software and the workload evolution to take advantage of those platforms. So for those three reasons, m- yeah, we're gonna see, you know, a narrowing of the field, uh, in that sense.

   29. RRRaghu Raghuram19:11

       That's-

   30. PGPat Gelsinger19:12

       Bas-
7. 22:08 – 25:15

   Why HBM is a hideous memory

   1. PGPat Gelsinger22:08

      scale.

   2. RRRaghu Raghuram22:08

      Earlier you talked about HPM being hideous memory.

   3. PGPat Gelsinger22:11

      Mm-hmm.

   4. RRRaghu Raghuram22:12

      You know-

   5. PGPat Gelsinger22:12

      By far the best one we have today.

   6. RRRaghu Raghuram22:14

      Exactly.

   7. PGPat Gelsinger22:14

      But-

   8. RRRaghu Raghuram22:14

      It's like the dumb thing about democracy.

   9. PGPat Gelsinger22:16

      [laughs]

   10. RRRaghu Raghuram22:17

       Um, I really can't recall when there was the last memory innovation. It might have been Optane. I think you probably had something to do with it.

   11. PGPat Gelsinger22:24

       Yeah, that one died, so.

   12. RRRaghu Raghuram22:25

       That one died. [laughs]

   13. PGPat Gelsinger22:27

       We killed it. [laughs]

   14. RRRaghu Raghuram22:28

       Yeah, we killed it, we killed it. So do you think memory innovation is around the corner? Or what needs to happen for... Obviously, we understand the scale problem-

   15. PGPat Gelsinger22:39

       Mm-hmm

   16. RRRaghu Raghuram22:39

       ... because it's even more-

   17. PGPat Gelsinger22:41

       Yeah

   18. RRRaghu Raghuram22:41

       ... larger there.

   19. PGPat Gelsinger22:42

       Yeah.

   20. RRRaghu Raghuram22:43

       But if you keep the scale aside, even from a technology point of view.

   21. PGPat Gelsinger22:46

       Yeah. And let's, you know, uh, you know, look at that history. I've probably personally been associated with at least five different new memory architectures, Optane just being one of them, right, you know, that did not see the light of day, right, uh, as well. Uh, I'm probably familiar with close to 100 that have happened in the industry over the last 30 years. And exactly how many major new memories have we had over the last 30 years?

   22. RRRaghu Raghuram23:13

       Zero.

   23. GAGuido Appenzeller23:13

       Zero.

   24. PGPat Gelsinger23:14

       Right? You know, which, you know, okay, DRAM, SRAM, Flash. Okay. [laughs] What else is there? DRAM, SRAM, Flash. You know, so I, I do think that it's-

   25. RRRaghu Raghuram23:21

       Flash stacking counts for a little bit.

   26. PGPat Gelsinger23:23

       Barely.

   27. RRRaghu Raghuram23:24

       Yeah.

   28. PGPat Gelsinger23:24

       Right? You know. But anyway, right.

   29. RRRaghu Raghuram23:26

       Yeah.

   30. PGPat Gelsinger23:26

       So, uh, you know, in that it, it, you know, it has been, right, you know, a, a disappointing-
8. 25:15 – 42:49

   How tall can chips get?

   1. PGPat Gelsinger25:15

      techniques.

   2. GAGuido Appenzeller25:15

      The, the future is stacked, yeah.

   3. PGPat Gelsinger25:17

      Somehow you gotta bring memory and-

   4. GAGuido Appenzeller25:19

      In more ways than one. [laughs]

   5. PGPat Gelsinger25:21

      You know, you gotta bring memory and compute together.

   6. GAGuido Appenzeller25:23

      Yeah.

   7. PGPat Gelsinger25:24

      Right? You know, in, uh, uh, fundamental, uh, new structures. Um, I do think that, uh, given the workload now has such capital being deployed against it, you know, in the AI space, that I do think we'll actually be able to break through some of those physics challenges as well. Uh, as well. So I do believe, and you know, I've just funded-

   8. GAGuido Appenzeller25:44

      New substrates and, yeah.

   9. PGPat Gelsinger25:45

      Yeah. Uh, you know, I've just funded a new memory company as well that's still in stealth, but, uh-

   10. GAGuido Appenzeller25:50

       Oh, amazing

   11. PGPat Gelsinger25:50

       ... you know, I, you know, you know, I'm quite excited about some of the innovations that will occur in new memories, and you're gonna see new materials, uh, you know, for memory pe- you know, a lot of people looking at things like ferroelectrics.

   12. GAGuido Appenzeller26:02

       Yep.

   13. PGPat Gelsinger26:02

       Uh, you know, finding non-capacitive, you know, uh, high-density memories, finding memories that can, you know, be high performance, high density, and stackable, right? But, you know, thing, you know, and people are looking at, uh, low-cost, uh, flash, you know, for that. I'm not particularly enthusiastic about that idea myself. I think, you know, speed and performance of the underlying cell are somewhat problematic with a flash or an MRAM structure.

   14. GAGuido Appenzeller26:28

       Yeah.

   15. PGPat Gelsinger26:28

       You know, but I do think there's memory innovations are coming for the first time in 30 years, and I do think that's gonna be an exciting space and, you know, we'll have a couple of companies there, I'm sure.

   16. GAGuido Appenzeller26:38

       Yeah.

   17. PGPat Gelsinger26:39

       Hopefully we'll do a few of them together.

   18. GAGuido Appenzeller26:39

       We'll be able to shine the scene. Yeah.

   19. PGPat Gelsinger26:40

       Yeah. Yeah. May-

   20. GAGuido Appenzeller26:41

       Hopefully.

   21. PGPat Gelsinger26:42

       May- you know, what do you think? Should we do... Okay, you know, but, uh, we'll, we'll find some things to do here.

   22. GAGuido Appenzeller26:46

       Yes.

   23. PGPat Gelsinger26:46

       But I think memory innovation for the first time in 30 years is nigh upon us.

   24. GAGuido Appenzeller26:51

       Yeah. Yeah.

   25. PGPat Gelsinger26:52

       It's, it's amazing.

   26. GAGuido Appenzeller26:52

       Long overdue. [laughs]

   27. PGPat Gelsinger26:56

       You know.

   28. GAGuido Appenzeller26:56

       Trillion. Okay, sorry.

   29. PGPat Gelsinger26:58

       Yeah. [laughs]

   30. GAGuido Appenzeller26:58

       Off by orders of magnitude.
9. 42:49 – 48:42

   Energy capacity equals economic capacity

   1. PGPat Gelsinger42:49

      Yeah, yeah.

   2. GAGuido Appenzeller42:50

      And you touched upon it with vertical delivery inside the chip-

   3. PGPat Gelsinger42:53

      Mm-hmm

   4. GAGuido Appenzeller42:53

      ... uh, on the chip. But, uh, I mean, there's 800 volt DC, there's power delivery to-

   5. PGPat Gelsinger42:59

      Yeah

   6. GAGuido Appenzeller42:59

      ... the data center.

   7. PGPat Gelsinger43:00

      Mm-hmm.

   8. GAGuido Appenzeller43:01

      Um, there's an entire ecosystem all the way to, uh, power generation and-

   9. PGPat Gelsinger43:08

      Mm-hmm

   10. GAGuido Appenzeller43:08

       ... and, uh, uh, the social political dimensions of that-

   11. PGPat Gelsinger43:13

       Yeah

   12. GAGuido Appenzeller43:13

       ... and so on and so forth.

   13. PGPat Gelsinger43:14

       Yeah, yeah.

   14. GAGuido Appenzeller43:14

       So where does innovation come in, and where does, like, it's pure execution?

   15. PGPat Gelsinger43:19

       Well, you, the, you know... Uh, and if we start at, you know, uh, first principles here, you know, our, our nation has done a terrible job with its energy capacity, right? You know, and essentially we went through 10, uh, 15 years where essentially I was taking coal offline at the rate I was adding renewables, and essentially the nation was flatlined in terms of overall-

   16. GAGuido Appenzeller43:42

       Yeah

   17. PGPat Gelsinger43:42

       ... energy capacity. Not good, because in a AI digital age, energy capacity equals economic capacity, right? So essentially, my economic capacity as a nation, flat for 15 years if you buy that thesis, right? Which I think is very, you know, uh, provable. Over the last five years, I've, you know, we've seen this enormous influx into just give me more capacity. Uh, but, you know, I think, uh, you know, even with that, we've gone to maybe increasing our national energy capacity 4% per year.

   18. GAGuido Appenzeller44:15

       Something like that, yeah.

   19. PGPat Gelsinger44:15

       Going from, you know, essentially zero to one, right? Wow, 4X, [chuckles] or, hmm, 4%, right? So to me, you know, that, you know, it's a bad situation, so we just need more energy capacity, uh, for it. So number one, and obviously we have some companies in that space.

   20. GAGuido Appenzeller44:33

       Yep.

   21. PGPat Gelsinger44:33

       You know, we talked about, uh, nuclear uprating as one of those-

   22. GAGuido Appenzeller44:36

       Yeah, yeah

   23. PGPat Gelsinger44:36

       ... with Alva. Um, and, uh, you know, I think nuclear, right, you know, as a base load is, you know, a fabulous thing that we want to go build more of. You know, unfortunately, all of our renewables have, you know, deep dependencies in China-

   24. GAGuido Appenzeller44:50

       Yeah

   25. PGPat Gelsinger44:50

       ... which is very unfortunate, uh, you know, for us. You know, we're, you know, uh, you know, gas turbine, uh, lead times are only eight years at this point, right? So, you know, you know, I mean, we really have a, a difficult environment to scale up rapidly, but one, we just need more capacity.

   26. GAGuido Appenzeller45:06

       Yeah.

   27. PGPat Gelsinger45:07

       And fundamentally, energy capacity somewhat dampens how euphoric we can get on AI, right? Because, you know, why build a new data center and buy, you know, the million GPUs if I can't power them, right? And I think you're gonna see more and more defaults happening on many of those data center projects because the energy won't be there.

   28. GAGuido Appenzeller45:27

       So you think it'll be a significant headwind?

   29. PGPat Gelsinger45:28

       I think it will be a headwind, and I think you're already starting to see some of the first indicators of that, the Oracle, you know, uh-

   30. GAGuido Appenzeller45:35

       Yeah
10. 48:42 – 52:59

    A VMware for agents

    1. RRRaghu Raghuram48:42

       Because now-

    2. PGPat Gelsinger48:43

       You're in charge.

    3. RRRaghu Raghuram48:44

       Yes. So, so I thought. [laughs] Um, now VMs are back.

    4. PGPat Gelsinger48:50

       [laughs]

    5. RRRaghu Raghuram48:50

       How do you think about that as a former CEO of, uh, VMware?

    6. PGPat Gelsinger48:54

       Well, you know, I, I, uh-

    7. RRRaghu Raghuram48:56

       Let's, uh-

    8. GAGuido Appenzeller48:57

       I feel, I feel excluded from this conversation-

    9. RRRaghu Raghuram48:58

       No, no, no. I was just-

    10. GAGuido Appenzeller48:59

        ... between former CEOs of VMware. But go on. [laughs]

    11. RRRaghu Raghuram49:02

        Part of it was I was gonna talk about networking as well.

    12. PGPat Gelsinger49:04

        Yeah.

    13. RRRaghu Raghuram49:04

        So, but I don't think we have the time.

    14. PGPat Gelsinger49:05

        But, you know, hey, you know, I, I think the, you know, every innovation has always led to the next abstraction, right? And I think many of these ideas and, you know, right, you know, I always like to say, you know, you know, what, what's, what's new in our world today, you know, data structures, algorithms, you know, and abstraction. You know, you know, we bring it back to those, you know, foundationals. And hey, you know, I think, you know, the virtual machine abstraction, you know, whether it's done from the infrastructure level or from the application level, you know, I think it's still a foundational abstraction model, you know, that deserves to always have a place, uh, in the, uh, compute hierarchy. So, you know, so, uh, what do you think? One of us go back and run VMware again?

    15. RRRaghu Raghuram49:44

        Yeah, that's what I tell... I, I think I told Guido, "It's time, it's time to go back and run VMware again."

    16. PGPat Gelsinger49:48

        Yeah.

    17. RRRaghu Raghuram49:49

        Yeah.

    18. PGPat Gelsinger49:49

        Yeah.

    19. RRRaghu Raghuram49:49

        Yeah.

    20. PGPat Gelsinger49:50

        You know, and, uh, you know, and I do think some of these abstractions, you know, but you also think about it and, like, one of the things we did at VMware, we, you know, essentially, you know, managed every aspect of the computing hierarchy.

    21. RRRaghu Raghuram50:01

        Yeah.

    22. PGPat Gelsinger50:02

        Right? You know, and you manage the workload, you manage the network, you manage the storage, you know, system, and how do you abstract those? And I think in the AI context, right, we think about, you know, these agent swarms. Well, who's gonna manage the agents? Who's gonna create the security profiles around all of the agents? Who's gonna be manage the performance of the agents? You know, essentially every fundamental element of virtualization and management needs to get recreated in this next computing hierarchy.

    23. RRRaghu Raghuram50:29

        Yeah.

    24. PGPat Gelsinger50:30

        And there's gonna be lots of wonderful companies that, you know, sort of break through for how those things will be, you know, done.

    25. RRRaghu Raghuram50:36

        Okay. Do you wanna end it on networking?

    26. GAGuido Appenzeller50:40

        No, no. A- actually I'd love to stay with the-

    27. RRRaghu Raghuram50:41

        Okay

    28. GAGuido Appenzeller50:42

        ... with the VMs for a second. And the, the most interesting thing for me is that we're now building VMs so that they're usable for agents as opposed to usable for humans. And what we've seen in sort of other companies is that that changes a lot the form factor, that changes a lot how you market them, right?

    29. RRRaghu Raghuram50:55

        Yes.

    30. GAGuido Appenzeller50:55

        You're working with... You want the agent to make the pick for, for, for your VM offering. Um, it, it changes a lot how much complexity you can have. It changes how, how quickly the startup should be. Humans are a lot more patient than, than agents.

Episode duration: 53:14