Cloudflare TV

🎂 Pat Gelsinger & Matthew Prince Fireside Chat

Presented by Matthew Prince, Pat Gelsinger
Originally aired on 

2021 marks Cloudflare's 11th birthday. For our annual Birthday Week celebration CFTV is featuring an array of new products and guest speakers.

In this session, join Cloudflare co-founder and CEO Matthew Prince for a special fireside chat with Pat Gelsinger, CEO of Intel.

Before rejoining Intel in early 2021, Gelsinger was CEO of VMware. In that role, he transformed VMware into a recognized global leader in cloud infrastructure, enterprise mobility and cyber security – almost tripling the company’s annual revenues. Gelsinger was also ranked the best CEO in America in 2019, according to an annual survey by Glassdoor. Prior to joining VMware in 2012, Gelsinger was president and chief operating officer of EMC’s Information Infrastructure Products business, overseeing engineering and operations for information storage, data computing, backup and recovery, RSA security and enterprise solutions.

Gelsinger began his career in 1979 at Intel, becoming its first chief technology officer, and also serving as senior vice president and the general manager of the Digital Enterprise Group. He managed the creation of key industry technologies such as USB and Wi-Fi. He was the architect of the original 80486 processor, led 14 microprocessor programs and played key roles in the Intel® Core™ and Intel® Xeon® processor families, leading to Intel becoming the preeminent microprocessor supplier.

Birthday Week

Transcript (Beta)

So welcome to Cloudflare TV. I'm here with Pat Gelsinger, who is someone who I've admired for a really long time.

He's the CEO of Intel, started his career at Intel, was the original architect of the 486 processor, which has got to be one of the, you know, just most revolutionary processors and game-changing processors, really put Intel on the map.

Went from Intel to be the president at EMC, and then took over as CEO of VMware, and then returned back to Intel.

So Pat, thank you so much for making time to be on Cloudflare TV with us.

Well, thank you, Matthew.

And, you know, great pleasure to join Cloudflare TV, chat with you today. And hey, it's been, for a farm boy from Pennsylvania, it's been a pretty extraordinary career.

You know, I, so one of my, one of my professors in business school was Richard Tedlow, and Richard wrote, Richard wrote Andy Groh's biography.

And so, and I wish I had thought to do this earlier, but as I was, I was getting ready for this, I shot him a quick note saying, you know, if Andy were still with us, what would he want to ask Pat?

And he hasn't written back to me, so I don't have an answer from him.

And I promise I'm not going to surprise you with a question, but what was it, what was it like to work with, you know, such an incredible leader?

And, and I, you know, he was, he was, I understand your mentor for a long time.

Tell me about Andy. Well, you know, the first, you know, let me, you know, the first story of when I first got, started mentoring with Andy Groh, and then one story at the end.

So we'll sort of bookend my time with Andy. You know, I was working on the 386 and was updating the executive staff of Intel on the tape out, which is basically finishing the design, putting it together.

And basically, as a 20, I think I was 21 years old, maybe 22 years old at the time, this young kid, right?

And I'm basically scolding the executive team that they got to fix my computers because they kept crashing.

The database sizes were getting too big to finish it.

And about a couple of days later, my phone rings, I pick it up, you know, who is it?

I didn't want to be interrupted. I'm working like crazy, trying to get this chip out the door.

And the voice comes back, Andy, Andy who? And the voice comes back, Andy Groh, but I'm like, ah, and you start shelling me with questions.

And he says, those are lousy answers. Be in my office in a week with better ones.

And that began my mentoring relationship with Andy Grove that literally continued up until the day he died.

And, uh, uh, you know, I always joke that mentoring with Andy Grove was like going to the dentist without Novocain.

He was hard.

He was demanding, but he made you better. And, you know, just had this, you know, just this insatiable right.

Pursuit of what was true, what was right, you know, what was the best answer, you know, in everything that he did.

And he demanded that of everybody you worked with.

And, uh, that was, you know, and I'll just say, it was a really quite an experience to be shaped, uh, by a guy like Andy.

And then, you know, late, late in his life, you know, he continued to mentor with me and, you know, it was a little bit over five years ago that Andy passed away.

And I remember my last dinner with Andy was maybe three months before his passing.

And, uh, it was one of those where we had a lot of tough stuff we were talking about in it.

And I composed a mail to him that I was just about to hit send. And I heard that, uh, he had fallen quite ill, uh, spoke to his wife, uh, Eva right before it.

And I still have the email that I was sending to Andy, right. You know, in my outbox, right.

And it's just sort of sitting there just sort of as that last little tribute to the great influence on my career and, uh, life.

And just to tell one of the little story, I'm in Hawaii, uh, next week for a sales, uh, recognition event, and I'm speaking to Gordon Moore and, uh, Gordon's 92 now.

And I'm having, you know, sort of a last interview with, uh, Gordon, uh, as well.

And, you know, to have been mentored by worked with, you know, Gordon Moore, Robert noise, Andy Grove, and now being the CEO of this company, this is pretty special.

What so, and, and feel free to say none of your business, but, uh, any, any, anything that you can share about what, what you guys were going at, or what the contents of that email were.

Well, you know, it was, uh, at the end of his career, you know, two things were really important to Andy was succession and values.

Yeah. Right. And in it, you know, it was very much, you know, this whole idea of succession and, you know, you know, to some degree he was asking me what he could have done better in that regard.

And I'm like, you're my mentor. I don't give you advice. You give me advice, right.

Uh, in this kind, you know, subject and, you know, corporate values, you know, how you reinforce it, how you build it, how, how values are the most lasting thing that a leader does.

Right. And as a CEO, I view myself not as the chief executive officer, but as the chief culture officer, right?

Because you will set, define, reinforce and demand of the leadership team and the company, you know, the values that you want to persist for years, if not decades after you've left this role.

So that's a pretty, it was a very powerful thing that Andy and I talked a lot about, you know, as he was coming to the end of his time on earth.

Did, um, what, what, what, if you, if you sort of talked about what, what the culture is that, you know, you, you, you, you are sort of leading at Intel today.

What, um, what does that, what does that, what does that look like?

What are you, what are you, uh, what have you brought to the, brought to the place?

Well, you know, as I think about it and, you know, I've talked a fair amount about it, you know, I've been back at the company now seven plus months, uh, for, and I've said, we want to reestablish the Grovian culture, right.

And a lot of people here, you know, at, uh, Intel, when I say that, you know, you know, veterans, uh, you know, 10, 20, 30 years with the company, they say, yay.

Right. You know, we're back right in that regard.

And other people are sort of saying, what are you talking about? Right. And in it, you know, it's this, you know, it is this demanding data driven, right.

You know, where, Hey, if you don't have the data and the point of view, Matthew leave the room, right.

I'm with the people there who can really chew on and have the real discussion.

The real discussion happens in the room. You know, we, uh, right.

You know, we will, we, you know, we'll work hard to come to consensus, but if we don't, we will disagree and commit to a path together.

There's no right. You know, subtle, right.

You know, things that are left unsaid it's said in the room, you know, we aspire to engineering excellence, you know, and if you're a great engineer, this is a culture that you thrive in and you love because of that directness, that pursuit of data and the truth, you know, it's also paranoid, as I said, you know, you know, we, we, we assume that competitors are going to do good things and how do we do better things in that process.

But, you know, also, as I talk about the culture today is, you know, the Grovian culture was very appropriate for that period of time and uniquely emulated, you know, things like key results and objectives and many of those things have lived on, but it also needs to be modernized.

And in this environment where things like diversity inclusion, uh, you know, uh, ESG requirements for companies have become very appropriate, you know, that wasn't part of the original culture of Intel.

So how do we both, you know, what is old is new again, but also bring some of these things that really make it, I'll say a better, more modern culture as well.

And if, if we end the Gelsinger era, if I could call it that, you know, having reestablished many of those things, as well as modernize refreshed, right.

A culture that lives on, I'm going to feel pretty good.

Do you, uh, do you have, is there, are you mentoring sort of the, the, the next pat today?

Do you have people that you, you, you sort of have the same relationship with that Andy had with you?

You know, I have a couple of people that I mentor with regularly as well.

Um, and, uh, I continue to have a mentor now that Andy's passed.

I do have one retired CEO. Who's my mentor. Um, my father just passed away a couple of weeks ago, you know, which was a very sad moment.

I feel like I'm, I'm losing all of those older male influences in my life and, uh, a career or something that I recommend everybody to have.

Uh, but one of the things I've told my chief of staff is, uh, Hey, I'm, I'm looking for, you know, a couple of those, uh, you know, uh, young, aggressive, uh, and when I was first interviewed, uh, to, to come and interview with, uh, Intel, you know, the interviewer, Ron Smith, you know, retired a number of years ago, you know, this is what he wrote about me.

He said, smart, aggressive, arrogant, he'll fit right in.

And, you know, so I guess I'm looking for a few of those people now that I've come back to the company to add to my, uh, portfolio of, uh, mentees.

Cause I do think there's a pretty unique aspect to take somebody much earlier in their career, right.

And help them being shaped by a, uh, a grizzled veteran like myself.

So one of the things you said was, um, that, you know, when you were, when you were 21, 22, you were helping design the three 86.

And then, you know, you're, you're known as the architect of 48, six. What is, what does that mean?

I like, I I've never designed a semiconductor. What is, what does that mean?

Like, what does that look like? Um, you're not, I mean, I, you say you tape something out.

I physically, you know, picture you with Scotch tape.

That's, that's obviously not how it works. What, what did you do back then?

And how was it, how is it different as you're thinking about, about semiconductors today?

Yeah. And you know, when, when, when I architected the 46, literally I was engineer number one.

You know, what is it going to look like? And we were making decisions like bringing floating point on chip, you know, establishing the first on chip cash in the X 86 family, you know, setting the micro architectural direction directions around a deep pipelined architecture at the time setting clock per instruction goals.

And, uh, what instructions we were going to add to the instruction set?

Well, that's what it meant to be the architect of the 80 46. And I took the 80 46 into production.

So it was like, I, from soup to nuts, that was my chip from beginning to end.

Right. And, uh, you know, I'm even blessed, uh, to have, you know, uh, a number of years later, you know, maybe, uh, you know, we, uh, launched the 80 46 in 1989, I think it was around 1992 or 93, the museum of modern art to New York had the chip art expedition, right.

You know, and here they were showing off and, you know, all these different designs and so on of Silicon chips.

And, uh, the 80 46 was selected as the chip that was put on permanent display at the New York museum of modern art.

So I'm an artist, you know, and trust me, I'm not an artist, but, you know, it was one of those, uh, uh, moments in time.

And so that's what it meant to be the architect of the 46.

And when we say we tape it out, that's when, you know, we finished all of the simulations, all the designs, all of the layout, you know, to send it to fab for the first time to the manufacturing, you know, when you get the first thing that's not in the computer simulation, but the actual chip now today.

Right. And I remember I'm a meeting with one of our young smart, you know, maybe he's going to be one of my mentor mentees, uh, in the future, this, uh, you know, young punk, he says to me, Pat, you're pretty proud of your 80 46.

And I said, well, yeah, I am right. You know, and, uh, you know, a million, a million transistors, you're not even the power control unit of a modern microprocessor.

That's now a hundred billion transistors, right. Right. You know, cause we were 1.2 million transistors on the 80 46.

Well, actually 1 ,180,486.

I added a few at the end to make it, uh, to round appropriately to be 46 today.

You know, our, our most modern high -end processor today is over a hundred billion transistors, right.

Which, you know, when, uh, you know, the 80 46 was 25 megahertz when we taped it out, you know, at approximately 12 MIPS or 12 million instructions per second, you know, the high-end, uh, Ponte Vecchio, a hundred billion, uh, transistor chip that we're just introducing now is approximately a teraflop on a chip, right.

You know, which is, you know, the progress that's been made, you know, since then it's just extraordinary as you think about it.

And just to put this in context, the highest performing supercomputer, right.

13 years ago was a teraflop.

So literally I can put on one chip in the pocket of my, you know, on the palm of my hand, right.

I can equal, you know, the IBM Roadrunner 13 years ago, that was the highest performing computer in the world, uh, at the time that filled a good sized warehouse room.

Now it's a chip. Yeah. I mean, the, the amazing thing about this is that, you know, it seems like it's an industry that's changed so much in so many ways, but the fact that like the Intel instruction set is a lot of, a lot of what you were architecting back then is still, still around, um, you know, and you've just made, uh, you know, relatively recently, a fairly substantial announcement, you know, 90, $95 billion CapEx investment over the next 10 years, a new, a new chip architecture.

Like what, what are the things that like, what are the things that need to change, um, going forward?

And what are you, what are you excited about as you, as, as you look at Intel over the next 10 years?

Yeah. You know, there's so many things that are going on, you know, and I've called it that, uh, in our industry today, you know, the four superpowers, this, you know, first ubiquitous compute, you know, compute is becoming relevant and everything is not just your personal computer, right.

You know, but it's, you know, your light switch, it's your car, right.

It's your autonomous delivery. You know, everything is including compute ubiquitous connectivity, right.

Where everything is getting connected. Uh, today, 60% of humanity is connected by 2030, 90% of humanity.

And after we're done connecting every person, we're connecting everything, you know, where everything is capable of computing and of connectivity, right.

You will have, uh, ubiquitous infrastructure, right.

We're now essentially in the cloud. I can give you on a moment's notice, the largest supercomputer that's ever been built, right.

Where it's sort of like, okay, you know, how many Tara flops would you like right now?

Right. You know, and right, you know, with edge computing, we will also not only have large compute, but we will also move it to the edge where you have low latency, high bandwidth connectivity to infrastructure as well.

And finally, right.

We can bring intelligence to everything with AI, right. Literally, you know, and I, I remember we're talking about the 46 Matthew when I was architecting the 46, you know, we had the early days of AI and it was cool.

It was fun.

It was sexy. And it resulted in exactly nothing. Yeah. Right. That was 30 years ago.

And as I like to say, it was an overnight 30 year success, right. Where the algorithms kept getting better.

The data sets kept getting larger. The compute kept getting bigger.

And about five years ago, all of a sudden AI could do lots of interesting things.

So finally bringing intelligence to everything. So compute to everything, connectivity to everything, infrastructure to everything, and now intelligence to everything, right.

You know, and as these come together, you know, literally we're seeing technology, these powerful forces, as I call them that are now permeating everything.

And every aspect of human existence is becoming technical driven.

And underneath that everything technical, everything digital runs on semiconductors.

So literally we're seeing the industry's role in humanity become essential to everything that we're doing and look for in the future.

So tell me what part of your life isn't becoming more digital.

And wherever we were, COVID was a big accelerator.

Absolutely. Work from home, learn from home, right. Health from home, every aspect of our lives were accelerated in that regard.

And, you know, so we're working on so many compelling things, right.

Where you sort of say, huh, you know, how do I, you know, go from now, you know, next year we'll deliver the first exascale computing.

Yeah. Well, we're planning the first zeta scale computing.

What does it take to do the next thousand X and what problems can be uniquely solved when I give a thousand X bigger computing, a thousand X bigger AI training models, a thousand X bigger data sets that you're working on.

That'd be an example of some of the things we're looking at, you know, probably by the end of the decade, we'll be closing in on quantum supremacy where we'll have the first realizable, meaningful quantum machines that will start doing, you know, things better than digital.

Right. So then we'll start saying, how do we start weaving quantum capabilities into the computing fabric that we have?

Yeah. We'll also be by the end of the decade, you know, we'll be looking to make the next big G right.

Where we go from 5G to 6G and 5G isn't just a faster 4G, right. It's fundamentally a secure low latency, high bandwidth network, right.

That in many cases is better than wired, right.

You know, we will also see, you know, broad deployments of optical computing as well.

You know, things that were, today on copper wires, et cetera.

Well, we won't have the power, right. To drive them over copper.

We see big breakthroughs happening as we move to a ubiquitous optical connectivity.

So, you know, those would be just a handful of some of the things that we're working on.

And, you know, we see no end in sight. And as I mentioned earlier, you know, Gordon Moore, I'm talking to him this next week, the famous Moore's law.

Well, you know, as I say, we're not done with Moore's law until we have exhausted the periodic table and we got a lot more elements to go.

What, so what, you know, a lot of what you just talked about is focused around, you know, the CPU and that being the center, you know, I think famously, you know, Andy Grove really changed Intel from being a memory company to being a CPU based company.

How much of the future of Intel is CPUs or, you know, how much are you excited about?

I mean, you're doing some really amazing things with storage and other, and you referenced, you know, 6G, which is obviously radio technology.

How are you thinking about how to sort of prioritize across those things?

Or is that the wrong way to think about this?

Well, you know, if I look at it from maybe two or three different dimensions, Matthew, you know, one is, hey, you know, there's memory and there's logic.

You know, we're everything logic, everything digital, everything that can be, you know, deduced into some form, right, of digital processing, we're going to lead.

At the center of that is compute, right? But everything, you know, as you go look at 5G, it's a big compute, right?

You know, as you go, you know, do signal processing and all those kinds of things.

So, you know, we're finding more and more uses for compute.

You know, we'll be less centered on the x86, but the x86 is still the center, right?

Because we're going to have GPUs, we're going to have intelligent network processing, what we call IPUs.

We're going to apply CPUs to more problem sets as we go forward, where, you know, AI processing, hey, today they're more specialized, but hey, we believe AI deserves to be part of every computer going forward.

And, you know, maybe my biggest training things are done with specialized compute, but we're going to be doing, you know, model updates and inference on everything going forward.

So, you're going to see, you know, AI capabilities become permeating, much like, you know, we used to think of microprocessors as being 8, 16, and 32-bit integers.

Well, now, if you didn't have, you know, if you can't do FP64, you're not about, yeah, you know, that's been a required part of the platform for three decades now.

Well, similarly, right, or, you know, AI capabilities where you're doing, you know, matrix processing, bfloat 16s, you know, being able to do sparse matrix processing.

Well, everything is going to be able to do that because it's become part of the expected workloads for everything.

So, you're going to see this broadening definition of what is logic, what is compute being applied across more and more workloads and across more and more domains where, hey, you know, I just gave a speech at the mobility conference, the biggest car show in Munich, and I described the car in the future as AV and EV happens to the auto industry.

It's a computer with tires, right, because everything that's happening is turning it into a modern computing platform that happens to have tires on it, right, you know, for these trends.

And we're seeing that across every industry.

What, you know, and that one of the things that I think has been top of mind for anyone, you know, in the auto industry and others has been really the semiconductor shortage that the world is going through right now.

And I think you'd have a really unique view into kind of what the, what is the supply constraint that's keeping, you know, across everyone in the semiconductor industry, enough semiconductors to build cars or switches or CPUs or anything else.


And, you know, how did we get here? Maybe it's the first question, which, you know, I believe that the semi-industry was headed for a shortage, right, where, you know, we've seen the industry growing four to five percent per year.

It's about a half a trillion dollar industry right now.

And, you know, it was accelerating for the digitization of everything, right.

So I think we were already headed to a modest shortage and then COVID happened and COVID exploded demand.

Right. Everything went online.

Everybody's shopping online. Everybody's doing Grubhub. Everybody needs three computers at home because now, you know, mom's working online and dad's working online.

The kids are taking classes online. So you've seen this explosion of demand where the growth rate of the industry went from five percent to over 20 percent.

And the supply chain went from five percent to negative. Factories are idled, right.

Supply chains are disrupted. So all of a sudden, what probably was a one or two percent shortage became a gulf, right.

And, you know, literally, you know, last week, new factories were shut down in the semiconductor supply chain in Malaysia because of a Delta outbreak.

You know, and we're not so we're not done, even though we're you know, we as a semiconductor industry are working like crazy to catch up.

And, you know, we're squeezing our factories harder, bringing in new equipment, building new facilities as rapidly as possible.

But we're still even facing incremental supply disruptions. Just last week, you know, new factories are being shut down as a result.

So the other issue of the semiconductor industry is it only takes us three years, three plus years to build a new factory.

You want more chips, Matthew? I'm on it. Right.

You know, I'll give me a call in 2024. We'll give you all you want. Right. You know, so like that, you know, these things are the most complicated, you know, advanced manufacturing capacity and a modern leading edge fab.

And I announced we did groundbreaking on two on Friday of last week, a modern edge factory is 10 billion dollars and takes approximately three years to build.

Right. I mean, these are extraordinary facilities that we're building and, you know, the chemicals, the electrical, right.

You know, the you know, all of the piping, the you know, we have individual pieces of equipment that go into this that are 200 million dollars for one piece of equipment.

I mean, extraordinary complexity. So, you know, the industry takes a while to ramp up.

So you've seen the supply being impacted, this demand explosion, and we're just racing to catch up.

And so now everybody is scrambling and now you have, you know, multiple order.

You know, hey, you know, I'll buy 10 from you, Matthew, and 10 from Mary and 10 from Fred.

Right. I only need 11, but I'm going to get, you know, start securing supply for 30.

Right. Just in case I can't find them all.

Right. You know, so you're finding all of this disruption and the supply chain.

And, you know, I mean, we have car factories that are producing 30 thousand dollar cars for absence of 50 cent microcontrollers.


Yeah. Yeah. How does a Detroit or Ohio car factory tell their employees that we're idle?

Right. You know, the parking lots are full. Right. You know, with 99 percent complete cars and trucks and so on.

Right. For minor pieces of semiconductors.

No, go home. We can't build any more right now. I mean, it becomes a very politically and economically untenable situation.

So industry is working super hard to catch up.

We're working hard to catch up, but it doesn't have short term fixes.

And as I said, you know, I expect the shortages to be the worst in Q3 and Q4.

The quarters we're in right now, incrementally better each quarter next year.

And it'll persist into 2023. And people keep telling me I'm wrong.

And every day I'm proving to be more right. So as you think about, you know, the investment, I mean, one of the things Intel's always been known for is that you are much more vertically integrated than a lot of the other semiconductor businesses.

And you're making investments in modern fabs and everything else.

How are you thinking about that? Like how what's what's the future of Intel in terms of running fabs?

Would you ever outsource more of that or would you ever let other people build chips in your fabs?

Yeah. You know, we you know, so we we rolled out a new strategy as I came in as CEO in March, and we announced what we call IDM 2.0, which will be very explicitly saying we leverage other foundries, but we're also becoming a foundry.

So the majority of our volume is going to be we are building our chips for our customers, period.

But we're going to leverage foundries.

Sometimes they're a little bit better than ours in some particular areas.

So we're going to leverage those to augment our supply chains.

I'll have some products that are, you know, both internal and external. So I have flexibility, but I've now for the first time opened the doors of our fabs wide open to become a foundry for the industry as well.

And there's really a part of this is augmenting our business model.

But part of it is also realizing that in COVID that we've been way too focused for the last couple of decades on cost of supply chain.

And what we need to be more focused on is geographically balanced, more resilient supply chains.

Right. So we said, hey, we are going to be a U.S.

and European champion to build more capacity rapidly. That's why we announced our two factories last week.

That's why I've said I'm going to open our next major factory locations in Europe and in U.S.

before the end of this year. You know, we are investing aggressively to both satisfy, right, this overwhelming demand for semiconductors, as well as building a more globally balanced, resilient supply chain.

Well, I'm incredibly excited. You know, Intel, when we were first getting started, your startup program gave us 20 CPUs for free.

And we built the original Cloudflare network on Intel CPUs.

And so I guess if I had one ask, my ask would be, you know, we've always been very conscious about the amount of power that we use and in our server chips.

And so my ask would just be, you know, please, more cores per watt.

And we'd be really excited. And we hope that Intel is a big part of our network going forward.

So I really appreciate the time that you've had. I know you have an incredibly busy schedule.

Thank you so much for coming, helping celebrate Cloudflare's 11th birthday.

And it's an incredible honor. And congratulations on coming back to Intel.

Well, happy birthday to Cloudflare. And it's my honor to participate in your birthday celebration.

And to all of your listeners, you know, it's a great, great honor to participate today.

But we look forward to what we're going to do over the next 11 years together.

So thank you so much.

Awesome. Thanks, Pat.