Arista Networks CEO: The AI Infrastructure Boom, Power Limits, and What’s Next

0:01Hi, everyone. I'm Nikolaj Tangen, the CEO of the Norwegian So-owned Wealth Fund. And today I'm here with somebody who is absolutely vital to AI, namely the remarkable Jayshree Ullal, CEO of Arista Networks. Jayshree has more than 40 years background in networking, so nobody is better able to talk about this than her. So Jayshree, super to have you on there. Great to be here, Nicole. I have a bit of a cold, so I apologize if I sound nasal. No problem. I don't have to sing on this podcast, right? So we should be fine. No singing. Let's start with the beginning.

0:47What does Arista do? Yeah, that's a loaded question. So Arista is building the world's most demanding mission-critical networks, which is high performance, high scale, low latency, high reliability, high availability, high automation, high telemetry. So it's a lot of demands across a lot of spectrums. And Arista has emerged, especially over the last decade, to be both the pioneer and the leader in this. Hmm. I've heard that if you think that kind of chips is the engine, then the networking is the highway between them, right? Yeah, that's a good way to look at it. I think the chips fuel a lot of, especially in this world of AI, processing with GPUs, APUs, XPUs, whatever you call them.

1:41But as you process this, you have to get worried about, are you processing it to full utilization? And so, you know, if a car is rated at 100 miles an hour, but can only go 30 miles an hour because the infrastructure cannot support going farther, then you're grossly underutilizing the car or in this case, the GPU. So, yeah, we are the information highway to all accelerators, users, mission critical workloads, you name it. So let's say now I sit in my office, I do a prompt on ChatGPT. how do I touch your products? Just what happens with the scenes? Yeah, so we are very much behind the scenes. This is something my parents always used to say, what do you do?

2:25Because you can't see what we do. And when you type in a prompt of Arista, basically what at that point we're doing is connecting everything to everything. This conversation on this podcast is not possible without it. So the traffic patterns of high intensity, whether it comes from a user, a device, a workload, a machine, a server, a storage, or now more and more the AI killer applications, we connect. And we do that through a set of switches and routing platforms. We call them leaf spine architecture, where you can have multiple leaves that connect into an aggregate spine. And these days you can also have an AI spine.

3:04So these are some of the world's largest high-scale platforms. But they need to be powered. Platforms and chips need to be powered with the right software. And that's where Arista's hallmark has been its extensible operating system, EOS, where we fuel these platforms and networking connectivity with more and more types of data. We can deal with structured data, unstructured data, flow data, and make that all work through our purpose-built software. And how is networking from AI different from what came before? Yeah, very different. So just to step back a little, if you go back even 20 years in networking, the internet was born through a lot of, in fact, the old days it came, the TCPIP and all those algorithms came from the research community, the ARPANET.

3:56And then there were multiple protocols, things like XNS, AppleTalk, OSI, IP. TCP IP became the worldwide language of internet communication in the 2000 era. In the 2015-2020 era, that wasn't enough. People were starting to build some of the world's largest clouds, whether you think of AWS or Azure or GCP, Google, or so many more. and they needed the kind of unprecedented scale to touch so many more companies and users. It had to be multi-tenant because they were not supporting one company, they're supporting thousands of companies. So a typical CIO of a mid-sized company stopped building their own things.

4:38They started relying on the cloud for their mission-critical assets, whether it was e-commerce applications or their databases or whatever. Now in comes AI the last couple of years. And that's adding another twist because cloud traffic is very different than AI traffic. AI traffic, you know, and we all saw this with the onset of ChatGPT and all the other language models, is conducting a class of complexity and search that is thousand or million times deeper and greater, broader. And so this required even more intensity of the traffic patterns. So typically we were doing with before AI what we would call front-end networks, connecting standard CPUs, storage, general purpose query, indexing, searching, that kind of things.

5:28With the advent of AI, we were suddenly connecting hundreds, if not thousands, or even hundreds of thousands of accelerators. And this GPU traffic, as I said, required you to very different traffic. You know, it's much more intensive in the fidelity of the traffic, the intensity of the traffic, the durability of the traffic, the peak burst. So we had, this is often called a back-end network or a scale-up and scale-out network. We had to build almost a different set of parameters and metrics. And until we came along, I would say majority of this was little islands of things. There was an InfiniBand network, a PCI network, a CSI network, and Arista was able to bring Ethernet and IP and standards-based capabilities like we had in the cloud into the back end for AI.

6:20What's the biggest constraint now in terms of building out these networks? You know, I would say the biggest constraint for the industry at large is power. because the power consumption of these GPUs and the network and the cables and the optics and everything is just unprecedented. In the old days, we used to talk about megawatts. Now we talk about tens of gigawatts in a data center. To put this in perspective, this is larger than a football field, and you're powering that with gigawatts of capability. So our AI providers or cloud providers are unable to find this level of power. If today you're searching for that kind of power, it could take you three to five years to find it.

7:04The space and the power are probably the biggest problem right now. But once you do find it, you have to outfit it correctly. And that's where we come in. Absolutely. And so if you find it, you have to call you guys. But now enormous amounts of money flowing into AI infrastructure right now. Just how do you view this? Is it a bubble? Is it too fast? Yeah, it has happened very fast. When I look at the cycle of how networking progressed from 100 megabit to 1 gigabit to 10 gigabits to 100 gigabits, that process happened every five years maybe. But the migration from 100 gigabit to 200, 400, 1.60, and eventually 3.2 terabits is literally all happening within 12 months each.

7:54So what used to be five-year cycles has become 12 to 18-month cycles. So when we think, okay, we're just done with 400 gig last year, we're already thinking of 800 gig and 1.6 terat this year. So the insatiable demand for speed, for scale, because you've got to aggregate all of these throughputs, and for predictable latencies all hitting us at once. You were at Cisco during the internet boom. I was, I was. How does this compare with what you saw at that stage? Much greater. The internet boom happened in a very steady fashion. There was a period of time where there was a peak in 99, 2000 due to the dot-com boom and bust, I would add, and then Y2K as well.

8:39But for most part, that internet took 15 years to develop. I was there from the very first acquisition in 1993 of Crescendo, when the company was less than a billion dollars, to when I left in 2008, it was a 40 or close to$50 billion company where the catalyst switches, which I think are still Cisco's most successful product line, were contributing at least a third. The point though was it happened slowly and steadily, but there was a peak, there was a bubble for two years. This one feels very different for a number of reasons. First of all, that 15 years that I was talking about is collapsing literally into five to seven years.

9:15The time aspect of this is shrinking. Secondly, the scale aspect of it is just exploding in magnitude. It's a thousand X more. And the third, perhaps the most important one, because people like to make this parallel to, are we in a bubble? I think it's a lot of responsible companies and therefore we're not in a bubble. Or if we are in a bubble, we're in a prolonged bubble. It's not Celex that sort of came in, were funded, and then they collapsed. These are responsible companies like Microsoft, Google, Meta, Oracle, that are taking responsibility of building all of these, not to mention the neoclouds as well, such as OpenAI and Anthropic.

9:57So, you know, I would say more than a bubble, it feels like an explosive mega trend that is here to stay for some time. We did have overbuilding at the time of the dot-com boom. How can we be confident that it doesn't happen again? Because it's not available. The power and the capacity is not, so nobody can really overbuild. It's going to take time to build it. The desires are greater than our ability to execute on it. So I think it'll still be a three-year execution rather than what happened in 1999, where everybody just hurried up, built these companies and said, build it and it will come. Here, there isn't a build it and they come.

10:35It is, please build it. we need it now. That's the biggest change. But again, that's not to say that I wouldn't correct myself and say it isn't a bubble, meaning it is happening in a furious, sudden state, but far greater than the steady state we've all been used to. But I don't think it's being done for a one-year factor. It's a three to five years, and it's being done by responsible companies who are voting with their money because they see demand. That's the biggest difference, that the demand is there. Talking about responsible people, so your biggest clients are the hyperscalers, right? Correct.

11:15Cloud companies, Microsoft. We call them the cloud and the light titans. How concentrated is the customer base? Well, you know, they've always been concentrated. Microsoft and Meta, as an example, have been more than 10 % of our revenue each since we went public in 2014. And I'm not going to apologize for that concentration. I'm going to be pleased and grateful for that partnership. So they continue to be 10 % of our revenue. But of course, as the denominator gets larger and larger, we are diversifying to a lot of specialty cloud providers, to the enterprise, to the campus. But I would be kidding if I didn't tell you they're very, very integral to Arista's strategy.

11:57And the partnership has never been stronger. Which country are you the most impressed by when it comes to building the AI capabilities? Which country? Undoubtedly, the United States. I think we're leading. There are pockets of things happening in other countries. But in terms of scale and the companies we're working with, majority of that is in the U.S. today. Any other countries which you think are doing the right things? Definitely. I think we're seeing pockets of this in the Nordics, in the Middle East, some of it in UK as well, some of it in Korea and India. So no question it's coming to many of the other international locations as well.

12:44When you look at where investments are taking place, are you worried that it's going to split the world even more than before? No, maturity of the investments I'm seeing are not splitting the world, but connecting the world more. So thankfully, they're not government sponsored. So they're all about AI and technology and how to increase the communication. I was just in the Middle East a month ago, and I'm just so impressed with the rate of progress there, where they are investing a lot of their oil currency into important AI investments and capabilities. So I don't look at it as a divide. I look at it as a unity.

13:21Talking about Impressed, you have now 21 % of the market for AI data center networking, which is incredible. Just how did you get there? Is that all we have? I was hoping it was higher, but we got work to do then. We've climbed our way beautifully from zero to 21. I think if you look at the high speed market, we're more at like 41. But if you look at the general speeds of all switching, then we're probably more like the number you mentioned. Look, it's been a labor of love. The company was founded and started shipping products in 2008. So we have not done it, you know, in a bolt of lightning. We have done it steadily, but surely.

14:02And really, I think to epitomize the culture of this company that Andy Bechtelstein and Ken Duda started, I would have to say build the right foundation for our software, for our hardware infrastructure, and keep improving it. So our market share really comes from the fact that our customers appreciate our innovation. We're built by engineers for engineers. We put a maniacal focus on quality, architecture, and doing the right thing for our customers and supporting them. Because we're sitting there in the heart of a mission-critical environment where if we go down, everybody gets to find out. if we don't go down where the world's well-kept secret, then that's fine with me.

14:41So I think we have acted in a very methodical fashion and not just rushed a product to market before it was ready.

14:55Now, you spent 15 years at Cisco building up their data center business. Just what was it like being at Cisco? It was more than the data center business. I love my time at Cisco. When I joined Cisco, Cisco was approximately$650 million in revenue and$6 billion in market cap. So it wasn't a startup, but it was certainly a very young company. And so I got a chance, and the company was largely building routers. So I got a chance to contribute in creating and building the Catalyst brand of switches. First, it was Catalyst 1 ,200, then 5 ,000, 6 ,500, 4 ,000, 3 ,000. So this switching brand went from being zero when I first came in there to I still remember the first year was 40 million, then 365 million the second year, and then a billion the third year.

15:46And those kind of things, you know, don't happen without the worldwide web and the internet taking off no matter how great your products are. So I felt I felt privileged to be part of a rocket ship. It was a great company. And, you know, but at some point, I felt like, wait, I didn't intend to be in such a large corporation. I wanted to do something smaller. But I fondly and nostalgically remember my Cisco days and those 15 years. What was the most important thing you learned from being on that rocket ship? I think the same thing we have here, which is that focus on customers. Cisco has a culture of customer focus.

16:20You know, I think Arista doubles down on that as well as innovation and engineering and quality. But I've taken calls on behalf of Cisco at 2 a.m. As an executive sponsor, if something was serious, I'd pick that phone up, you know, even though I value my beauty sleep. That was more important at that point in time. So I would say the culture of customers is super important. Now, you could probably have been the, you know, the next CEO of Cisco. So why did you want to jump spaceship? You know, I don't know if I would have been the right CEO for Cisco because I was an accidental executive at Cisco.

17:01I didn't plan to even be the executive I was. I love products. But don't you think most people are accidental executives? I mean, it's not like you're five years old and you sit in India and think, hey, I'm going to be a superstar in Cisco. Oh, heck no. You could ask my parents how little of a superstar I was when I was five years old. But leaving that for the moment, I guess what I'd say is I loved product. I loved working with the engineers to enable products to customers. And I was too steeped in the technology interests to think I had broader capabilities or desire to be a CEO at Cisco. So actually, when I left Cisco, the fact of the matter was I thought I would do something entirely different like clean tech or energy or solar or battery.

17:51And I really wanted to get back being in Silicon Valley to the roots of what I love, which is being an entrepreneur. And so I joined Arista when it was zero revenue and with 30 engineers. And the most important thing I loved about that, which is working with the people I enjoyed. So let's say now I have a startup, I got zero revenue, 30 people hanging out in an office. How do I recruit a superstar like you were at the time into the total uncertainty? Yeah, no, I think Andy and I go back to, what was it? We knew each other. So this is Andy Bechtolsheim. Andy Bechtolsheim, the German co-founder of the company.

18:33So, you know, we had bought his company, Granit, when I was at Cisco. So, you know, we had worked with each other. We worked also when he was at Founded Sun Microsystems and I was in AMD building chips. So basically, we've known each other for, you know, since 1987, 35 years or more. So I think, you know, when he saw that I was leaving and I didn't, I truly did not know where I was going. I did not want to make a decision on where I was going. I wanted to clear my head. You know, he said, hey, I got the startup. And I said, oh, my first reaction was, oh, no, Andy, I don't want to do networking and switching again.

19:08But as I got to understand what the founders were trying to do and what the engineers were building, I was intrigued because it was networking and switching with a twist. It was not the internet and catalyst switching like I knew it was a completely different ballgame. And what exactly was that? What was that twist that made you interested? The software, the software architecture was so unique. you know when you look at what all the other networking companies have done they usually have five operating systems for five use cases times 50 images you're managing 250 here we had one operating system one binary image with a lot of class foundation that we call the state-driven published subscribe model where if something fails it doesn't just fail it automatically recovers that agent and you know it's the opposite of a christmas tree light where the christmas tree light fails, everything goes down.

19:58Here, we built something where one light failed, we could actually replace that LED and keep everything up and the customer would never know it. So that software has what has earned the right for us to be in all these mission critical, high profile customers you named. And then of course, we've taken it to a variety of use cases. So I was drawn to that technology. I was drawn to the people. But I would be kidding if I didn't tell you there were moments of anxiety because the year I joined was 2008. It was in the middle of the financial collapse. One of our earliest customers that I won was Lehman Brothers.

20:31And you know what happened to them after that? I do. So I had an ominous start to my entrepreneurial career at Arista, which is now 17 years and going strong. So it's not like we didn't go through ups and downs. We absolutely did. The financial collapse was one. Our largest competitor, Cisco suing us was another. COVID and the whole supply chain crisis was another. So we have fought a lot of adversity to be where we are. And it's good to be in this position. What does it take to take on a giant like Cisco? You know, be a startup and it's just like, hey, we're going to beat them. Just want to, what does it take?

21:10I don't, most of my colleagues would tell you, I don't think I ever felt I was taking on Cisco because initially we were so focused on very specific and unique use cases that Cisco didn't even want to be in or pay attention to, like high-frequency trading and low latency. That was a use case they just didn't deal with. And our first 100 million came largely from that. Our second use case was the cloud. And again, Cisco used to call it MSDC, massively scaled data centers. But they didn't really pay attention to that sector as a sector of customers and try to understand why they did what they did.

21:53In fact, the early days of AWS, Amazon and Google, they didn't use Cisco. They didn't use Arista because they built their own because they were so dissatisfied with their requirements and what was commercially available. So once again, I felt like we were fulfilling a white space that the large corporate corporations, and in particular Cisco, were not doing at that time. What is an example of outstanding customer service?

22:24there are many examples and many stories i could tell you but let me actually ask you that question back if today you picked up the phone and called arista support what do you think and you you had a network problem and you had a what do you think the response or expectation from you would be well that you came and sorted out my problem well typically how it works is we'll say okay who are you? Have you paid your bills? If you're just a random asker of questions, they're like, wait a minute, we don't have an account on you. We don't do any of that. We immediately say, what's the problem? And our average resolution time, we don't, you know, it's like treating a patient in an emergency or ICU.

23:06You don't ask them 20 questions. You get to their symptoms and treat them. So our average resolution time is 25 minutes. It's an unheard of record. Why? Because we're filled with experts and our quality and support go hand in hand. You build high quality products. If somebody misconfigures them or there is a support problem, you've got to be right there. So we don't outsource this to anybody. It's a follow the sun model. It's worldwide coverage, any day, any time, any zone. And we respond and solve the problem. Some problems are solved in five minutes. Some may take hours, but the average time is very short.

23:40And it's a very high quality experience. How do you make sure you stay ahead? It takes some personal time. So I will block out think times in my calendar where I just don't do meetings because everybody wants me to do a meeting. I make a point of staying current on events and technologies. AI, as an example, I made a personal point in 2022 and 2023 to come up to speed. I wasn't paying full-time attention to it. And when the chat GPT moment hit in November of 2022, I think it was a massive, you know, sort of milestone for the tech industry. And I remember I had to give a talk at the optical conference, OFC, in January of 2023.

24:32And I could have talked about optics and log haul and short haul and DWDM and co-packaged optics and DSPs and all of that. But instead, I used that optical conference to actually talk about AI as a driver. And so to do that, I had to learn and study myself because you're talking to an audience of PhDs who probably know much more theoretically. And so I take great pride in always learning, always continuing to push the envelope both as an individual and, you know, from my peers in the industry and inside of Arista. We have a very rich organization of top technical experts. What's the furthest you can see into the future?

25:16Not that, well, I'd be wrong. I can see far enough to say some things might happen, but I would say my accuracy is best in the one to three year. After that, it's a little more blue sky. And what is going to change over the next three years? I think the way we think of AI right now is going to change dramatically. So today we are building the mainframe version of AI. We're talking about how bad are the clusters, how many millions and billions and trillions of parameters and tokens can you put in there? What does that manifest in terms of teraflops or terawatts or terabits of capacity? But in my view, the next three years, while that is important, I'm not underestimating that, I think what's going to happen is this is going to become much more of a distributed problem.

26:04We're not going to be able to fit it all into one size. And just as the mainframe move to client server and eventually distributed PCs and clients and phones, I think AI is going to take many shapes and sizes, including this one. And what are the implications of that? Significant, because miniaturization is a lot harder than saying, let me just build the biggest, baddest, everything. That means I've got to be able to translate my training models into reasoning and inference and do more with less in a lot of those use cases. because if you can just throw bandwidth at the problem and throw capacity and compute capability, then, yeah, you can process the world's largest training workloads.

26:55But if you can apply that to reasoning and extend that in a more distributed fashion all the way into your general-purpose compute and storage, I think that'll be AI to every desktop, not just AI for the biggest, baddest. Are you worried that the technology world is splitting in two here between the US and China?

27:19There is a political aspect to how we are polarized today. I wish innovation and technology could be playing an even field, but I understand we're all countries governed by different. So I am worried that for the United States to stay ahead and a step ahead, we have to be constantly innovating and constantly partnering with the best talent. I'm an example of an immigrant who came here who wasn't born and raised here. So whether that talent is in China or India or London or wherever, we need access and we need to be collaborating with the best of the minds.

28:03Let's move on to culture. How would you describe baristas' culture? In one word, do the right thing. Because it's too easy to be, especially when you're a public company, focused on the quarter, focused on something. And no doubt every company has a business model that we have to focus on and all that. But if you look past that and do the right thing, then all the intermittent things take care of itself. One of the best examples I can give you is, you know, early in our career, we had a huge quality problem with one of our chip vendors with the hardware. We worked with them. They worked with us.

28:38And I had a choice of just doing a minor patch on the software. But I went all out to explain this to our customers and say, look, you're not seeing the problem right now, but this could be a slow decay. And we want to volunteer to replace all your gear at our cost. You know, that could have nearly made us bankrupt at that time. but we chose to do the right thing. So more people remember that we did that, even though it was bad news at that time, than all the good news, right? So I think that focus on doing the right thing as a company, for our employees, as a culture, for our customers, is what I would describe as the one-liner.

29:15How do you make sure that this attitude is shared across the organization? It's not easy, especially as you get bigger. But what's beautiful is, as a management team, You know, we have pretty much been the same management team for the last 15 years. You know, we lose people occasionally due to alternate goals or financial success. But I think when you have the same management team and leaders, and obviously now we're bringing in next generation leaders, and you reinforce that culture, then that permeates throughout. And the other big thing I would say is, while we, of course, have middle management, we We make sure that the middle management plays an important role in this culture and isn't just a professional manager, but a coach player.

30:02That they're contributing with their own, you know, two days of coding. My president and CTO, Ken Duda, who's the founder of this company, he still codes. So that tells our employees that we're leading by example. We're not saying one thing and doing another. How do you keep learning? I probably don't learn enough. But as I said, one of the important, my sources of learning are many. When I go to see customers, they teach me a lot because, you know, they look at it through a different lens. So that's one aspect. I spend a lot of time with our engineers to understand what they're up to and what some of their pain points are.

30:40So they're always teaching me. Of course, today through, you know, you can go to Claude and become a software developer with great efficiency or Gemini or ChatGPT. Those tools are certainly making it easy to understand complex topics. I try not to rely on them except in my points of weakness, but that's certainly a source of learning. And my most favorite part of learning is actually reading, which I do less and less of. But that's my go-to. So, Jayshree, what do you look for when you hire? Yeah, so obviously when you hire somebody, it's very clear they need to be competent. So there's an element of aptitude and quizzing them on their capabilities.

31:23But I look for competence, character, moral values, culture, and what kind of a human being. Are you a good person? Because I want to make sure that this hire is not only capable of producing results, but can also be with us in the long run and embrace and permeate our culture. Again, back to doing the right thing. so we definitely enforce the look at the long road and so we look for like minded people in fact it's many people say oh you know if you hire somebody who's related to you is it nepotism I think there's a good chance you're going to get some really good people that way because you know the relative is good so they must be too so I'm always willing to give internships to the Arista employees extended family if we have openings of course What proportion of your employees have Indian background?

32:26I would say because we have a large India facility, it's probably around 15 to 20 percent. Can we go back to your upbringing in India? You grew up in New Delhi. So what were you dreaming about when you were five years old? Yeah. Well, yeah, I was born in London, raised in Delhi, and then have spent majority of my life in the United States. My dad was an educator. So he was dreaming about my education and making sure I do well. He started the first five IITs in India, which, as you probably know, are very well-known and high-reputation famous. So I think education was a huge part of being a middle-class family an institution.

33:13So doing well in school, you know, excelling, not just doing well, excelling in school, going to the best schools. My parents made sure they could afford the best education. So I didn't dream what I'm doing now. I dreamt more one step at a time about how do I do well in school? How do I go to the best colleges? And there was no long-term plans per se. But I I could tell that my parents had high expectations of me. Probably me myself didn't have as many high expectations of myself. What does it do to a person to have all these expectations on you? Sometimes it forces you to perform better than you would because it's a very interesting thing that anytime my dad or mom went to meet my teachers at a parent-teacher association, they'd always say Jayshree has a lot of potential but she's not using it all she's not fulfilling to her highest potential and so they'd come back obviously demanding more expecting more but I think what it taught me was you know one of my best friends in India you rank all the students so she always came first and I didn't come first and I noticed that she would work you know, about eight hours a day in terms of studying and to do well in these exams, I would work more like three to four hours a day.

34:38And to me that, you know, rich ROI, if you want to call it that, to get a rank that was not as good as hers, but still I did well, I was in the top 10%. You know, I tried to convince my parents I was playing the long game. And today it may look like I was, but at that time it was more an excuse to not study eight hours a day. You studied electrical engineering, which was not very common for women at that stage. No, not at all. So how come you ended up there? Yeah, I ended up there through elimination. I loved physics. I love chemistry. I love math. I did not like biology. I was not as good in the arts department.

35:15And I thought, what is it that can give me this intersection of problem solving? And computer science was still a pretty rare field in my era. Probably if that had been there, I would have done more of it. I had to take my computer science classes in the math department for Fortran and C++ and assembly. So it was, I took a few classes, but electrical engineering was the hardcore engineering and it's coming back. I wish more people will do it now in this modern world of more and more silicon and chip and hardware design. So I think, and I had been influenced by my father starting the IIT is that applying science to real world problems, which is what I think engineering really is, was something that excited me.

36:00Now, you're considered one of the hardest working people in technology. Just how much do you work? You know, I think when you love what you do, it doesn't feel like you're working hard or working long. You're enjoying it. Also, as a CEO, the buck stops with you. And I have a responsibility. So I would say my mom says the same thing. You work too hard, you work too hard. And I say, Mom, I love what I'm doing. So I haven't kept track of the hours because, you know, it has always been important to me to balance my work with my health and my family. But frankly, I clearly need a few more hobbies. Maybe then I'll work a little less hard.

36:42What are your hobbies? Do you have any hobbies? I used to run a lot, but I've hurt my knees. So I walk, I do a lot of hikes. Our dog is a big part of that. My family, we're all into music of some form or shape. So while I don't plan to do any public singing, it's still a lot of fun. the acoustics in a shower sound really good anybody can sound good and these days you can use enough synthesizers to sound good too so I love that, I love to read I probably have a What do you read? I read a lot of I haven't read as much recently but when I do read it's usually on long distance flights and either I will print and publish a spec and read a technical document or I will read a completely brainless novel like a John Grisham or something where I don't have to think about things because that's my downtime.

37:44But I would rather read than watch a movie. When you look at your life, what are you the most proud of? I'm most proud of my family. I would say my daughters, my husband, my extended family, they give me a lot of joy because at that point, it's not just about me. It's about, you know, recently I got an honorary doctorate degree from my alma mater, Santa Clara University. But my daughters really did a PhD and a doctorate in veterinarian science. So I'm happy for their values. I'm happy they're compassionate people, and they're happy that they pursued their path of excellence. What kind of advice do you give to young people, including your daughters?

38:30This is a lot of the advice I give because they all, they seem anxious to get to a certain point in a certain time in their career. And I say, look, pursue your path of excellence. You know, every one of us has a gift, which also means we're good at some things or very bad at some things. But if you can find the things you're good at and improve that and build upon that and become excellent at it and keep working on it, I think everybody defines their path to excellence. And it isn't always measured by money alone. It should be measured by reaching that destination of what you deem to be success.

39:10Jishri, it seems like your teachers back in India were totally wrong. It seems to me that you have absolutely used your potential to the full. Big thanks to you for keeping the world connected and all the best of luck. Thank you, Nikolai. Pleasure. Thank you for all the thoughtful questions. Thank you.