From Ghaziabad to Silicon Valley: Nikhil Kamath x Nikesh Arora | People by WTF | Ep. 11

Nikesh Arora. Oh boy, where do we begin? A boy from Ghaziabad who's now one of the highest-paid execs in the world. That's Nikesh for you. He's collected degrees like people collect Pokémon. Constantly pivoted way before it was cool. Worked at Google after landing an accidental interview with Larry Page, then at SoftBank, and now leads Palo Alto Networks, a big name in cybersecurity. Let's find out how not sticking to a lane worked in his favor. Hi, Nikesh. Thank you for doing this.

My pleasure.

I don't know where to begin. There's no fixed agenda for today, or not. There isn't one particular thing I want out. Our audience is largely the entrepreneur wannabe entrepreneur crowd in India, and we meet people who have gone down their path, been really successful, such as yourself, and try to tell them what they can learn from someone like you.

Or what? Yeah, from either doing what I did or not doing what I did.

Yeah.

Perfect. All right.

Sometimes we find not doing what someone did is as useful, if not more.

For sure.

So maybe we can begin by you telling us a bit about yourself, and we start from there.

Where would you like to start?

Childhood.

Ah, as you know, I grew up in India. My father was in the Indian Air Force. So, you know, you grow up in the Indian Air Force, you grew up with just enough means so you have a good life, but you're no different than every other person who grows up in India where you have to be resourceful, you have to be, you have to work hard to get above the crowd and sort of make something out of your life. And in that context, my parents are amazing. My father passed away a few years ago, but he worked in the Air Force for his entire life. He was a lawyer, and his job was to solve thorny legal issues both within and without. And, you know, when you live in a family like that where every day he's trying to get people to do the right thing, The sort of the— you absorb the implicit culture of the high integrity you see around you. So, you know, I have my father to thank for a life of integrity where he had to make decisions against all odds. He had to make decisions about lots of things where, you know, it was crystal clear for him, given that the role he has was always to pick the right side, that somehow that became the mantra in our family that you always had to do the right thing irrespective of the cost. And so that's sort of, you get that from your father. And my mother, you know, rare for her time, she's a master's in math and Sanskrit. So she was very along the lines of that education empowers and whatever you do, you have to be smart, you have to be well-read, and you have to do that. And she provided the nurturing element in the family. So I was very blessed. The difference is when you work in the Air Force, you move around every few years. So it brings a sense of impermanence, it brings a sense of instability perhaps. And the question is, what do you make out of it? So on the flip side, you get to adapt to many situations and you can pick up your bags and move because you've done that naturally over your life. So those are some of the, I'd say, the key building blocks of one's personality as you grow up.

Right. And how did the transition from growing up in Ghaziabad and going to school in Delhi to where you sit today. I must say, you have a very pretty office, but everybody seems really scared of you.

Ah, that's interesting. How did you pick that up?

Before you arrived, people were like, he's coming now, he's coming now, he's coming now. And they really were freaking out. Like, I've been to some secure buildings in my life, but You have some crazy amount of security going on here as well.

Well, remember, we're in the security business, so we are a trophy company to break the security defenses of.

Right.

Because as you can imagine, the security industry comes from an underpinning where it was, you know, your perception of a security person generally is that if you think back a few years ago, is that kid sitting in his parents' basement playing Call of Duty on the side, hacking into the FBI computers and showing, look, I can do it. There's no economic motivation. There's more of a trophy kill. And that sort of feeling persists in the security industry. If you're gonna go after somebody, go after the people who are entrusted with protecting everybody else. So we do have to have more security than you expect.

Has it happened?

There's a general belief that most security companies and now infrastructure software companies are constantly attacked.

Right.

Because 10, 15 years ago, the philosophy was you wanted something, you went after it, and you got it. You wanna hack into Nikhil's Life, you go hack Nikhil, you find out what he wants. Then they said, well, that's a lot of work. Why don't we just go hack into a Gmail platform or email platform, and then I can go and get whatever I want from whoever? So people started going towards what we call supply chain attacks. You went after a big piece of infrastructure, and whoever was utilizing that piece of infrastructure would then be fair game. So we joke in our industry, hacking has gone from a hobby to a profession. When you do something professionally, you do it right. So yeah, it happens a lot.

Is that because the incentives have gotten more aligned now?

Well, if you think about it, there's something to the tune of north of $10 billion that is either extorted or ransomwared or taken from individuals in some sort of Ponzi schemes, which is based on digital sort of hacking.

Every year?

Yeah. So it's a lot of money.

Yeah. But that does sound like a Earth-shattering problem, that amount of money.

Well, it's an earth-shattering problem because think about it, it's still the Wild West, right? Because there is no recourse. Think about it, it's the lowest conviction vertical in the bad actor industry. You can hack somebody from 10,000, 5,000 miles away. You can get paid in some version of cryptocurrency and you're untraceable. So it's like much better than if you've watched, remember the old Westerns you would watch and you'd had to wear your hat and carry your guns and there was a 50% probability you'll get shot by somebody. In this case, It's pretty low impact.

If Palo Alto is a $100 billion company, I would've assumed that this is a much bigger industry as well.

It's getting bigger every year. But don't forget, there's the economic incentive part of it. There is the intellectual property part of it. There's the nation-state part of it, right? We think most future wars, as you can see, the current wars that are in play are part cyber wars and part technology wars, right? People are trying to figure out the lowest cost way to create instability, chaos, and destruction of life and property without it costing them too much. And the first thing that happened in the Russia-Ukraine conflict was the entire logistics systems of Ukraine were taken down by hackers because that was the way they could sort of destabilize and immobilize the army of Ukraine.

Mm-hmm.

It wasn't a bomb that landed, it was a cyberattack. So you can see that sort of the future norm as it relates to creating instability, taking control, stealing property, both intellectual or economic, is going to be sort of eased by cyber.

I have a private equity fund, Nikesh, and I have been trying to figure out what industry might thrive in 10 years post all the disruption of today. I'm sure disruptions will continue to come by, but How does cybersecurity change? And as an investor who's looking to get allocation, if I have a broad portfolio, what percentage of that do you think should go into something like cybersecurity?

Look, if you analyze cybersecurity, it didn't exist as an industry sub-vertical of significance till about 25 years ago. If you go back, you know, I was at Banaras doing a commencement address at BHU. And I was reflecting when I went to school there in '89, we had one ICL 1904. There was nothing called an iPhone. And that computer was 5,000 times less powerful than one iPhone.

Right.

So in what was then 30 years, you've got 5,000 times the power in the thing we carry in our back pocket. But it's about 2005, I think, was when we started getting the app economy and everything started getting connected. So that was not even 20 years ago. 20 years ago, we were not worried about getting hacked because most of us were using our phones to call somebody or at best send a sort of old form text.

Right.

So if you believe the entire connectivity revolution has exploded in the last 20 years and continues to sort of grow exponentially with where we are, in our parlance, we call that the attack surface continues to expand. Now you can attack 5 billion online people around the world. You can attack every company. In the past, no company was connected to their consumers. So if you believe the attack surface continues to expand exponentially and every service becomes somewhat useless without connectivity— now your car has to be connected. If you can't connect your car, you can't get into your Waymo. And tomorrow, I'm pretty sure robotic arms will be connected and humanoids will be connected. So if the attack surface continues to expand, There is the demand function is secured. Now the question is, can you satisfy the demand function with the right technology? And that becomes sort of the cybersecurity industry. So I don't know how it'll compare to other industries, but I'm pretty sure it is a gift that'll keep on giving.

Will it come down to who has more compute and by virtue of that they can hack into something? If I had more compute than Palo Alto, for example, would it be possible for me to hack into it?

So there is that prevailing theory that when quantum comes about, it'll break every key, which is true. It'll break every key that is there today.

Can you explain how that is? I read that, but I couldn't understand because I'm a totally non-technical person.

So the way it works is when you communicate with something else and you're trying to send data in an encrypted fashion, I encrypt it, I send it to you, we decrypt it to your end so nobody can intercept it and steal it or observe it or read it. Right? Now the industry has a protocol where there is a protocol which defines how to decrypt, how to share keys. So you have the key that is required to decrypt the communication I sent you. Those keys are effectively codified in such a way that it's hard to break them, but it'll require a lot of compute to break them. But with the arrival of quantum, they will be able to break those keys in seconds or minutes as compared to days.

Because quantum has more computing power.

So much compute power, right? So if you could do that, you need a whole new set of keys, or you need more complicated, you need keys that quantum can't break, which means you need a new set of protocols to define it. So there is this prevailing theory that when quantum comes about, it'll break today's encryption. So everything will be decryptable and hence observable and hence could lead to potential hacks. Now, the sad truth is most hacks today are way less sophisticated than that. Typically, there are human beings who make errors in configuration, human beings who click on the wrong email that comes to you, human beings who leave their password on a yellow sticky on the side of the computer to make it easy for people to get into. So I think from that perspective, we have enough of a crisis right now. We don't need yet another one from a compute perspective. But I think on the flip side, the possibility of this notion of AI, which will start figuring out on the fly where the gaps, misconfigurations, or the errors are and be able to be much more intelligent at providing more real-time protection are gonna be. So I think the opportunity's gonna be more in AI-based analytics and real-time protection to start with than just really throwing more compute at the problem.

So from 3 lens, as an investor, as an individual, and speaking to the people who are starting small enterprises across the world, As an investor, if I want exposure to cybersecurity, how should I look at it? Like, what should I look out for in a company?

If you're looking at a startup category, the most likely categories which will see outsized returns are categories where a new attack vector is being born. And there are many ideas about how to secure the attack vector. And many people are experimenting. Take the current example of AI. We all talk about AI, and we were just getting our arms around ChatGPT and LLMs, and suddenly you have coding assistants, and now we're talking about agentic AI. All these things that are out there. I don't think there is a general agreement on what agentic AI means in the world.

Mm-hmm.

Right? And I was talking to somebody this morning, and I was saying, Sumi, My manifestation of agentic AI is a Waymo in San Francisco. You get in a Waymo, you let the car decide where to take you. You let the car decide when to brake, when to turn. That's giving agency to the car. Hence, it becomes agentic AI.

Right.

Now, I don't think most enterprises or human beings are ready to give an AI agent control. Simple things like, if I told you, it's easy to imagine that if you were coming to San Francisco, you'll ask your favorite AI bot or LLM or agent to find you a nice restaurant where you like to eat. It'll understand your preferences. and know who your friends are and say, I wanna eat with my 2 friends. I wanna eat a nice Indian meal. Find me a restaurant which is more like my taste and make the reservation. Now, if you trust AI to do all of that for you and show up blindly over there, that would be called giving agency to that agent to let it decide for you.

Right.

I don't think we're ready as human beings to let even that basic— to hand over that basic agency. So sort of a long way of saying that if you're building technology, Think about those scenarios. Think about a scenario that if Nikhil gave agency to an agent and somebody took over that agent, what chaos would they be able to cause?

And this will happen when AI moves from, in many ways, for a lot of people, including me, it's very question and answer right now. You're talking about when it is able to do more sequential tasks in a way.

Able to do things.

And everybody's trying to solve for that, right?

Yes. So there's 2 elements, right? There's the planning element and the doing element.

Right.

Today it does a task. Book me an airline ticket. When do you want to go? It's kind of like writing a piece of code.

Yeah.

But it says, find the best option.

Yeah.

Infer what the best option is.

Yeah.

Figure out what's most convenient to me. Figure out the time I should be flying from here. Figure out that allows me to get there at the right time based on my calendar or whatever you have here. Something is planning this. That inference engine is what everybody's trying to build for every use case.

Yeah.

That inference engine, then we connect, let's call it, to a do engine. And if you can infer and do, that becomes your agent.

Yeah.

Now the good news is that's gonna be fun. Exciting news for cybersecurity is that once that happens, I don't have to bother you. I can just take over your agent and cause chaos.

Right.

Now you can think of simple agents making restaurant reservations, which is, I'd call it non-destructive. As the worst you're gonna end up with a bad meal. You can also make it more complicated and say, you know, it can change the entire configuration of my firewall. It can change the entire configuration of the heating cooling system of Palo Alto because you have an agent which checks climate, understands how to redo the entire heat settings for the company. So there are all kinds of different examples, from control systems to industrial systems to robotic systems you can think of, which eventually will want agency. And they will become an interesting place for bad actors to try and take control because those agents will allow you to act on behalf of whoever's agent you've taken over. So if you had— so I say that these are companies who are building these solutions, they'd be very interesting because it's an unknown problem. It's a field which has not been fully explored. There is no resident expert today. There is no resident install base. So it's a sort of a blue sky problem. And startup founders are notoriously good at trying to attack those problems, solve them. Either they'll build a big company out of it or they'll get bought out by somebody who wants to solve their problem.

Yeah. I'm working with this company here while I'm here, and they were demoing their new browser yesterday.

Yes.

And if you had many tabs open, it could do a few things which seemed very sequential in nature. The Gentric browser? Yeah.

Yes.

That's almost here, right? Because that was a live product. They've not launched yet, but it seems to be getting really, really close.

Yeah, I think the idea of automated tasks in sequence has been around for a while. People have called it a playbook. People have called it a workflow. I think the question really becomes is when will you let a planning agent control it versus you control it yourself?

Interesting.

For a meaningful task. You know, you can say, I can do a 3-step task. I can search the web. I can find you, you know, which websites are selling great shoes. I can make a catalog of them. I can sort them by size. I can marry that with your preferences, which I have in memory, and sort. Those are all planned tasks. You actually wrote the plan. Something executed your plan. If I give you a nebulous problem saying, find the best shoe for me that I should wear tomorrow, there's too many subjective decisions that need to be made in that process, which means your planning or inferencing brain has to start working.

Right.

Which is creating certainty from a lot of nebulous data. I think that's the holy grail. And when you can start getting really good inferencing and planning agents, then it starts to mimic human intelligence.

So as an individual, As an individual investor, if I were to look at a new vector, a lot of people do say that self-driving cars are not really the best example of AI, at least generative learning.

Yeah. Well, look, AI was traditionally known, what used to be machine learning, right? It's just the idea that you could algorithmically write code and go decipher things.

Because they're not really predicting what will come in the road or what will happen next, right? They're more like—

Well, they are acting based on information that they're presented with in the instance. And over time, you go from predictable actions to unpredictable situations and figuring out, inferencing what you do in that scenario. So there's a little bit of inferencing going on, but there'll be more inferencing going on. And generative is a whole different conversation, right?

So that's good. That's useful. As an investor, I can look for new vectors, new surfaces where cybersecurity might be a thing and invest in that. Invest in small companies which might want to build a niche specifically in one thing.

Yeah, well, I think if you step back and look at the broader impact of AI, I think there are a lot of industries will get upended in the next 10 years. If you take the whole notion of product development, you have a trading platform of sorts, right? You're— a lot of your trading platform, in my understanding, and this is where you're the expert and I'll posit something, is you're helping us consumers interact with large data behind it.

Yeah.

You're making it easy, you're creating forms, workflow, and eventually you're consummating transactions on my behalf.

Yeah.

But the entire UI is designed to make me understand and to get the right inputs from me so you can actually understand what I'm doing, right? I could pick a certain ticker, I can choose a certain amount of shares, I can put it in there, I can buy, sell. If you have call options, I can, you know, whatever. So like a typical trading platform. If you look at the world, 75% of product development in technology is human beings teaching consumers how to interact with backend engineering databases and transactions.

Yeah.

Whether it's flight tickets, whether it's Yeah. Trading.

Yeah.

If I have a smart enough capability through a combination of generative AI and workflow automation where I can talk to this without requiring UI, which we can imagine, right? We can imagine if I have an inference engine, it can say, hey, take the position I had yesterday that I bought, sell it, take half the profits and reinvest them between $100 and $110 in Reliance or whatever your favorite stock is, if it's Robinhood, whatever you have, right? You and I can imagine how that query or that natural language, that you can literally, you're literally in your brain, you're planning the steps that are required in your product to go execute that transaction.

Right.

But it doesn't happen today. In the future, it's effectively a planning engine, planning the tasks, calling the 4 tasks saying, sell yesterday's position, take the proceeds, understand how many shares I can buy, take half then buy those many shares between $100, $110, go back to the user and say, I bought $100, 100 shares for you of this, and on the remaining cash is in your balance.

Yeah.

We can actually write the steps in our head of that transaction. The question is, do I have to code all of that and take every scenario and do it? Or in the future, there's an agentic AI or an AI interface that allows it to do it, and there's an inference engine, which means your entire process of product development at zero data is gonna change.

Yeah.

Take that and multiply that across every user experience that is out there. Take a different example. Today, a lot of applications are designed generically for everyone. It's kind of like very little user preference or knowledge. It's just kind of marginal, right? Like you a little bit understand what you know. But if you can have full memory of my preferences and you're designing an application, whether it's a nutrition app, right? You go to a nutrition app today, it tells you you have to pick a whole bunch of stuff. By the time you're done filling forms, 5 pages later, it tells you eat less carbs. Thank you very much, right? Or gives you some advice. Over time, I think applications will become applications for one instead of generic applications. So I think there's a lot of implications of AI in the technology world from a product development perspective, AI in the enterprise software world where I think analytic products will die. It'll all be products that help you do something. There's a huge debate. I mean, look at the world of advertising. The whole notion of digital advertising, which is about a $450 billion industry, is to encourage you and me to buy something or transact. People buy ads on the internet so that eventually you and I will either go trade or you and I will go buy something. But if our agents start interacting on our behalf and they start assessing what the right answers are, what is the role of advertising in the future? So there's just so many implications across the entire technology landscape. I don't think your investor brain should be worried about one particular cybersecurity vector to invest in.

Right. The interface in businesses like us, I agree with you 100%, will become irrelevant. If I could use the analogy of, say, cryptocurrency and the blockchain, I think where people like us come in is we are essentially not a company of interface, but a company with a regulatory moat.

You're what I call a system of record.

Yeah.

You have track of how many shares I own. You're a system of record. If you were to vanish tomorrow, I wouldn't know what my net worth is if I was a $0 customer. So, yeah, so you're a system of record, and system of records will stay, but the entire product development process of how I interface with the system of record will change.

System of record will stay when there is regulation behind it which mandates them to stay, or will the same apply to a technology like blockchain where there might not be a regulator mandating the system of record?

It doesn't have to be regulated, right? A system of record is almost like, I would say, something that is built over time, right? Well, either it's mandated by regulation, mandated by enterprises. I have a— I have—

you—

I'm pretty sure you and I both have payroll systems, which are systems of record. They're not mandated by regulation, but you have a payroll system. If it goes away tomorrow, you wouldn't know how much you paid someone and how much you have to pay them. The whole mode of interaction of me with your payroll system may be fundamentally different.

Yeah.

And your head of HR, and that might be different, but the system of record will stay because it's required as part of your business process to exist. Now, system of record, could also be there purely because of the fact that I have great market share. Hence, I have the record of that data somewhere, and it's too painful for you to replace that system of record, even though it's not mandated by regulation. It just happens to be that was where it is. So there is a prevailing theory that the big will become bigger because if they're smart enough and fast enough and agile enough to be able to change the mode of interaction With the system of record. And I think that's going to make a lot of enterprise software rethink how it's built, how it's run, and how it's managed.

So if AI becomes democratic, democratic in the means that everybody evolves to a point where everybody's like each other, how big does the role of a brand play in something like this? Why do I pick A over B? Is it just utilitarian or is there a brand play involved?

Well, I thought you were asking a societal question, which is a much harder question. I'll stick with the brand question. I'll stick with the brand question over here. Now, I think, look, brands are correlated to experiences and trust and perception, right? Why do you spend a lot more money for a Cartier bracelet for your wife or your girlfriend than you would for something made by your local jeweler? The raw material's the same. The cogs are pretty much the same, but you pay more for the perception, for the brand, for the experience, the trust, the loyalty. So I think the brand encompasses a lot more than the raw material from that perspective. I don't know how democratization of AI changes that. Let's take a look at it from a slightly different angle.

Mm-hmm.

You know, the way I say, this is partly the societal comment, right? When things like Google came about, if you go historically, historically information was power, right? Caste system is based on power, which is basically fundamentally based on information. The highest caste owned the information and dictated how society should behave. And you look, you know, historically a lot of power has been because of an information disadvantage or asymmetry. The internet came about, information asymmetry began to vanish, right? The guy who had the newspaper in New Delhi was no longer smarter than the guy in Varanasi next to me who got the newspaper a day late. Remember, there are places in India that got newspaper a day late. But today with the internet, there's instant access. So democratization of information allows a whole series of applications and experiences to be built. The question is, does AI democratize intelligence?

Depending on the definition of intelligence.

Well, if you listen to what Sam Altman is telling you and all these wonderful people building AI models, they're talking about AGI or artificial general where these models are gonna be able to mimic the human brain or be smarter because they have infinite memory, infinite pattern recognition. Our brains are restricted by our capacity, but this thing is not. So let's assume for a second, without taking the draconian outcome, let's just assume that it normalizes intelligence.

Mm-hmm.

And what I mean by normalizing intelligence is if I call ZeroDime, I'm sure you have a customer service team, right? There are, let's say there are 5 different people. They have different capabilities, different understanding. I could get a marginally different answer from one of those 5 people if I called depending on their intelligence if everything was not codified by your systems, which I'm guessing it's not always because there's a human involved. Now we're saying AI is gonna get so good that it will make that experience consistent on a constant basis. So it is just normalized intelligence across 5 people in your team or maybe 5,000 people in your team. So when intelligence gets normalized, what are the consequences of that? Then nobody's smarter than the other person.

Then what is the differentiation?

The differentiation is solving the unknown problem.

How do we articulate what the unknown problem is? I would assume that humans from the beginning of time have wanted to differentiate through many different means. Capitalism—

Well, why do we give Nobel Prizes? We give Nobel Prizes to people who solved unknown problems. I'm sure there's a lot more unknown problems in the world that we haven't touched or scratched the surface of, which have to be solved. There's a combination of, you know, now the question is, does human intelligence learn how to manage all this democratized intelligence? Or is AGI going to manage this democratized intelligence? You and I are just going to be drinking in the pub or hanging out at the bar and saying, geez, everything is being done by AGI and AI. So I don't know the answer to that question. But I think the bigger question is, like we saw the democratization of information, this is going to democratize intelligence. And when intelligence gets democratized, what are the business implications, societal implications of that? And nobody's smarter than the other, right? Like, you know how to put a rocket in the air? I do too. Let me talk to my pet agent. It knows the answer.

Does capital become the moat then? Incumbent people with capital?

I did say the big get bigger, right? As long as they play their cards right.

Because they have more compute power in that world?

No, I think it's not just capital. There is a debate today that all the known information that is in the public domain has already been used to train AI, right? If you hear the debates, all these models are being trained by all the information out there. I think there's 10 times more information in private domains which is not available in the public domain. All the drug discovery data which is sitting in every drug company that has made any drug in the past, it's all proprietary data, right? All the intellectual property, all the data used to build NVIDIA chips or Intel chips, all proprietary. AMD, all this is all proprietary data.

Maybe quantum computing changes that and everything comes online because everything—

I don't think quantum computing changes that. I think the question becomes, At what point in time do these private databases start being made available in training AI for specific tasks? Could I design the best chip designing model? Of course I could if I had all the data that Intel has and Nvidia has. There's no human being who has it in their head, but possibly some version of AI can understand patterns and build that knowledge base. But once it builds that knowledge base, can it use an inference engine to solve the unknown problem?

Right.

Right? Today we've relied on human intelligence to understand everything, the art of the possible. And humans have to ideate and say, well, how do I solve the unknown? Now you understand the known, but it's a high bar, right? You have to go to, I don't know how many years of PhD studies do you have to go to? How many research papers do you have to write to get to a point where you say, I think I have a good understanding of this sliver of the problem I'm trying to solve.

Right.

But tomorrow we say, guess what? That's in your edge model sitting in your pocket. It's only as smart as all the intelligence in the world that is being used to solve this particular domain. Now work from there. So I think that's the bigger debate or the bigger, opportunity, or the scary part is what happens when this thing gets smart and we start democratizing intelligence. I mean, I gave you the example of going to Varanasi and saying that was a 5,000th time less powerful computer. In 30 years, we were able to build a 5,000 times powerful computer.

Right.

Imagine if the state of AI, which is possibly or potentially moves exponentially faster. If AI is 1,000 times smarter in 5 years, let alone 5,000 times, what is the art of the possible?

I love how you're explaining. Would you like to prophesize how the world might look in 10 years? Just wager a guess.

No, I don't know if prophesizing is to any value, right? I think you have to take an optimistic outlook towards it. You can prophesize a draconian set of outcomes. You can prophesize an outcome optimistic set of outcomes. We all were, you know, we've all lived through this in recent past where we had a pandemic. And if you looked at people prophesying the pandemic, you said the pandemic is gonna persist for 5 years. I'm pretty sure there would be as many doomsday prophecies as there were possibly more than optimistic prophecies. But look around you.

Yeah.

A very extremely resilient society. I think we forget historical lessons and people will say, oh my God, what's gonna happen? AI's gonna take my job. What's it gonna do? Well, guess what? You know, I don't see anybody buying picks and shovels and digging up the ground to plant wheat, right? Things have happened and we all found something else to do. So I'm sure we'll find things to do. We'll find ways to propel society further, right? And I'm sure things that get solved easily will create new opportunities we just haven't thought of.

I was talking to one of our private equity common friends And he seemed to think that the models, the language models will become like the device, like the iPhone today. And the real marketplace is going to be what gets built on top. Do you suspect that to be true?

Yeah, I have my own version of that interpretation. I think it's broadly consistent. And the way I say it is that What you're getting in the forms of a model or AI is the equivalence of a brain, right? The early version of the brain had parroting capability. It could read everything in the world and repeat it to you.

Mm-hmm.

Over time, it built pattern recognition. Over time, it's building some judgment as based on pattern recognition, what is good, what is bad, what is right, what is wrong. Over time, as you feed it more, it's starting to get PhDs and starting to pass the bar, and it's starting to get smart enough to, you know, so it knows the entire world's information. Your problem set is known, your answer set is known. It's able to answer the questions, able to get smarter and smarter. Now, given the breakneck speed that you're seeing the evolution of models, they're getting smarter and smarter, and there's sort of an arms race, no pun intended, between all these people. And yes, so Will people build smarter and smarter and bigger and bigger brains? Yes. But if in your company at ZeroDah or at Palo Alto Networks, we said, hey, here's a really smart person, they have the best brain in the world, eventually you have to adapt that brain to the problem you're solving, right? That's the art of building something on top. That's the art of creating a wrapper around that brain and applying it to something. Could I take the smartest brain? I could.

Mm-hmm.

I have to teach it everything about drugs. And say, the problem you're trying to solve is solve cancer, solve the problem. But that requires teaching that brain, training that brain, putting guardrails around the brain, making sure the same brain that can solve cancer can also cause cancer, right? If it's smart enough. So you got to figure out you got to apply it to good. So I think the act of wrapping these models or brains with useful things that we want it to do will be where the art will be, which is what I think our mutual friend is talking about, building applications around them.

Mm-hmm. Perhaps. As an investor, better use of my dollar to be building a wrapper on a brain or trying to secure the brain?

Well, clearly the trying to secure the brain, you're asking my question, the cybersecurity guy, I'm just trying to, I'm still staying in the philosophical part of it. Okay. Look, I think every who's solving a problem, necessary or unnecessary, right? I mean, the application world, some people are solving necessary problems, some people unnecessary problems. Do I really need more social media and social applications? Possibly not, but somebody's trying to solve it. I think everybody is going to see how they can apply this brain towards solving their specific problem. And as an investor, a lot bigger share shifts will happen.

Mm-hmm.

in what is, I'll say, a stabler ecosystem for now. A lot of people will lose share, a lot of people will gain share depending how they use the brain to solve the problem, which is what happened in the last technological inflection, right? Why does Amazon exist? They used the technology inflection of the internet and connectivity to solve the problem, known problem differently, and caused a share shift between traditional retail and today, right? Why did Google exist? Because it shifted out of traditional advertising to online advertising because it used the inflection point of the internet and mobility, et cetera, to cause that to happen. So the question is, or the point is, people will use this technology inflection of democratization of intelligence to create, to solve the same problem differently. And the ones who solve it differently, faster, much more efficiently, lower friction, lower cost, better economic value, and better outcome are gonna have a disproportionate share shift in their direction from legacy players. Now, if I had a dollar for every time I heard a legacy player saying, I'm smart this time, I'll figure it out, Mm-hmm. then none of the new economy industry should exist and startups shouldn't exist. So by definition, people will build better solutions, better mousetraps. And the idea, the opportunity for investor has never been larger in that context than just focusing on securing the model. Which, look, which will not be non-lucrative, but it's almost securing the model is harder to discern which person is gonna secure the model better. It's easier possibly as an investor to say, take a human experience and say, I get it. That experience can be fundamentally transformed by applying smarter, more intelligence to it.

And securing the model might not be so interesting if so many models get to the point where they're almost the same.

Leave that to us. We'll take it. Another legacy incumbent saying, oh, we've got this problem solved.

Very interesting. So that was from an investor's lens. Me as a young person starting a company, How do I look at cybersecurity? How do I— what are the 3 things I have to do?

Broadly starting a company or just you want to start a cybersecurity company?

No, no, broadly starting a company. I'll come to starting a cybersecurity company in the end.

Look, what's fascinating is for the first time in my life, from a technology perspective, I'm seeking out young 23 or 25-year-old founders to learn from them. Normally, I'm the sage old man with white hair, and people come to me and say, "Hey, can we talk about starting a company? And how do you scale it? How do you make it bigger? How do you live a life like yours at Google, SoftBank, or Palo Alto?" Right now, I reached out to the founder of you know Windsurf, or I reached out to the founder of you know Console AI or Factory, and I'm sitting down picking their brains. This morning, I was talking to the founder of an autonomous car company, saying, "Tell me how this new world works. How do you see it? How does this AI going to impact it? How are you building a company which is so fundamentally different than mine?" Because I'm sitting there saying I might be legacy because technology is moving so fast and it's being done so differently. And I read about these companies which hit $100 million in ARR. Not that it's an achievement, but it's much faster than anybody else that's doing it. I see people who say, oh, I was watching an exchange on X between 2 people saying, oh, thank you. We built this feature in 72 hours and shipped it. It couldn't have been done between the 2 of our companies if you guys hadn't put your heart and soul to it. And oh, by the way, thanks to our coding agent. So I'm sitting there saying, The pace of innovation has fundamentally inflected. And I feel like I just don't wanna be sitting here gazing at my own navel saying, oh my God, we've got this covered. So it hasn't been more uncertain than today.

Right.

So all I can advise people who are out there is, can you teach us something? 'Cause they've got it figured out. But more so, I think the wave of technology that is coming is gonna allow people to build businesses faster, more agility, a lower number of people, and fundamentally rethink them. And if you're not doing that, if your rethink is marginal, if you're looking for a 10%, 20% improvement, don't bother because things are about to move 10x. So I think self-reflect on the idea. If your idea is not 10x worthy, you're solving the wrong problem.

I feel the exact same way. I've met a bunch of really bright kids every day for lunch and dinner the last week. And I just feel stupid. Like, I'm like, that's why I'm here for longer than I was meant to. And I want to come back because I feel like suddenly this area has become some kind of a hub.

It's fascinating. 2 or 3 years ago, and it happens every 4 or 5 years, right? When I moved here in 2009, and I just moved because I was going to be part of Google's team. And I moved here at that point in time. the buzz was social, right? You know, it was Snapchat and Facebook and YouTube was going from strength to strength and Instagram was getting bought and WhatsApp was getting bought. That was kind of the time in 2009, '10-ish timeframe. And then people say, oh my God, this wave is coming to an end. What's gonna happen? And suddenly, so the crypto wave arrived. Suddenly everything was crypto around you and there was crypto, crypto, crypto, blockchain, you know, a whole bunch of coins and a whole bunch of stuff. The SPAC wave. And again, people called the end of San Francisco and said, oh my God, it's a city you can't go to. And suddenly it's the AI wave. AI companies out there buying up real estate in San Francisco and building businesses. And if you ask the world, where is the hub of AI? It seems to be here. Every AI model that is being built with a reason, except perhaps in China, is headquartered here.

Do you think India should try and build a model? We are trying.

We should. I think the question becomes, is a CapEx question. You know, we can talk about, I mean, you're much smarter about the Indian ecosystem than I will ever be. India has not shown a propensity for large CapEx projects with unclear returns. Perhaps that's the best way to say it.

Yeah.

Right? You can't walk around and raise $50 billion and build 2 nuclear plants to fund AGI today, irrespective of who you are and irrespective of who your investors are from the Indian subcontinent.

Yeah.

And, you know, so that's where I think the constraint there is the appetite for people to deploy large amounts of capital.

I agree with you, but on the flip side, in a world which is getting increasingly fragmented on the geopolitical lens, and tariffs and all of that. The question is also, if somebody turns off a tap that we have built so much upon, what happens then?

Explain that.

Like, let's say we have used a model based out of China as the base to a lot of stuff that we have worked on, and one day they shut it down and India doesn't have its own. What happens then?

Look, there will always be the next version of the model. So even if you were using one and they didn't give you access to the new one, then you have the less smarter model because the next one's smarter, faster, better, cheaper, right? So that risk is constantly there. It's not just access to the current model that you have. One would hope that with the arms race we talked about, there's north of 10,000 models on Hugging Face, many of them open source. So at least there's art out there, prior art out there, which allows you to experiment and play with it. But I think—

But even when they open source the model, open source the weights they have used. They open source in general.

There are some models you can modify weights. Look, I understand the value of having a frontier model of your own where you can control your own destiny. It comes at a high cost today. The question is, yes, should— I mean, the unequivocal answer is yes, we should in India. The question is, who has the financial capacity and desire to go deploy tens of billions of dollars to go fund that? And where is the talent to go do that? If both those conditions are met, yes, we should. And then you'll talk about power. Where's the gigawatts needed to go fund that training of the model?

Right.

So the other option is you find models that are out there. Again, not attempting to prophesize because we agreed we weren't gonna do that. But there is hope that the current pace of development that is going on in the model world, everybody wants access to various talent pools who can leverage those models because that creates market share. So I'm hopeful that if geopolitical conditions don't get so bad that nobody's talking to each other, which hopefully will not arrive, there will always be the possibility of collaborating with a model somewhere in the world which will be willing to work in a part of the world which has 1.8 billion people, give or take. Which is 24%, 23%, 24% of the world's population, you'd want that model to be experimenting in that kind of environment with so many smart people, with such a large middle class, with so much progress ahead of it that it would be silly to ignore that market, the needs of that market from a model provider perspective.

And data of 1.5 billion people is equally appetizing.

That's right.

Very interesting. Coming back to the cybersecurity question, you told me as an investor, what niche I can look at, the verticals. You told me as an entrepreneur, if I want to start a company, what are the things I should watch out for, what kind of cybersecurity company I should try and build. The last thing I wanted to ask you on AI is somebody described it to me in this manner that they said world output is about $100 trillion. Let's assume 50% of that is in services. Even if AI is able to disrupt 10% of that, $5 trillion will justify all the money spent on infra and CapEx that has gone into the world of AI right now. If you were to have a dollar to invest into every AI company today, do you think it's overinflated? Do you think you'll make money in a decade at today's valuation?

That's a hard call, right? Now, normalizing for execution, some people will execute better than the others, right? Some people will be able to sustain better than the others. And as you know better than I do, in the world of businesses which are not making money, your survival is only guaranteed till the next fundraise.

Yeah.

Right? So if I had a dollar to give and you said, well, it'll make money in 10 years, the question is, can you sustain the fundraising through that period until you turn cash flow positive? Otherwise, you've seen companies which have gone away because they were not able to raise their next round, or the round was so dilutive that nobody wanted to deal with it. Or so, you know, all those conditions notwithstanding, I think it's a bigger question from an economic value proposition you're asking. You know, is there enough there there here in the entire industry that it justifies what is being spent today without regard for which company you invest in? Let's take the company dynamics out. Like Prima Facie, it It seems to be the prevailing wisdom, right? The fact that people are raising money, you know, the companies that raised money at $2 billion 6 months ago raising money at $4 billion right now is unprecedented. You never heard that degree of development and conviction that is happening in the market today. So clearly there is a gold rush. Now we've seen past gold rushes.

Mm-hmm.

So I'll give you the answer in 2 parts. You know, it's kind of like the, you know, overestimating the short-term, underestimating the long-term answer, which is, I think the long-term trend points towards something like you hypothesized, that there is enough opportunity to eliminate redundancy, run work, repetitive tasks, and solving the same problem multiple times, which was never leverageable. Right now you hire a consulting firm to do a project in company A, how many B, C, D, E through Z are going to pay the same amount of money to do the same project? Because there's no democratization of intelligence. Intelligence is captive and is being sold one At a time, when it's democratized, every gets the same time, which means it lowers the cost to deliver it. So that if you apply that logic, the logic suggests that yes, we can shave off a lot of a lot of inefficiency, a lot of increase economic value tremendously, which should justify this desire to invest in democratizing intelligence, which should justify the fact that once this democratized intelligence made available generally to all of us, we will find a way of creating more economic value because we'll get faster. higher quality, lower friction, and more competence/better output. Short term, it goes back to the execution questions as to which is the right one of these. Does it make sense? Is there enough economic value in the category that you're putting billions of dollars to work that it justifies the category? So I think all those questions, it could be a Robinhood moment. It could be people taking from the rich and giving to the poor.

I was having this conversation with someone last week. on how we can get India to be a lot more innovative. And we were trying to figure out why, on the scale of innovation, new ideas per se, not enough have been rooted from there in the past. We had this debate around a whole bunch of things, you know, budget for research in academia, stuff like that. But it seems to come down to one thing, that the risk capital available there, like you just said, A kid out of Stanford could get $20 million to go and build an idea.

Actually, better still, a kid dropped out of Stanford. They have a higher chance.

Yeah, I met like 2 dozen of them last night. And I was telling them, the seniors who are just like finishing college, I'm like, there has been no better time for anyone to have graduated from Stanford than you guys, because whatever you come up with, you'll get funded.

That's been true every 5 years, by the way, but okay. You're telling us we're getting old, right? Yeah.

But what do you think it is? What should India change? Is it risk capital?

Look, this debate comes up everywhere, right? For the last 25 years, I've been asked, why is there one Silicon Valley and why are there not many Silicon Valleys in the world? And you want to make Silicon Valley of India in Bangalore or Israel is a Silicon Valley of— sorry. Is there a Silicon Valley for cybersecurity, et cetera? I think it boils down to a combination of things. And I think it's hard to get the combination of things together in one place.

Mm-hmm.

Partly it's sort of things like you've identified risk capital as part talent and resources. It's part infrastructure, it's part availability. All of that is part ease of doing business. And all those things sort of fit in the mixer and you shake it and, you know, a lot of that gets in there. I think it's also partly the acceptance of failure.

Yeah.

Right? Culturally, I grew up there and you grew up there and maybe it's getting better now. Failure is not as easily accepted anywhere in the world but in Silicon Valley. There are founders here who've had companies fail and started a second company and third company. There are founders who've done something wrong, come back and started companies and the market gives them the benefit of doubt and funds them again and again and says, yes, I understand you have tremendous appetite for risk, you dream big. So it's partly that in terms of— I think a lot of it is cultural.

I agree with you 100%. I have a speech trainer who's Israeli. His name is Michael. And he keeps telling me in the Israeli society, people celebrate failure, especially in the startups and the tech companies over there. I don't know how to get people to incorporate that. How do they change?

I've bought 20 companies in the last 7 years, about also. And for the most part, they've been startups. And I'd say more than 50% of them have been from Israel. But even there, there is a different risk trade-off post-success.

Mm-hmm.

I can meet founders here, look them in the eye, a founder will tell me, well, we're building a $100 million company without blinking. And it's like a kid who dropped out 2 years ago, he's gonna build a $100 million company. Israel, they feel very lucky if they build a billion-dollar company, they're happy to sell it and say, I'm out. This is an outcome I never expected. So there's a lot of cultural stuff in there. There's a lot of cultural acceptance of failure, cultural sort of role modeling. Maybe it's in the water here. People think if I go to Silicon Valley, drink the water, it's going to be a better outcome than if you drink the water in New Delhi or Gurgaon, right? So I think there's a whole host of factors. But I think capital is definitely in today's world, especially where we are, capital is definitely a constraint. And also don't forget there's pattern recognition, right? Name the last 5 $20 billion companies that came out of the Indian ecosystem.

Yeah.

And lived to tell the story. So there's also that, like, you know, every time you turn around, it's going to happen. It doesn't happen. People say, oh my God, maybe there is more than Ice that is visible to me that holds me back from doing it.

Fair point. So you went to IIT and then MS from Boston and MBA from Northeastern.

The other way around, but same difference.

That's a lot of studying.

Not really. It's a lot easier to study in the US than it is in India. So you can study and work here at the same time.

So it was study and work at the same time?

No, the the I came here to go to business school at Northeastern. I wasn't smart enough to get into the IIMs in India. They're too competitive.

But you were at IIT, which is possibly more competitive.

Well, at that point in time, a lot less than it is today. But I think they have more of them right now than they were then. I was there with about 6 of them, and we were called ITB. We were not even IIT. We were not even a full-fledged IIT at that point in time. Yeah, I did that. Couldn't get into business school in India. I didn't even try that hard. I think I left in the middle of the CAT exam and went to watch a movie. with my girlfriend at that point in time. So maybe it was me. Came to the US, they said I had to teach computer science. I hadn't studied computer science, it was a slight problem. So I spent the summer brushing up on whatever I could learn and I taught computer science at Northeastern. I went to business school there. It was '92 when I graduated, it was a recession. It was hard to get a job. I wrote 400+ letters to people. applying for a job and, you know, like sometimes the printer ran out of ink, but I knew it was a rejection letter, but pretty much it said in the first sentence, but they forgot the signature part. So I had like 400, I still have them. I have 400 rejection letters in my life lying in my house. And one theme that was kind of persistent through all of them when I applied, 'cause you know, you know how you have like those yearbooks?

Yeah.

My MBA yearbook said most likely to end up on Wall Street. My Wall Street rejection said not enough finance. So there's a bit of a cognitive disconnect between my peers and—

Yeah.

And the street. And so that's why I got a master's in finance from BC. I felt like I had to compensate for lack of finance, and I ended up getting a CFA as well at that point in time. Yeah, but that's what caused the back-to-back 2 degrees. And I was lucky enough to— in those days, I think systems weren't as robust. I had 7 rejection letters from Fidelity before I got a job for the 8th person. So thank God they weren't comparing notes.

Does education hold the same value today? In today's world, would you spend as many years doing what you did?

You know, the older we get, the more we value education and experience. The younger we are, the more irreverent we are of experience and education. So I think—

I agree with both.

Staying with the stereotype, I should value education and experience more. You don't have to. But I think the 23-year-old kid who I met who's building a $100 billion company will tell you that I'm full of shit and all my experiences are not. And he's got more risk appetite than I do.

Outside of experiences, would you go back and do the education if it took you 8 years, 10 years, how much ever time it did take?

You know, for somebody who had never worked in the Western culture, who had never spent much time interacting with people here, I always say education is as much a social experience as it is a learning experience.

The network that you built?

No, not even the network. I mean, like, I have young kids, right? And I run into people in Silicon Valley especially where the debate is, oh, I'm homeschooling my kids because school is not smart enough to teach my kids. And you run into this all the time. People say, I gotta homeschool my kids. Now, of course, I don't know how you get through life as a parent. You have to arrange every sport game, every, like, literally, you are now the orchestrator of your child's schedule. But let's put that aside. They have this perception that my kid doesn't learn enough.

Hmm.

But then the question I ask is, when you go to school, there are 10 kids in the class or 20 kids in the class. You learn social interaction, you learn competition, you learn not getting what you ask for on a constant basis. You learn not getting attention from people. You learn different personalities, how to interact with them. Social experiment.

Yeah.

Now, if you're going to be so smart that you don't need people around you, yes, God bless you, and maybe that's a good thing for you to do. But I think 99.9% of us need people. So, to some degree, the collective learning experience is important, and it sets you up on how to live life around people, how to live life around different personalities. Now, if you spend too much of your time over there, you become an academic, and all you do is learn how to. prophesize and write papers. And so would I go back and do as much? Maybe, maybe not. But at that point in time, that was the benchmark.

If it were a social experiment, I would probably learn more or be better off with more diversity, right? So rather than going to an affluent college where everybody is the similar, you know, they have either an IQ or some kind of other thing that has brought them there, would I be not better off just going to a park and playing with different kind of kids for 8 hours a Sure, but if you had to make peace with the same people and see them every day for two years, you'd be much nicer and much more different and much more managed.

If you showed up at the park and said, "I don't like these kids. I'm not coming tomorrow," and you play with a different set because you have no skin in the game, right? Like you know, and families.

College skin in the game.

We love our families and we hate parts of our families, but we're stuck with them.

So you got to drop out of free colleges more than they drop out of paid colleges.

Didn't know that, but possibly true.

Right. So you transitioned to tech after Fidelity. That's a completely different career.

Not quite. You know, we could say IIT was tech or is tech. When I went to business school, I was teaching some version of tech. At Fidelity, actually, I worked in their technology department in finance the first 5 years I was there. So still analyzing tech, the wrong tech. When I went to Putnam, I was analyzing tech and telecom as a buy-side analyst. So I was always paying attention to the business of tech, perhaps is a better way to describe it. I was never spending time coding or writing code. If I did, maybe I'd have founded a company myself and be much smarter. But no, I was always involved in the business of tech.

Right. That's actually a novelty, which is a new thing almost in society, Not founders of companies, but executives. I think you're like the poster child of that, who have become extremely wealthy and billionaires and all of that. Do you see a difference?

Sorry, I'm trying to find where the slur is in that, but go on.

So these 2 paths of either founding a company or being an executive like yourself, I think there's a smaller crowd like you which have done exceptionally well. How do you think those journeys are different and what kind of personality type is suited for which?

That's an interesting question. Look, we live in Silicon Valley, so there was a big debate, as you may remember, maybe 12, 14, 18 months ago, Brian Chesky sparked it off, the founder mode, where somehow founders were more— a higher moral right to go run, build great businesses long term and bigger ones than possibly executives. And I think, look, at the end of the day, eventually a company is a combination of a product and a business around it. A product-only company is not a company. Eventually somebody has to show up with the economics for that product to have been worthy of being invested in. We just had this conversation about AI. $5 trillion of services will be saved, hence it makes sense to invest in all this. Now the question becomes, becomes, you know, horses for courses. What is the right time for a product impassioned individual to go ahead and build the best product in the world? As long as they can get the right capital attracted to go build the product. At what point in time does that product need to be wrapped with the right application or business model where it becomes a business that actually generates value for the buyer or the other side. Now, I think the answer is slightly different in, let's call them consumer businesses and enterprise businesses. In the consumer business, you can build a great product, you can create value, and you can find some basic fundamental economic mechanism like charge a subscription or sell advertising against it. Those are a little easier. Advertising is harder than subscription. If you're running Netflix, at the end you pay $19. You don't wanna pay $19, I'm gonna keep working the product and eventually you'll pay me $19, right? And I don't have to worry about each one of you because there's so many of you. If some leave, it's okay. There are many consumer businesses where you're fundamentally a product company with distribution and marketing, right? And then over time, your moat becomes distribution and your product sometimes doesn't stay as good, but you still own distribution, right? And there are examples of that in the world. But eventually, if you don't build a good product, even those companies begin to peter off. So let's leave it there. In that case, it perhaps makes even more sense for a founder to persist. where there's this constant need to product innovate because that drives the entire distribution and consumption of the product. In the case of enterprise, you see a different variation. And if you look at this example used of executives, and I saw a recent article somewhere, it talks about executives. I'd say 75% of those executives named there are actually enterprise CEOs.

Mm-hmm.

Not consumer CEOs, funnily enough. And the reason is because enterprise, is a combination of great product, building a business ecosystem around it, and making sure that you build an effective motion around delivering that innovation to various businesses, which is a combination of not just like— I met a founder the other day of an enterprise company. He's like, I love building the product. I hate going to customers and trying to explain to them why this is a great product. I lose patience because they don't understand why this is a great product. And then eventually I have to take them to dinner to get them to buy. And the founder doesn't like that. He wants to build great product. Business executives say this is par for the course. I gotta do all of these things and about. So look, I am as product obsessed and petrified as they come. I'm constantly challenging myself, especially my biggest, as I said to you, one of my current fears is what happens with AI. Does it make all of us obsolete in the way we run our businesses and the way we build our products and the product we build? So there's an existential crisis with AI around all those 3 dimensions for most companies in tech right now. So then the question becomes, and this goes past founder or executives to the notion of what kind of team does it take to win? Like, I tell my team, you're the best at what you do, but I reserve the right to jump in and help you do what you do. My job is not to do what you do. I can't do everything my CFO does. I can't do everything my chief product officer does. I can't do everything that my chief marketing officer does. Have I seen it? I've been around long enough. Have I seen it? Have I done it enough? Yes, I have. So I can jump in and work with them on certain things, course correct, point to the right north star, and say, we're all going there.

Yes.

So then I just need to make sure that I have paired myself with a phenomenal set of technical people who are as product savvy as they come. And sometimes you'll find great technical people, but they don't dream as big as founders. They're not bad technical people, they're equally smart. They just don't have the risk appetite. I can bring the risk appetite. So I think part of winning, and I'm sure you see that in your business, part of winning is surrounding yourself with the right team, not just the best at what they do.

Right.

But the right intermix of the people and personalities you need. And I think that's been one of the things that I've been, particularly excited about having done in my life, whether it was a team I built at Google and the people I worked with at SoftBank or people I built Apollo Networks. I think I found the formula where we can find the right set of people who complement each other and bring the right balance of conservatism, risk appetite, execution, and ideas and competence that allows us to at least try and aim for above-average outcome.

So in a way, you're saying Do you think the mark of a great leader in enterprise is hiring the best lieutenants?

Oh, I think you're oversimplifying it, but I won't take the bait. I think the mark of any winning leader is mobilizing a set of very good people around them to create great outcomes.

Right.

And it doesn't have to be just the hiring part. Hiring is not enough. You gotta make sure that they perform. Hiring is just the beginning of the journey, and you have to make sure you're constantly course correcting and adapting and making that happen. And you have to keep your eye on the prize. North Star, it kind of tends to move sometimes.

The most vivid memory or recollection I have is when you joined SoftBank, you were all over the news in India for your salary and stuff.

And in Japan as well, believe me. And in Japan as well.

How was that? How was that experience? How was Masa? How was working with him? How was it to be as popular as you were for your salary, of all things?

So that's That's par for the course, but it's not what you aspire for. It's kind of, that's a way of keeping score. People keep score that way, right? That's the way of, it's like, you know.

Do you still keep score that way?

No, it's the way people keep score. And again, it's a tough question to answer. There's no good answer to that question, right? If you say I keep score that way, then it makes it sound you're materialistic. If you say I don't care, makes you sound like you don't care.

If you were to ask me, I would've said, Yes. That question came up.

That's how people keep score. Score, again, I think being well compensated is a good outcome. Yeah. You know, it sometimes becomes more or less, it's a different issue.

We live in a capitalistic world and we all have voted for capitalism. That's how people score. Yeah.

Yeah. Although we, you know, we are constantly fighting the battle between democracy and capitalism, which sometimes end up at odds, but that's a different conversation.

Can't a democracy have capitalism? I think we have democratically voted for capitalism.

Not quite. I was with this very interesting business person who tried to run for public office and did not succeed. And I asked her, I said, what is the biggest— you asked the question about the biggest glaring memory or fact. What was the biggest difference in how you lived your entire life till you were 50, and then you tried for 5 years? How was the biggest difference? Like, she's like, you know, I'd classify my first 50 years of my life as an attempt to pursue capitalistic outcomes. The next 5 years, democratic outcomes. And she said the biggest difference in capitalism is $1, one vote. Democracy is one person, one vote.

Do you think it's $1, one vote today?

In capitalism?

Yeah.

Your shareholders decide. My largest shareholder can dictate what I should do if my 100% shareholder can fully tell me what to do. In democracy, there's no one shareholder, right? There's no 5 shareholders. There's no 5 board members who have a higher weighted voice because they represent majority of the capital.

What system do you think America of today adheres to?

Let's go back to the other conversation. I'm not a political commentator, so it's kind of like, honestly, I'd probably put my foot in my mouth unnecessarily.

So after Google And then after you joined SoftBank, how was it working with Masa? Just tell me, like, people that you've worked with, like Larry Page, Masa, if there is one thing you've learned from each of them.

You know, it's kind of— it fits the way I think about people because very often you get cliché questions like, who's your role model? Or, you know, like I say, well—

Do you believe in the concept of role model?

I believe in the concept of admiring people for something that they're really good that.

And not admiring things for other things they're not good at.

Yeah, but like, you know, our most interesting person in the world right now is possibly, you know, outside that, the— what is that guy, the Mexican guy who's the most interesting man in the world?

It's the most—

Never mind, there is, there is, you can go back and Google it. But today, the most interesting person in the world is possibly Elon Musk, right? Everybody knows about him, everybody wants to know about him. Now, you could admire him for a lot of things, but, you know, do I want to Do I wanna have 15 children with 5 different people, or do I wanna be, you know, in politics? So there are a lot of things that you can admire about them. There are some things you may not wanna be. So I think modeling your entire life around one person is a big ask.

Yeah.

I think admiring people for what they do. And when I turned 40, I made a conscious choice. I said, from now on, when I meet people, I'm gonna think about what they're really good at. Because, you know, getting to 40, I went back and thought, you know what, we have a tendency of looking through a filter and saying, well, this person's not, because they're not great at X, or they don't do X or Y well. And they realize that you actually have a mental frame of what made you succeed.

Mm-hmm.

And hence people were more like your personality or did things like you were more respectable or liked by you. And then people who did it differently were not as interesting because perhaps they didn't have the right answer 'cause they hadn't done it. But you can flip that lens and say, you know what, some people have succeeded in certain things in a way I would've never have imagined. There's something interesting. How do they do that? And you look around, you know, you're blessed and I'm blessed. We always run into people who achieve something in life or about to. Your 12 founders, I'm sure, are gonna be great in 10 years from now, or some of the people you and I mutually know are probably great at what they do.

Yeah.

So in that context, you can look at people and say, what do I admire about these people? And you mentioned people at work, but look, I remember a conversation Larry recited to me when he'd just come back after meeting Steve Jobs. I never met him, but Larry did. And Steve had called Larry and said, you know what? But in that time, Google was a small company and Apple was still a big company. And Steve told Larry, Larry, look at the model. The model is we do one thing, we do it really well. If you do one thing really well and win around the world, you're gonna win. And here was Larry, you know, busy doing Gmail, Chrome, Google Search, Maps, Local, a whole bunch of stuff. And Larry had a different point of view. He said, look, I don't believe that. I believe it's a question of capacity. If you have great people, you can motivate them, they can go build great stuff. As long as you focus on building a great product, doesn't matter if you do one or many. So Larry was product obsessed. In fact, much to my annoyance perhaps at that point in time, but I see the wisdom of his ways. When he became CEO of Google, he said, I'm gonna have 10 direct reports. There's gonna be one CFO.

Mm-hmm.

One lawyer and one business person and 7 product people. He said, Nikesh, you can do all the business stuff.

Mm-hmm.

And I was so proud. I went to my first one-on-one with Larry and said, okay, Larry, you know, we should chat. And I had this beautiful presentation ready to walk him through how we managed the 30,000 people. And he looked at me, he says, Bill Campbell tells me you're decent or good at what you do. I have to go build great products in the company, so you keep doing what you're doing. I'm sure if I had a few hours, I could help you do your job 20% better. I said, Larry, let me know when you have 2 hours. And I left. And he never found the 2 hours for me because he was busy fixing product. He was product obsessed. But I learned from him that, and he used to say that, you know, like tech companies which lose sight of great products eventually fail.

Mm-hmm.

So, and you look around you, that is true in technology. You know, companies like Sun Microsystems, Yahoo, all these companies were great tech companies at their time, their heydays, and eventually they lost sight of product, and eventually they didn't, they didn't succeed. And take Masa, you know, I say, as we grow up, we spend our life— perhaps more so in India, we talked about cultural things, right? We spend our life de-risking our children.

Explain.

The kid is born, you want to drop him at school, say, hold my hand. Every time you cross the road, you look left, you look Right. Don't eat that. Don't touch that. That's dirty. Right? Don't put that in your mouth. How many times have your parents or my parents admonished us as a mechanism of showing their love and trying to teach us the way of life in constantly allowing, sort of taking from their mental model what was risky, what could have some consequence, and trying not to have their kids do that?

Right.

And do that in life. And of course, our favorite Indian person was like, it's time you got old. It's time for you to get married and settle Right? Because we're deemed unsettled and possibly prone to high risk if you're not married. So this whole sort of cultural philosophy is to settle your kids, make sure they have a stable job, stable life, kids, and they're fully settled. That's the best risk manager in the world. It's your mother. Right? And here we're sitting and talking about taking more risk, dreaming bigger, and building great things. So with that as a frame of reference, Masa was the kid. You know, he turned 50 when I started working with him. And I'd say between the age of 50 to 60, he was sort of Benjamin Button, right? He was going, touching everything his mother said don't touch. And he was eating, putting everything in his mouth and being more and more risky and dreaming bigger—

Right.

Than anyone else I've known in my life. Forget any entrepreneur. Every entrepreneur was trying to take risk to build a stable business and get rich. Masa's the other way around. What are we gonna do today? Great, all in. What are we doing tomorrow? I'm all in. Masa has a phenomenal risk appetite. And if you look at it, that's the way he led his life. And funnily enough, he struck it 4 times.

Hmm.

You know, he built SoftBank from nothing. He built Yahoo Japan. He built He invested in Alibaba, he did Arm. All these things have worked for him. And by the way, he has risk appetite because his volatility is high. He was the richest man in the world for a few days, and then he was almost bankrupt at one point in time. So you learn that risk appetite is actually something we govern, but it is significantly influenced culturally.

Seems like a good way to If we have one life and life is meant to be experienced, masaz mein.

Yes, Nikhil, I think that's true. And you and I have the luxury to be able to say that. It doesn't work well with Maslow's hierarchy.

Explain.

Well, Maslow's hierarchy suggests that you know what it is. And like, when you find a person who has no food and no money and has to feed his family, he's not worried about taking more risk and living one life. He's looking to create that stability. So the risk appetite can change when you reach the aspiration state of your life. The point is he recognized—

Do you think AI changes that? More people move higher up the order in the hierarchy?

You're moving questions too fast. I wasn't done with the last one thinking through that one. But do I think AI changes our risk appetite? I don't think anything changes Maslow's hierarchy. In fact, I think there is a risk in the short term that AI creates more fear around Maslow's hierarchy than not.

Because?

Look, if you read everything out there, if you see the signs, it seems like repetitive tasks are gonna get easily overtaken by AI. And that's kind of its contribution to society in the next 2 to 3 years, which directly targets the early in career part of people's lives and the part where they're trying to go pay college loans in this country or where they're trying to settle down and get sort of their basic needs met. So the question is, if I can't get out of college with a loan and find a job that I'm going to be adding value on, then that probably impacts my Maslow's hierarchy more than it impacts my risk appetite.

But if there is no social mobility really in the world and Rich people stay rich people, and they stay higher up in Maslow's hierarchy. Do you think generationally people become more risk-averse, or they're able to take risk more easily? Like your kids or the kids of your kids would be freer in taking risk?

It's kind of a very interesting social experiment here. I think you can talk about predisposition, but not personality. What I mean by that is because your kids or my kids are gonna grow up in a slightly different circumstance than perhaps you and I grew up with, they're gonna have different predispositions. Like, you know, when I grew up, my father didn't have a car till I was 18, right? You know, we ate meat once a week 'cause that was the norm.

Same.

Right? So that's a different cultural upbringing than my kids. My kids don't have those constraints. They don't live With scarcity, so but the question is: It's very hard, and this is you should ask a sociologist this question as opposed to me. But it's very hard to take that and ascertain where the drive comes from. I've seen very well-to-do kids who have tremendous drive, who want to prove something. I've seen very poor kids with no drive. So there seems to be a difference of you know gene. in you which defines how you drive yourself and how you think about ambition and proving yourself. You know, I'm sure the upbringing, the factors around you have a role in finally defining what your passion is and where you go prove yourself. But I still feel that there's an inner drive that comes from something different for people, which is somewhat not 100% correlated to upbringing. Like, I have siblings. You have siblings? We're very different. We're very different people. Parents, our parents are the same. We grew up in the same house. We grew up with the same value system. We ended up with very different outcomes and very different motivations, very different ways of living our life. I don't know why that is.

So nature versus nurture.

Remember, nurture is the same. You can argue majority of the DNA is the same. Maybe one gene is more dominant than the other in certain cases. So it's minor modifications. Right.

Any fleeting thought, Nikesh? I must say, this is one of my favorite podcasts ever.

You must not have done many then.

I've loved this conversation, but any fleeting thought that you want to like leave behind?

I'll let you ask the question. That's too generic a comment. We can talk about anything you want.

Okay. Since you've given me that privilege, I give you $100. And I let you either go long or go short one sector for the next decade. Long one first.

Look, by definition, technology is lifelong long, right? If you look historically, 35% of the S&P is made up of technology companies. 30 years ago, it was none. Right? So by definition, what has happened is technology has taken over majority of sectors in life, and there's a possibility it'll continue to take majority of sectors of life. I'm sure this is gonna come bite me in the future, but on the short side, services has to be a short.

Yeah.

Right? By definition, the whole idea of services is that we deploy humans towards repetitive tasks where we sell intelligence or process with people. If if the impending opportunity in front of us is that process is going to be simplified or overtaken by AI, and intelligence is going to be democratized, then there has to be a reshaping of the services economy.

Makes sense. But thank you so much for taking the time and doing this.

My pleasure.

Hope to see you again soon.

Thank you.

Thank you. Cheers.

Bilal, you'll make sure every time you do. You made me do that. You'll just click it over here, right? All right, thanks. Nice to see you, man.

Was that fun?

That was fun, yeah.