Perplexity CEO on Chrome, AI and challenging the tech giants

0:03We're not doing a welcome to the time. That's a good point. We haven't got a welcome in there. Let's go. We can wing it because we're pros. I'm like, what does it normally say? Welcome to the Times Tech Podcast. I am Danny Orton. Danny in the Valley. And I am Katie Prescott. Katie in the City of London. Hello, Danny. How are you? I'm living the dream. How are you? The dream? San Francisco in August? Yeah, it actually, I will say it's been a super strange summer. But all of a sudden, like this week, it's like we have literally a week before the kids go back to school. And all of a sudden, it's like summer, beautiful weather, blue skies, maybe go to the pool later.

0:50Oh, dreamy. I heard one of your sons running earlier with a picture. And you did a game job of saying, that's lovely, darling. Well, you know, it's interesting. This is a tech podcast. First of all, I don't say darling to my kids because I'm way too American for that. But, interestingly, they both came in showing pictures because my son, my elder son, he's really, really into drawing. And so he watches YouTube videos, drawing videos. Oh, my daughters do that. Yeah. Yeah. And then this morning, the kids are home. I'm working. My wife's working. It's all a bit mad around the house. And I go down to make coffee.

1:24And my elder son is basically doing a drawing video for my younger son. Oh, that's so cute. Now we're going to make the mouth and now you do the arms and blah, blah, blah. So they both came up to show like basically what they drew. But it was my six-year-old drawing what my eight-year-old told them to draw. Are you filming them and making a fortune on YouTube? This is your latest side hustle. Yeah, exactly. I'm going to have them do unboxing videos and make millions. Exactly. I've had enough of vibe coding. Like that kid. What was his name? The kid Ryan. Do you know about this kid? No. this dude this little kid ryan something was making millions at age like 10 but making all this money just like unboxing toys and stuff he had like millions of followers on youtube what unboxing toys i have to look this up do you not know this this is a whole genre and whole i don't want my children to find out youtube videos yeah unboxing like the newest toy the newest video game the newest thing right our guy ryan the unboxer how much you think he's oh goodness me i don't know three million dollars 100 million dollars oh you're joking that's insane isn't that insane as often discussed we're in the wrong jobs every week that's the theme of this pod.

2:56But anyway, we digress because we have a very big show today, a small matter of perplexity. This is the AI search company that we were laughing about last week. After here we go, another massive number, their$35 billion bid for Google Chrome. That's right. They called it Project Solomon. Every good bid needs a codename. And we read out the letter and you know be honest we rolled our eyes a bit because this is a three-year-old company they're very interesting they seem to always be in the news but they're three years old they're worth according to their venture capitalists 18 billion so they're like yeah yeah we're gonna buy this thing that's at least two or three times worth what we are worth but yeah we're gonna just have it's fine because we've got some investors who'll put the money in i'm really looking forward to today because their chief executive aravant serinovas is going to join us and we can ask him whether that was yeah exactly straight from the horse's mouth straight from the horse's mouth about ai search and they've also launched their own uh web browser called comet there'll be tons to discuss but there's uh before we get to that there's some news stories happening i don't know if you've seen there's never any news in tech a couple things yeah so i think we should start briefly with the soap opera that is u.s chip u.s semiconductor policy just when you thought it couldn't get any more chaotic donald trump as he does he's on his plane yeah force one it's like yes exactly that one the old air force one he's like uh yeah i might uh i might impose some new tariffs on some uh chips maybe 200 percent 300 percent Who knows?

4:44Maybe 500%. Add a zero. Pick a number. Exactly. But he's like, off the cuff, as he does, he's like, I'm going to impose new tariffs this week. They could be two, three, whoever, you know, pick a number. And, you know, cue global freakout. Total global freakout. I'm laughing, but actually it's really, really not funny. the UK is better known for design in semiconductors than it is for manufacturing but there are a small cohort particularly of quite high spec chip companies here who are really really worried about this and I've been speaking to them over the last few weeks about what it means that everyone's been a bit reticent to say anything or do anything because they just don't know but on Monday the government sent them all a letter saying hold on to your hats essentially It's not a direct quote.

5:37Hold on to your hats. There is going to be some news. We're expecting an executive order soon about tariffs on semiconductors. It didn't put a number in the letter, but it did say there have obviously been reports that it is going to be significant. And the letter is asking the companies what they're doing about it. You know, are you changing location? Are you sitting on your hands? What are you doing? Tell us and we'll pass it on to our negotiators. So certainly it seems like Whitehall is expecting that executive order that he talked about on Air Force One on his way to Alaska to negotiate with the Russians.

6:18I mean, you know, many things on his agenda. moving swiftly um so so there is an expectation certainly here that that's coming and that will worry companies certainly the british ones who were optimistic that they might be exempt as a friendly country it doesn't sound like they will be yeah and i do think you know because there's some firms like tsmc you know which is the big maker of all of these the most complex chips taiwan semiconductor samsung they're building these mega factories in the u.s so long term that might be fine i think that's ultimately the goal of these policies but it is interesting like semiconductors very quickly have kind of gone from like yeah they're kind of important they're in computers they're in data centers yeah yeah yeah to like and this is kind of a trope but i think it's true is like oil they're like the new oil like they're this like strategically important geopolitical thing that everybody is suddenly deeply concerned about because in the world of you know the ai god machine these are the things that get that are going to make it and they also can make ai powered weapons and all kinds of stuff so everybody is suddenly like very focused yeah that's it it's the economy and the military it's been interesting speaking to businesses here as well about their dealings with the u.s administration so in i think it was 22 biden brought in the chips act and was putting 50 billion dollars behind r &d for u.s chip company or chip companies doing research and development in the u.s and also building fabs in the u.s and they're saying now it's much harder to get money out of the US government those subsidies don't seem to be as forthcoming so even though Trump's talking a good game about the importance of semiconductors and I think Intel's found the same recently it's actually harder to access that capital and that's another funny anecdote someone told me they were at the G7 semiconductor meeting which is apparently a thing I didn't get an invite well luckily because the Americans there were slightly persona non grata there was a bit of feeling amongst the other countries that you're the ones causing all the problems at the moment so you know we're going to stand over here and you can stay by yourselves well that's the other thing is trump is talking about taking a 10 stake in intel which is just another extraordinary you're like wait what yeah but speaking of we do this for you dear listener next week we are going to have Lisa Su, who is the CEO of AMD, on the pod.

9:08I'm traveling down to their headquarters at Santa Clara later this week, and we're going to talk to her about just this whole swirl in which she's involved, because of course AMD is one of the two companies, along with NVIDIA, who kind of struck this side deal where they're like, we'll give you, White House slash US government, 15 % of any earnings if we can allow to, you know, basically reopen our business and sell into China again. I can't wait for that. That's going to be fascinating. Yeah. So that should be really fun. That'll be coming up next week. So again, speaking of, you know, coming straight from the horse's mouth, we should get some really good sense of what is actually happening amidst the swirl of, you know, geopolitical wrangling by the negotiator-in-chief.

9:58The poor little semiconductor stuck in the middle. Oh, one other thing before we get to perplexity. OpenAI. This week, it's emerged over the past several days that some investors are going to buy a bunch of shares from a bunch of insiders, a bunch of employees, other investors, at a$500 billion valuation. Wow. Happy employees. $500 billion. And this is just like months ago. I think the last one was$250 billion. I was going to say, what was the, when the soft bank? $300 billion. Yeah,$300, I think. Yeah. Yeah. So like weeks, a couple months later, we're now at 500 billion. And what's really interesting is that the company is not raising a single cent.

10:42This is just people wanting to sell out. And people are so desperate to get in. But yeah, 500 billion sold and give us all the shares. But can you imagine being, and I don't know how, you know, there's certain restrictions for insiders, for employees, but can you imagine being a CEO who was there three years ago when it was worth whatever 20 billion 15 billion and now it's worth 500 and you can cash out and go and lie on a beach in thailand but i think it's really interesting they're not raising any money they're just allowing some people to cash out and because you know their their revenues are like doubling now they're i think by the end of the year the speculation is that they'll be at 20 billion dollars annual recurring revenue you know from again effectively as a standing start of zero three years ago i mean they're still losing tons of well it beats unboxing toys doesn't it i think it's even a little bit more lucrative but anyway i just wanted to i just wanted to bring that up because it feels like it may be like a moment it may not that we'll look back on back remember when they just like didn't even raise any money but just allowed a bunch of people to sell out at 500 billion yeah that was crazy but you know it is just quite it feels unstoppable almost like nvidia or bitcoin you sort of think yeah can't go any higher and just keeps going up and then it just keeps going but back to this week perplexity last week katie you actually read out their letter to google dear mr pichai we'd like to buy chrome for 34.5 billion dollars love from perplexity thank you very much yeah to which most people said yeah okay sure uh you know great sort of bob's my uncle kind of thing and at the time when i spoke to people in the industry they dismissed it as a bit of a stunt and people on social media said the boss our event had gone crazy but they did make a formal offer and they promised to keep chrome open source keep google as a default search throw three billion dollars into this company so was it trolling was it a genuine shot at a monopoly busting deal because of course chrome is potentially in play it is potentially in play of course because and maybe even by the time this podcast gets published we may have a an answer about its future so there was a judgment in the u.s that google search had a monopoly google had a certain monopoly of a search and one of the options to end that monopoly would potentially sell in chrome but certainly google has not put it up for sale yet so um it was but before we get into the interview i think we should just have a tiny bit of history about perplexity so it was founded in in 2022 by four guys from various ai businesses it deals with 30 million queries every day uh and its last funding round as you said earlier valued it at 18 billion dollars do you ever use it i mean can you describe what it's like to use for i do it's a bit like well it's a bit like chat gpt or any of these things you know where you you ask a question and it gives you an answer early on they did some really interesting things around kind of putting sourcing into their answer and be like and then suggesting follow-on questions down below So it is quite useful.

14:27And they also recently launched Comet, which is their own browser, which they built almost as like a kind of digital assistant. It can kind of finish tasks on your behalf, help you find stuff. You know, it's a glimpse potentially of the future of web search where you have this. It's kind of like an agent in a way where it'll just do stuff for you. i think the thing i would say about perplexity is you know they're also the other thing to keep in mind keep in mind is like they're being sued by a lot of different companies including our parent company i should say well subsidiaries of our parent company yeah which is kind of like par for the course basically any ai company worth its salt these days is getting sued and this is part of the very big kind of thorny question of like well what's fair use who should get paid how does this whole new model work in this world so it's all about content being used in in the engine so so subsidiaries of news corp i should say that's the who own the times they've sued perplexity ai for alleged copyright and trademark violations you say sort of one of the many lawsuits that's going on in this space at the moment but perplexity has denied them yes denied all allegations and you know so that that kind of that train will trundle trundle on you know with them and and open ai and anthropic and all these guys all the other guys but there i will say and and i think that's what i'm interested in hearing from uh irevind on is you know i think the broad idea out here people talk about moonshots the moonshot is like to unseat google which you know as a three and a half year old startup you know that's a big mountain decline but i think that's what they're going for and whether it's trolling or whatever perplexity has managed to stay in the news which i think is you know two degrees there is a there is value in that of just like remaining relevant i don't know if you remember their name also came up in the potential bidding for tiktok which again was the same thing where you're like uh you little startup are going to buy yeah it's a lot of swagger isn't it i mean that's uh that's a cocky move yeah so so anyway so but they're just a really interesting obviously super ambitious company and so it should be a fun chat i guess my first question is for you what's the future of the internet what does that look like if you say i don't know pick a number two years from now five years from now relative to our experience up to you know this year where it's like you know you google you do this you do that everybody kind of knows what that experience is so the the number one thing i would mention is The future belongs to the curious, and the internet will be shaped by them, and the internet will be enjoyed by them, those who are curious and those who have high agency.

17:22Now, the trajectory towards getting there, what exactly it means in terms of benefits of being curious, not fully clear yet. But a lot of the things we do today feel like going through a thousand cuts. It feels horrible to do a lot of tasks online today. AI was supposed to be the solution for all of this, but the only thing AI helped us so far is chat and search and summarize content out there, do some research tasks. But everything else, we're still doing the same old way. You're booking your ticket. You're picking your flight seats. You're identifying some event or concert on some obscure blog.

18:04You're moving around meetings. You're replying to emails. you're shopping with all these cluttered ads that's a lot of cognitive work currently we're doing which is not really making us smart it's actually like boring and mundane i think like if we can take all these annoying parts away from you and delegate that to an ai uh the hope is that you are really just going to do what you want you have a lot more time on your hands it's not really about giving you back time. It's more about giving you back fun you truly deserve to have. So my hope is just a less clunkier, more fun, whatever is annoying, delegation, having the power to get knowledge anytime, anywhere, and taking actions on top.

18:51That's the sort of internet that we want to help create. And we kind of call it bringing the fun back. And fun doesn't mean just playing video games or watching TV shows. Fun just means doing whatever you want like reading another person's article or listening to a podcast simply because you want to ingest the knowledge and you and you as a business what's your route to get there is it you know first stop replace google as the default kind of home page of the internet i don't think that's how you do that like like if you really are interested in creating this future replacement of google is not the objective function.

19:29It's like actually solving the hard problems along the way is the true objective function. Replacing Google is the mere side effect of that, in my opinion. For example, like why do we need to, you go back and ask, why do we need to replace Google? You replace Google because Google is clunky. That's the reason you're replacing Google, not because Google is evil or anything like that. And why is it clunky? Because it's built for a different ad economy, an economy that needs it to be clunky, right? and they cannot change that. Yeah, when you search for something, you have all the sponsored results.

20:02Like way at the bottom are the quote-unquote organic results, the actual kind of what the algorithm would otherwise put up there if there wasn't a bunch of paid ads appearing before them. Yeah, so if you go back to the origins of the internet, it was actually meant for connecting people, connecting people to information, open web access, open web. The goal was information dissemination. And I think we need to go back to those true goals. through the paradigm of AI, where it's less about just organizing information now, it's more about giving access to knowledge and expertise. That's very actionable.

20:39Like at the end of consuming answers, you are leading people to the next question, next question, and the information bandwidth experience is just another level that they can actually execute something at the end of a learning session with an AI. and they also get to be curious about many more things now that they could never get to because earlier they would spend two hours just like opening all these different tabs and piecing things together in their head. Well, so because, you know, for people who haven't used it, perplexity is like, you know, you can tell me whether this is the correct characterization in your eyes, but as effectively an AI search engine, you go on there, you ask a question, it gives you an answer, then prompts you with a bunch of other potential questions or like kind of follow-on questions.

21:23There are citations of kind of the sourcing of those answers, et cetera. Yeah. I guess the question is when you talk about Google, the clunkiness of Google, like as you said, there's a whole architecture of the internet that Google has built and Meta has monetized. So if you're talking about making the internet fun again and giving people a different experience, do you not then have to recreate the rest of the internet, the lives beneath it, the whole economy? That props us all up. Yeah, you're right. We do need to. And people are going to pay for expertise, right? People are going to pay for delegating their tasks that they don't want to be doing to an AI.

22:07And that's increasingly like the growth of the subscription model. So do you think like broad strokes, we're going to kind of move away from an ad-based economy on the internet where everything is quote unquote free, but you're just bombarded with kind of suboptimal results and you're wading through banner ads and all that stuff to a world where I pay for perplexity or a few other surfaces that basically like my digital butler, my digital concierge, my digital research assistant who helps me navigate and do stuff I don't want to do. So let me preface by saying, I think ads are not going to die entirely.

22:47It'll take a new form even in a world where you have agents. okay let me give you a simple human level example uh leave leave the ais when you have an executive or when you have some person in your team and and let's say you are you danny are a little hard to reach yeah and someone is trying to sell you a podcasting product something to edit videos or a new microphone that's much better and the only way they get your attention is they talk to your person on your team who's helping you like prepare for the podcast so they could first try to convince that person and your assistant can then come and convince you you know what danny i really heard about this amazing thing you got it right so that's one form of ads in a world where you don't actually directly interface with a brand advertiser but you trust someone to be a intermediary between yourself and that advertiser right now that intermediary is google except you never even ask for that.

23:51You don't have any contractual relationship between the Google search engine and yourself. And it doesn't deeply understand you. So all the advertisers are just bidding against that query, the query that you put, not you, Danny. It's not personalized to you. But if you and an AI agent have a pretty good understanding and it deeply understands you, your preferences, you can describe it in simple, natural language. It keeps changing over time. It's kind of like you've hired a human assistant and brands compete to get that assistant's attention, it can ignore it. It can have a separate contract with you that's under NDA.

24:29The brands cannot see it. But you might also be open to learning about certain brands. Maybe you're open to exploring new mics. So that's another mechanism in which advertisements could work in the AI world. My goal, honestly, is to usher a new era where click-based advertising completely dies. I don't want that to happen. I want it to be more preference-based, fine-grained, and really personalized. And only if the user wants it. That level of control, that would be great. Because I don't think everybody's like, I don't think fundamentally we are opposed to ads, right? Like we don't want spam and we don't want clutter and we don't want clickbait.

25:12But we do want to learn about new interesting things. so ads are aligned with the whole idea of being curious except they need to be done the right way not in a greedy manner so our goal is like let's first get the core foundation right which is making sure you can build a great personal assistant which is which is partly i mean that's manifested again correct me if i'm wrong in comet comet is your new yeah exactly is your new web browser but it also acts as it could do stuff for you right like i asked comment because i have an ability you guys sent me a code so i could download it and i'm like you know what comment um could you just reply yes to all my linkedin requests for friends and it just did it it didn't open a new page it's like yep these people have requested i've said yes i've said yes cool hey comment can you look for flights between here and LA on Southwest.

26:08And then it does all of this work in the background. It says on the side, he's like, these are the flights coming and going. I could choose these and get you all the way to the kind of, to the pay, to the payment page effectively. That's right. So what you're trying is what you're trying to do is basically become the, the new theoretically smarter gatekeeper for people's digital experience. I would just like to think about it as building your personal assistant at the end of the day. And your cognitive operating system. That's the other way we think about it. Delegate all the boring work here.

26:47Cue it. You don't even have to wait and watch it do stuff. Like just delegate and move on. Keep doing stuff you enjoy. Come back and watch. So I think that's the sort of mode that we want to get Comet to feel like. where you can actually have processes that run on your behalf as it gets more and more reliable. That's how we earn your trust. We're not earning your trust by putting no ads or all these things. We're obviously not having ads right now, but some people think privacy or no ads are the way to earn users' trust. We think the real way to earn users' trust to be really useful. Add new value.

27:23It's so useful that I will pay for it. Exactly. And if a billion people pay you to do something on a monthly basis, and you have a very big business, for example. Exactly. So think about it this way. Google has, I don't know, like 3.5, 4 billion users. Exact number is not mentioned. $200 billion in advertisement revenue. Let's assume 250 even. And that's like over 5 billion users. So whatever is the revenue, I think$100 per user, like$50 per user. That's what it is, right? Per year. Can I do that? Like maybe a billion people paying$50 a year or like$100 a year on average, it'll be a trillion dollar company and it won't have all these like ugly dependencies Google has on like pandering to the advertisers and the user and themselves and the shareholders.

28:11Like it gets very complicated for them, right? The challenge though is we've been trained for the last quarter century to get all our stuff for free on the internet. And I speak as a journalist who my industry, we're like, yay, the internet. We're going to put it all out there for free. And then realizing many years later, They're like, huh, bad for business. We need to kind of put that horse back in the barn. And it's been a very difficult, messy process that has seen a lot of companies disappear. Yeah. Okay, so number one, we should position perplexity as a business that doesn't thrive on every user paying.

28:46So when I say, if you have a billion users and you on an average make$50 per user per year, great business. But that doesn't mean the$50 needs to come from every user. or$50 needs to come only through users. Let me give you an example. If you delegated, let's take your flight or hotel example. If you delegated Comet to book a hotel for you or a flight for you, and the airlines and the hotel pays me a cutout of the transaction they got, which honestly they'd already do to human travel agents. That is the whole business of online travel agency. Booking.com makes$165 billion in gross bookings a year.

29:28And they take 15 % of all the bookings, so they make$24 billion in revenue. And they pay Google$16 billion, roughly. At least something like$8 to$15 billion in advertisement. And take the rest$8 billion for their profits. Great business. So something like booking, like hotels, start paying perplexity a cut for the transactions they get through Comet's agent. And they have to spend less on Google's ads as a result of that, because they're getting new deal flow through something else. And people like this experience more because it's more personalized and they don't have to deal with all this clunky, like old legacy UIs that are built by websites of the past.

30:09Then that's great. Like it's a win-win for everybody. And for us, it's a way to make money that's not from the user. So there are like at least four or five different ways in which you can make money right now through a product like Perplexity and Comet that we feel like getting to a point where we make$50 billion to$100 billion a year in revenue and$25 to$30 billion or maybe$50 billion even in profits at the end of the decade or maybe slightly after. I don't actually know all these timelines at this point. It's certainly possible. All we need to do to get there, the foundation, is build a great assistant at the end of the day.

30:46So that is the fabulous future. We're right now in the messy present. So the messy present includes my parent company suing you. the i think it was the biggest japanese publisher just sued you guys in the past week and the whole idea and you guys and every other ai company are being sued broadly over the same idea which is you're mining all of this content producing answers based on that content but not compensating the creators of said content and therefore you kind of if that continues you hollow out media which you know without making a judgment on it is just like you know for the media that's a bad thing right so how do you think about that when you think about you know as you're trying to kind of get to that future and where we are right now and particularly the media case so so first of all there are three things number one people want a better internet number two publishers want to get paid and number three people are already using ai to have a better internet.

31:53So all these three things are true. As for the first, people wanting a better internet, like I said, internet used to start it off as a place where people found information. Then it got cluttered with bad content, human slop, spam, ads. And that's the reason there's a lot of high friction interfaces right now. There is no joy. You've got to bring it back. And AI is definitely helping to get there because until now, AI was only accessible for the gatekeeper in terms of ranking. But now you, the user, has the AI to consume the content on the web the way you want and ignore whatever slop you don't want to see.

32:32That's true power at the hand of the user, which is great. That can help you to bring back the joy to the user, right? Fantastic. Now, publishers also need to get paid because a product like ours would not work if new content is not getting created. And so we just think that taking the old model that's built around traffic and clicks is not the right way to be ready for a newer future, where people still have the power at their own hands to decide what they want to consume and what they don't want to consume. And delegating AIs to go read for them if they want to. right so that's why we're gonna we're gonna announce in in a few days a new publisher program that's centered around like an apple news like model but in uh more friendly to both people and ai's consumer content so that's the model we want to get to where is that something akin to like spotify for example whereas like you know it's kind of you get a cut of every play or every read uh it it's not exactly like spotify it's more similar to apple news where there is a subscription called like say comet plus and we take a huge chunk of that revenue and give it to publishers so there'll be a comet plus subscription like apple news plus you pay for that and you can access content of publishers through that both as a human and delegating your ai to access that and we're going to like grandfather it by making every Perplexity Pro and Mac subscriber automatically like have Comet Plus.

34:08So that way immediate revenue gets created from the sake of the publisher's vantage point. And we're going to give away like a huge chunk of it to the publishers. So we think as we grow in our subscription business, both through the Perplexity Pro and Mac subscriptions as well as separate standalone Comet Plus subscription, we think like this can create a revenue mechanism that's already in the range of tens of millions for publishers but grows to several hundreds of millions or even billions over time for the industry is what you're saying yeah right yeah our goal is to set up a way where our business growth will make publishers also grow and and by the way we we don't want to be the only player doing this we actually want to put it out put out the structure and full transparency so that we actually want other AI companies to also do this instead of just trying to pay for tokens and train models.

35:03Because those licensing agreements are very short-term. Like the former chief scientist and co-founder of OpenAI said, Ilya Sutsky, pre-training has hit a wall. So most of the data on the internet has already been trained on and all the additional tokens that are being created right now are not as valuable anymore. right so so we need a model that rewards for consumption of content in the context of knowledge and answers and actions and that's what perplexity is trying to create it's a we are a different company and therefore we're trying to create a different model and the theory is that if this works there would be enough money being generated by those subscription or the share of that kind of subscription that it would cover kind of all of the theoretically lost ad revenue from a click revenue that you know a bunch of your free users and the whole kind of your whole ecosystem is using that's right have you talked to publishers about this have they have they been like yeah we're right on or we're like we're very we are we are very encouraged by the the the talk so far and We hope to announce the initial set of partners when we go public about this.

36:19The other question is, I think it was earlier this year or last year, you guys' name was thrown in the mix for a potential bid for TikTok. Recently, you made a bid for Google Chrome. Yes. Which is many times larger than you guys. the immediate reaction i'll say my immediate immediate reaction is like you're like trolling everybody so you're like this is a way to stay relevant to stay in the news to keep raise brand awareness but there is not a snowball's chance that you guys are actually going to be able to buy something like that is that fair or no well so two things here i mean we can we can do this for, if that was our objective, we can do this for Perplexes buying the Times or New York Times or Wall Street Journal, right?

37:13Or we're going to buy NVIDIA. You guys should buy NVIDIA. It's a good business. Exactly. So we can say all this stuff, but that's not what we're doing here. So the two bids have two different objectives, but also some similarities. There was a unique moment in time when TikTok was supposed to be divested from ByteDance at least TikTok America of course it got coupled with all the whole tariffs negotiations so beyond the point it went way beyond our pay grade so we don't have a chance there and the ruling there is that there's actually not going to be any change other than a new set of shareholders coming and owning TikTok America but the business is still going to be operated by ByteDance and why would it have helped is we actually did feel like there's a valuable distribution of users who are consuming content that's not necessarily useful.

38:13And there's a lot of fake news spreading there. And we as a company that's really grounded in accuracy and facts and utility can bring a meaningful change to a platform that's used by 100 million Americans on a daily basis. And that was the objective. to and many other people clean up tiktok yes and add legitimate value through community nodes ais better ranking that's more focused on giving power back to the user in a way that's we're doing similar to comet as for google chrome same thing again even all even after we said things like everybody else is like yeah we're also going to make a bit but our goal here is make it very clear to the to the government to to the doj and and and the judge that google's arguments that there would not be any other credible buyer are are not valid okay we're not saying hey you absolutely have to diverse chrome and like you know screw over google or anything like that that's not our point if you feel like that is the right judgment at the end to curb their monopoly power then we are ready to take it and run it and do it in a way where it's not going to have a short-term impact on their business but because we did outline that the google will remain the default search on chrome even if we run it and we did say we would invest at least three billion dollars in the open source chromium project and we did say that uh we would you know like like continue to support and maintain chrome for several years uh even though we have our own browser and and and if you if you ask like who else could have done this bit could open ai have done this yeah sure but open ai all the ip flows back to microsoft so it's like passing ip from google one monopoly to another yeah meta amazon apple all of them are like like monopolies in their own way and and and even if you leave the monopoly argument aside has anyone else launched a good ai browser so far exactly so we we felt like it was not like even though you might think it's a troll like we actually felt like i i genuinely felt like we we had the most credibility to at least place a bid now whether 34 and a half is the actual value or is it like 100 or is it 50 i don't know right like like these are different like i we went by the estimates that people have which is like 25 billion to 50 billion in that range so we picked a premium on top 25 yes it's higher than our valuation but such bids have happened in the past where a company has made a bid for a higher number than their own market cap to get something that's extremely valuable, which is 3 billion users here.

40:56It is a long shot bid. It is hard. It's going to be a low probability event to have the judge rule that manner. But we at least want them to be aware that there is a bid from another AI player who is trying to give Google genuine competition in this space. How does this all go wrong for you? And what I mean by that is we are in this crazy moment at 2025. I just did a piece last week about the$3 trillion estimate for data centers. Everybody's losing their minds. Single coders are getting paid hundreds of millions of dollars, etc., etc. I've been around long enough to know that when bubbles happen, most companies die and a few emerge from the ashes.

41:39How do you not die or maybe less dramatically become Bing? by the way bing if spun out would be worth 100 billion at least you should buy bing 100 billion for bing really well bing the search engine i'm not talking about the edge browser i think bing is supposed to make 12 billion a year in um ad revenue something like that across all people who are at least 70 years old possibly but but but but again like uh it's kind of like useless without the uh windows ecosystem but but bing's api business itself can make a billion a year in revenue so it's not it's not like it's it's considered a joke but it's not like that much of a joke as uh people make it out to be now how do we not die i think our goal is honestly to build the best personal assistant that can be built and and uh give the power that luxury that a few billionaires in the world have which is they can get anything delegated to anybody at the tips of their fingers in in the hands of everybody and anytime you've given luxury that was only accessible to a few people in the world uh to to many people and and you've done it effectively it's usually worked out pretty well it takes time obviously to achieve it but It usually works out pretty well, and that's our goal.

43:06We die if we cannot execute on this. We're not actually betting on AI progress to keep continuing in a dramatic fashion like the model labs are. Even in the case where the bubble bursts and something like OpenAI collapses entirely, where the whole$500 billion target project makes no sense, then perplexity is still fine. Open source will keep catching up. China will keep releasing open source models that are great. And that's already happening. We're already deploying a ton of open source models in our product. So our dependence on these model labs will keep reducing. The cost of serving our product will keep coming down.

43:44The amount of people using the product as a result of that can keep going up. And as we drive more reliability and value, we can monetize all the queries and use this even better. So the way we die is by completely botching up execution. Right. But you're fighting against Anthropic and OpenAI, who are, to be clear, losing tons of money, but also raising tons of money and making tons of money. Yeah. So there you said it. Like, they are raising tons of money for what purpose, right? That's the key point to consider. They're raising tons of money to go build a machine god. Because that's the only way it makes sense to keep pouring this much money into the business.

44:29it's like oh what if they actually build like five percent chance they build a god and the god is worth like trillions and trillions right so they're they're distracted but they're victims of their own like capital like they have to invest the capital back into data centers infrastructure model training runs and like continually showing like new progress benchmarks on like these obscure coding and math and competitions and like things like that and like we have no none of those distractions we can literally focus on hey can i build a great travel agent for danny can i actually like make sure this thing can go and shop reliably can i make sure like this can can actually like place uh coffee mug orders for me when i want to like open a new shop like like that's the sort of things that they don't care about and like that's actually where the value of the internet is like like really just reliably delegating research and commerce and like like yeah all these different task that no one wants to do.

45:27My theory is that you saw it with ChatGPT5. They rolled it out and it was like, oh my god, this is going to be the most amazing thing ever. And everybody's like, it's marginally better. In some ways it's worse. If that kind of striving for increasing intelligence, I feel like that's going to slow down, which then makes product more important. For OpenAIS, but then you should ask, did it make sense for them to have all this valuation in capex then right okay like they do have 700 million users that doesn't mean uh it's um it's actually that valuable right like there are products you you could argue like google has 4 billion users and um i think i think that's where like perplexity is a lot safer bet for people to make because it's it's yes it is priced high but it's not 500 billion high uh yes it it it has you know 20x fewer users in open AI but it's growing faster too right so I think my sense is that yes you're right any saturation and model progress is going to drive more focus on product layer but then what happens to all the data centers you build no one's there to use it what happens to all the researchers you hired they're just going to go for the 100 million offers given to them by Zuckerberg right like that's already happening and then the whole brand of your companies you can build new differentiated amazing models that's why you're valued at this price if you're just another product company then you're no longer worth that much your brand is no longer worth that much so i think anthropic is making a slightly different bet which is hey we're just going to go very deep into coding and developers are always having paying power they're going to keep paying for it and we're going to automate the code like like software engineer and uh i think that's a bet that I feel like even if they're not going to lose, it's a great bet.

47:21Their models, by being good at coding, are generally useful to us too on agentic tasks. So that's why I think there's more synergy and overlap there. In a year from now, let's have you back on the pod in a year, and then we can check in. But in a year from now, is there something that you think we'll be doing with agents, with AI out there in the world that seems impossible today or surprising today or unlikely today i would say that a lot of things that we repetitively do today like checking stock prices or seeing if someone's like posted an update on something or seeing if a shop is like like you know that's that's been closed for a while but needs to be open a lot of these things are maybe the answer sounds boring to a lot of people listening here and they're probably expecting me say something crazy but this is already hard to do today because we i know that because we are trying to build these things yeah i'm pretty confident one year from now like uh most of these asynchronous non-periodic tasks uh can be done by just like uh you prompting it and it'll just be taken care of for you my ai assistant will just handle sections of my life that today i have to yes and it's going to feel like the browser is your personal intelligence cloud which has processes running all the time doing your bidding.

48:47Right. And it's going to keep earning your trust by being reliable. Like go book me the Taylor Swift ticket next time she announces a concert. Like you don't even have to worry. Like you have access to my wallet, do it. I think that's the sort of thing we are trying to build. And I hope one year from now, if we talk, like we've made like a pretty significant progress. You're like obviously hunting very big game. You're talking about like kind of remaking the internet, kind of supplanting Google, like the very big things. Where does that come from? Is this always how you have been? Not at all. Like, I mean, I realized that it's easier to attract smart people by working on harder problems.

49:27That's for sure. I am genuinely attracted by anything that's making me more curious. knowledge was like like a core asset for me to grow up and and valued a lot in our in my bringing uh emphasized a lot it's the extent that my parents are prouder of my me getting a phd and like publishing a lot of good papers during my time as a researcher than you know like being a ceo or running a company and like doing interviews they're happy about it but they're more interested in me still remaining true to my academic roots. I grew up in Chennai, India. And then I came to the United States, did a PhD in the University of California, Berkeley.

50:13That's what got me into AI. And academics never rest. They always want to keep working on the next thing. And everything they did serves as the foundation for them to start building on top. So that's how I run Perplexity, too. We initially started to try to do searching over databases, like just employment records and then searching over like larger databases and then searching over the entire web and then, okay, not just search, but research agents. And then, okay, not just answers, but actions like actually executing stuff. So, I mean, obviously like I would love to build an OS, like I would love to do the phone.

50:47I would love to do all, whatever everybody wants to do. But I'm fully aware that like the browser is just the beginning today. And like, we need to take it through, for example, one year from now, I want to do the things I said that can be possible, right? And hopefully like a lot more. And it'll eventually build up towards a world where we all have a lot of time on our hands to just go learn anything we want.

51:14Wow, what a great interview. How amazing to hear from him. So there you go. Aravind Srinivas making the case perplexity may not be just trolling the giants, but actually trying to take them on. These are interesting times, have they not? They certainly are. they are crazy i mean the fact that that company went from in a year a valuation of 500 million to now 18 billion again it's just like what where does it all end when does the bubble pop and what does that look like and the idea that they would attempt to buy google chrome like google the search company that has spawned its own language you know it's um yes yeah who'd have thought it but yeah there we go uh that's why it's fun covering this industry because you have people who are just like yeah i'm gonna kind of completely upend or take over or displace the big fish in whatever industry especially these days i feel like everybody's like well AI makes everything possible I'm just going to go for it and like I said most of it won't work but if you never know what's going to happen from from one chapter to the next you do not you do not so I uh on that note speaking of because you know you always have to have a backup plan I'm gonna buy some stuff on Amazon and I'm gonna turn on my camera

52:46and I'm just gonna see what happens what are you gonna buy Danny i mean what are you what are you going to unbox next i mean this is a family podcast but there's not there's no niche for like the adult unboxing segment what's in the box what's danny got in his box this week we all need a backup plan katie i don't know what you i don't do you have one do you have a backup plan nothing is as good as this i need to get one when ai comes for our jobs because normally the journalist backup is pr but i think that um ai is already come for them exactly well ai is coming for that too right well next week will be fun um as i said i'm traveling down to scenic santa clara california um to speak with lisa sue of amd so that should be super super interesting and i'll and i'll report back on my unboxing see you next friday see you next week bye

53:45you