Summary

The Offline Network Episode 23: AI, Travel & Stablecoins (aired 2025-10-22). Guests: Gaurav Jain, Ankit Sawant, Nakul Gupta from xAI, OnArrival, Multipli. Gaurav: "And we can go into tokens as well, because right now, token economics is a very big thing, which is talked about." Gaurav: "And I think no other model was as frequently talked about if you really think about it, right?" Topics: venture capital and funding, AI and LLMs, crypto and Web3, consumer brands and D2C. The Offline Network is India's live show on startups, tech, and venture — streaming M/W/F at 4 PM IST on YouTube.

Full Transcript

Utsav Somani: All right, folks, welcome to the Monday stream of T.O.N. I hope you've had a good weekend. There were a few exciting things, fun things that have happened on the weekend. India female cricket team won the cricket World Cup. So big congrats to them. There was one other thing that came out. Dipinder Goyal's book came out. I got a copy on Blinkid and I think it's fascinating. Like just one third into the book and you can see some of his, I mean, the childhood memories and stuff and what truly played a part on shaping him into this person that he's become today. We've read all the headlines. We've read all the numbers. But a book like this truly tells you what a human being is like, right? And that, I think, is a fascinating one. And one exciting story of one third into the book. So one story that stood out was that Priyank Suroop was roommates or was he started a venture with Dipinder as well. And Dipinder was busy changing his then wife or then to be wife, Kanchan Joshi. And the second, I mean, the first venture where they were trying to digitize SMEs and launch websites for them didn't take off. So he didn't partake in the second one, which would now be known as Zomato. So fascinating. I want to complete this book across this week as well, but do pick up a copy. I think it's a fascinating story. We've all seen the share prices, headlines, all of that stuff. But I think it's always good to just step back and learn the more human side of things as well. Today, we're going to talk about AI, travel and crypto with our guests as well. So let's welcome our first guest from XAI, Gaurav Jain. Gaurav, welcome to the show.

Gaurav Jain (xAI): Hey Utsav, nice to meet you. And thanks for the invite.

Utsav Somani: No, thank you so much. I saw that you were doing an AI meetup in Delhi. So I thought you will be in same time zone, of course, traveling back from your role in SF. So I thought it'd be good for us to chat together and catch up on the show, because we've spoken about headlines, numbers, big deals and NVIDIA and OpenAI, everything that's happening in the world of AI. So I thought, why not hear from a practitioner himself? What do you do at XAI?

Gaurav Jain (xAI): Yeah, so I joined XAI in June 25. So about a few months ago, and I'm in the inference team, mostly dealing with when the models are served for real time use case for end users. So a lot of the optimizations have to go in to make it efficient for users to have a seamless experience. So yeah, and most of it, what I do right now is more like you would have seen Grok imagine a lot of the image generation and video generation use cases, which are being released in mobile phone, in the Grok app, X app, and as well as grok.com. So that's where I'm working on the inference side of things. I'm happy to dive more deeper into inference as well, if you want. But yeah, I'll let you be.

Utsav Somani: What does it mean for our listeners?

Gaurav Jain (xAI): Yeah, so in general, we train all these large models. You would have heard about a lot of these, a lot of AI compute and infrastructure spend on training these models. But once you have these, you train these models once, but then there are users all across the world who want to use these models. So how do you use them? You essentially deploy them. You load the models up into these GPUs and all kinds of TPUs and different machine learning chips. And then you create these front ends for the user front end, like chat app is one of the use cases. Now, all of your requests goes back to these systems. And then you give your user input to the model and it goes through the entire model and gives you an output, right? So that essentially is model doing one pass to generate one output token, right? And we can go into tokens as well, because right now, token economics is a very big thing, which is talked about. But yeah, at a very high level, that is what inference is. Now, you have users across the world querying the model at the same time. So you can batch a lot of them at the same, you know, batch them together, make it more efficient. A lot of the optimizations happen. It's essentially planetary scale computing. So if you don't make it efficient, you're just burning a lot of power, as well as a lot of money.

Dhruv Sharma: Gaurav, let's see if you can actually talk about this. But Grok, relative to its peers, seems like that student in class who always does well in exams, but is basically a backbencher. It's a fun model. It's got personality in how it gives out responses. Does that have anything to do with how it was trained in the first place? What training data you may have used?

Gaurav Jain (xAI): So I joined XAI in June, right? So like a few months ago. So that's one thing. But in general, like I would say that has been the notion, essentially, even when I was outside, you know, when I was working at D-Matrix before this, it used to be that Grok probably was... Like people would talk about ChatGPT way more often, right? And I think no other model was as frequently talked about if you really think about it, right? But now Grok is in the conversation. Grok is in the conversation, not only on XAP, Grok is in the conversation when you look at, imagine video generation model. We have people who are working on it day in, day out. And as for when it comes to the training data and everything, I would say at a very high level, each and every organization is using the same training data. I don't think there is really as much novelty in the training data or there was as much novelty in the training data to begin with, right? But now you can think about the advantages with Grok comes with. You have this XAP where you are continuously interacting with Grok, right? And then you have different arenas where you have deployed Grok. Imagine is one of the other use cases. And now we have Grok as an API service, which is also released and people are trying to use Grok API along with X and trying to get, you know, your primary use cases of just Twitter sentiment analysis has become much more intelligent now, right? So, yes, I think one thing to really think about it is Grok really started in 23, right? And if you look at OpenAI or even Google's DeepMind acquisition, they are more than a decade old now, almost nine to 10 years now. So the rate and the velocity with which the team has got together to build Grok and not only one component has built these various components. And I think it's insane. Sometimes I really feel like it's a place where it just keeps on moving. And the best part about this is like everyone is moving in one direction, which we believe is like to not just solve the language use case, because language use case as a chat. Yeah, people are interacting with, but they're much more complex things which you really need to deal with. You would see that people are trying to attempt Olympiads, people are trying to attempt to solve drug use case, like, you know, generating new drugs to cure different diseases which haven't been cured, video games, physical world models. So, yeah.

Dhruv Sharma: Can you give us a sense of Grok DevTools?

Gaurav Jain (xAI): That's a really good question. I mean, I do try to follow what's happening in there. But I think as far as I know, there is an API which is there, which allows, because Grok, I think, is an absolute best when you think about it from using from multi-tool use cases, right? Which is what like everyone wants to do now. When you think about things beyond chat, you have to think about how do you integrate these different parts of different components and give an end-to-end seamless user experience, right? So on that end, I know Grok has the ability to search, has the ability to use and query your GitHub repos. Similarly, look at different docs and different integrations like with what many of the other sort of model companies, foundation model companies have started coming together with. But honestly, I'm more than happy to also share the link after that to see how extensive the suite is now. It's changing every day, really. Like people are working on it day in, day out.

Utsav Somani: And you mentioned, I mean, given your technical vantage, if everyone has access to the same data, I know you merged with X and the same companies now known as, I mean, the XAI or the XCorp. I mean, of course, you get an advantage because of real-time data which is being generated on the Twitter feed or the X feed. But where is the highest level that is coming out now? Like how do you optimize this to become even better than other models?

Gaurav Jain (xAI): Yeah. So think about it from this perspective, right? Like we have Tesla and SpaceX as well, right? Like they are all part of at the end of the day. So like they're very tightly integrated, right? So you have Grok, which has been deployed into Tesla cars. People are using it day in, day out. And again, like you have this voice API, which is integrated into the Tesla system. And then you can chat with it while your car is driving, taking you from point A to point B. And so that is something which is coming in, right? And then at the same time, you have different use cases where Grok is being used for customer supported Tesla. It is being used in different places inside SpaceX. So a lot of those things are already where Grok is deployed and which are unique. You cannot think about any other, like probably Google has that advantage because Google can go and talk about like deploy their models in Waymo and something like that, which I don't think is the case today. But the fleet which Tesla has, I don't think any other car company or even a model company has. So that's one thing. And the other thing is like really thinking about the principles, right? Like I'm sure you would have heard Elon also talk about this a lot, is that like world models and like for these models to automatically or in a sense, like be able to generate these video games all to all, give a user an experience, really like really make you feel like you're living in a simulation. And so from that perspective, we get, I think we are positioned in a very good place, I would say. Really like, but yeah, a lot of those details are still something which I don't know if I can speak about or not. And I probably might also be something which I don't really know, right? So yeah.

Dhruv Sharma: Gaurav, when was Grok 4 launched? And that's the current generation, right? Yes, July. July. Okay. So just a month after you came. And I've read, like I've read this in the public domain, but I think Grok 4 was trained at the Memphis facility, like Colossus. Yes, yes. In-house, yeah. Yeah. Can you, do you know anything about Colossus? Have you been to Memphis, that facility? Can you share something interesting?

Gaurav Jain (xAI): Waiting for the trip. But I would really say that like if you, there are a few folks on Twitter who I myself follow because of like where they work at the facility and they keep on giving these, you know, sort of regular updates, even on Twitter. So they're really fun people to follow. And you really get sort of a real time update about like how the facility is coming up and how these large 100K, 200K clusters are coming up sort of in a few months time and very rapidly.

Utsav Somani: And on Twitter and LinkedIn, you've spoken about, I mean, sorry, I'm calling it Twitter, old habits die hard. You said XAI is rethinking assistant intelligence other than just generative chatbots or AI works both ways. Break it down for us. Who said this? Sorry, maybe I'm picking up the wrong commentary. All right. Anyway, so skip that. But so let's bring the conversation back home. So you held that AI happy hour in Delhi. What is your landscape overview of what's happening in the world of India?

Gaurav Jain (xAI): Yeah, unfortunately, that did not happen because like, which was very strange for me. So I reached out to a bunch of cafes. I thought it would be pretty easy to, you know, call a cafe and say that I need a space for about 20, 25 folks to meet on a Saturday evening. And they were like, no, we don't take reservations on Saturday evening. So instead, like what I started doing was set up one on one chats and I've met a bunch of people already. Yeah, I'm pretty bummed about it that I couldn't get to meet more people at the same time. So lesson learned that do it in advance and not a couple of days before when I actually want to meet. But yeah, in general, I think I do have some opinions about the Indian landscape as well. I've been meeting founders and also trying to just read more about what's happening over here. I think India is in an interesting position right now. If you think about it from the perspective of that, where are we in terms of like, because questions very often come up, where are we in terms of frontier models, right? What are we doing? Are we training our models? If you're not, then what's happening, right? So I think it's a pretty unique position like India overall, right? Because I don't think compute is one of the biggest challenges. And I don't think India is really good or in the sense like has any advantage when it comes about compute. And I think we are pretty lacking if you look at China or US. And in fact, obviously it's ramping up, but I think the ramp up is honestly too slow in my opinion. Yeah, you talk to founders and I have talked to a bunch of them. A lot of them have this opinion that, hey, there's this fund which is organized by the Indian government about that. We need to source more GPUs. But I think when you compare yourself with other countries like the US and China, they're not waiting on the government to give them funds, right? They have this investor money and they have their own research, in the sense revenue, which they are putting into AI CapEx. Look at Google, Facebook. And I really think from an Indian landscape perspective, it has to be accelerated when you look at, let's say, I think when I entered the conversation, you were talking about Zomato, right? Zomato, for that matter, is probably one of the first companies, like first startups of this new age that has gone on to YPO, right? I think that's one of the companies which I saw growing up, right? I don't remember the last company which has sort of taken, like there are very few companies in that space today. So these are the companies I think which should lead the charts. Like I think Ola tried doing that, but I think they went on a very different trajectory than they should have. But yeah, these companies should be leading the charts about like, hey, we will source these GPUs for you, you train these models on them. That's one part. But yeah, the other part is also the consumer landscape. And consumer landscape, which I certainly think many companies have now realized that India is a very hot landscape. It's like the mobile app era of 2010s when Ola had just come in and I think a couple of mobile apps had just come, MobyQuick, FreeCharge had come in and people immediately got on to them. I think we still haven't hit that inflection point where there is an AI app from India, which I really want to use, and I'll be very bummed if I'm not able to use that, right? So we are still very early in that, but India really could have certain unique use cases where it can crack the market. And I think which sort of will make it another billion dollar company for that matter.

Dhruv Sharma: Gaurav, anything to close us out with? Gaurav, what's it like to work at XAI?

Gaurav Jain (xAI): Oh, pretty fun. I mean, I think I'm learning every hour or every minute for that matter. Really interesting conversations, really interesting peers, and extremely, extremely talented people. I think one of the most talented set of people I've ever met in my life. So learning and also learning to learn as well as deliver at the same speed, if that makes sense.

Utsav Somani: Awesome. More power to you. Thank you so much for tuning in and educating us. Thank you, Gaurav.

Dhruv Sharma: Have a great rest of the trip, Gaurav. All the best.

Utsav Somani: All right, listeners. Now we'll change gears up a little bit. Let's talk about our favorite thing, I think for most people, travel. Let's welcome Ankit, who's the co-founder of OnArrival to tell us about what's happening in the world of travel. Ankit.

Ankit Sawant (Co-founder & CEO, OnArrival): Hi, what's up? Hi, Guru. Thank you for having me.

Utsav Somani: Thank you so much for dialing in.

Ankit Sawant (Co-founder & CEO, OnArrival): Yeah, I hope I'm very clear.

Utsav Somani: I mean, tell our listeners, what does OnArrival do?

Ankit Sawant (Co-founder & CEO, OnArrival): So in very simple words, OnArrival is basically trying to build an AWS for travel. I come from a travel background. I've worked almost 12 years now, adding three years, almost 15 years in travel. One of my observations was, to think about it today, one of the most costliest infrastructure in the world is AI, but anyone with a credit card can access it as a developer. But equally costlier is travel, but no developer can access it outright. It's so difficult to get APIs for flights, for APIs for hotels, trying to understand what is happening with your flight information, whether it is on time. So many things are complicated, workflows, because travel was one of the first verticals which got digitized in 1963. So it is carrying the baggage of 40, 50 years, 60 years in terms of how the technology is. And no one has been willing to solve that baggage. And I think with the advent of AI, that has made it very, very clear that whatever has happened in the last 50, 60 years in terms of selling travel online will not function over the next, I would say, decade or so. Because AI needs access to the inventory in such a short amount of time. It needs more attributes to the inventory than just pricing. It needs far more richer information that it can marry with your profile that maybe would some likes to only travel on business class or through when traveling with extended family only travels on premium economy. Small nuances that it can pick up and now fine-tune your results. So all of this, for the longest of time, was left for users to figure out in the UI. But now there is someone in between who is willing to take that decision for you. But it needs access to that information at that speed. So that's what we're trying to build. We're trying to build a new infrastructure so that anyone new who's trying to sell travel or anyone who's already selling travel can actually build a B2C experience on top of it. Like AWS is my favorite example, but also something like Stripe for travel to make it easy for developers to integrate travel.

Utsav Somani: So embedded travel, basically.

Ankit Sawant (Co-founder & CEO, OnArrival): In a way, embedded travel, yeah. Anyone can embed it in any format, anywhere.

Utsav Somani: And I mean, you and I are part of this WhatsApp group, fintech.org, and I remember you posting a comment there about how banks might be bigger than OTAs at some point in terms of travel.

Ankit Sawant (Co-founder & CEO, OnArrival): Yeah, because this is something that already happened in the US. The fastest growing OTAs are actually JP Morgan Chase, Capital One, banks who have invested significantly in bringing in, creating out, I would say, creating a travel experience, because look at your own personal consumption behavior. You basically use a credit card to buy travel somewhere because you want to optimize for points, which can be used to transfer in some loyalty program, right? So in India, largely 50 to 60% of online travel is booked through credit cards. They are the biggest driver of credit cards. Travel is the biggest category of co-branded credit cards which sells in the country. HDFC Smart Buy, which is the one which everyone uses for point redemption, that itself does almost $700-$800 million of travel transactions a year. So if it was an OTA, it would have been the fifth, sixth OTA in the country. So, but the consumer largely is sitting with the bank. The credit card is the biggest driver. And we thought that if the banking travel experience is as good as an OTA, there is no reason for someone to go to an OTA because the credit card, you're getting underwritten by the bank, you're getting a better offer from the bank. Now the portal on the bank is also as good as an OTA. So why not make them the powerhouse for selling travel?

Dhruv Sharma: Yes. This is so interesting because the adage is every company is a fintech company. What Ankit is really saying is every financial services company is in the end a travel company. But Ankit, I mean, you've obviously shared a lot in your sort of opening remarks or comments, but can you still give us the landscape of travel? Who are the actors? But more importantly, who are the actors you work with? Who are these sellers? Who are the ancillaries? Help us sort of visualize.

Ankit Sawant (Co-founder & CEO, OnArrival): So we primarily today work with any financial institution who wants to sell travel, who has a credit card in offering. It can be a four branded credit card. It can be a bank itself selling the credit card because they drive in India. So the OTA, OTM travel market, GMB wide is 15 billion. But the spends on credit cards that happen in travel are larger than 15 billion. They're almost 20-22 billion. So they are the ones who are driving consumption. So we primarily today work with them. However, we work with also expense management companies who are embedded in companies to help you file your expenses to launch corporate travel version of the products. Because once you do expense management, there is a saying in travel that it's like a triangle. You are doing expense management, then you're doing travel and then you're doing a card. So then you're doing all three from a corporate. So that is another set of customers that we work with. And the third set of customers that we're working with are some of the OTAs in outside India markets who are not that good with their tech stack. But they want to basically, I would say, like how, why would people choose to work on AWS because it is a better infrastructure. So that's the option that we give to them. And in terms of supply, typically we act as a platform. We aggregate everything that MakeMyTrip has done over the last 20 years. But the benefit today is, I think, we are able to do it much more faster because today by cofounder, every integration that we do happens through an agentic AI that is only built for doing travel integrations. It does nothing else. So we are able to do an integration maybe in a week or two, which historically used to take three months, six months, which has been trained in-house. And we are able to provide today, whoever we work with, we provide them a guarantee that our pricing will at least be better or match MakeMyTrips of the world or PureTrips of the world at this point in time.

Utsav Somani: And how do you source inventory for this to give a better price?

Ankit Sawant (Co-founder & CEO, OnArrival): In India, we try to work directly with as many suppliers as possible. So the thing is, everyone assumes the travel market is only what you see for the consumer on the top with the OTAs. But there's a huge B2B aggregator market in between that has been operating for the last decade plus. I think the biggest example is some of the companies are listed in India. There are significant companies which are outside of India. So a big portion of travel has always been done through B2B players. B2C example, every country, I would say that B2C travel is mostly oligopoly. Every country has only three or four players which are very, very strong. But B2B travel is the opposite of that. It is a huge spread. So we work with them to start off to meet the B2C expectation. And then we basically decide who do you directly work with and go deeper into that pricing and inventory mix and max.

Utsav Somani: And one interesting thing that you mentioned, agentic AI that helps you create your product faster. I mean, you've written a blog post about this also, I believe. Tell us a little bit more about this. Like, I mean, how agentic AI or even AI can come into the travel space in your company also.

Ankit Sawant (Co-founder & CEO, OnArrival): There are two types to think about integration. One is we have created an integration where we on paper guarantee the customer that you will go live in 30 days. Otherwise, you have a penalty for us. Like I'm talking about, since we start the work on integration, there are some prerequisites that they have to do. And in 30 days, your product can be launched to all of your consumers, which is a crazy amount of guarantee to give. I don't have an insight on what they will do, but I'm confident on my co-founder that he will do it. And as far as integrations go, the entire travel industry is all about integrations. You need to have better, deeper integration with sites or hotels or what is happening with your reservations and so on and so forth. These integrations have, I would say, 15 years, 20 years of trading data today. I would say that they have been there in so many forums. How do I integrate this? There are so many blogs written about this. So many things which have gone so technical. So we are able to use that to basically train the first level of integration work that any, I would say two developers or three developers have done maybe in a month to be done in three, four days. And then there are people who would then overlook that integration and then take it to maybe a 95% success in another, probably four to five days. And then the last level of testing that happens, which is, I would say that you can't, today you can't not have a person do it. So that's where the last 5-10% heavy lifting is done by the person, actually an engineer. But that whole entire cycle, when we started, we were doing it in four weeks. Then we brought it to three. Today we are doing it two. And now we're going to attempt something to be done in seven days. So that's an internal benchmark to see whether that is possible or not. But if it happens, it will be like crazy in terms of the velocity it will provide to my team because we are a small team today. We are just 20-25 people as of now. So we have been leveraging. We are making sure that I think one of the questions I ask even in non-tech roles every day is that how much of AI do you use? Which one of them you personally pay for? And if the answer is usually no, I think that then the person would not probably be very keen to use AI also as they grow.

Utsav Somani: Interesting. That's a solid hiring insights for the world that we're heading into.

Dhruv Sharma: Yeah. Ankit, if you take a step back from your role, I mean, the work you're doing at OnArrival, as someone who's been in the travel industry, what do you think is the ideal travel experience as a traveler, as a business traveler, traveling for leisure?

Ankit Sawant (Co-founder & CEO, OnArrival): So I think the biggest missing point is post booking experience. You book your flights by searching anything and everything, whatever discount you can aggregate, whatever you can do. But then what happens after? There is a huge gap between once you have bought a flight and you have a PNR to thinking about, okay, now I have to do a check-in. Now I have to figure out which seat to select. Now I have to understand whether this flight will have maybe even a power connection. Does it have an internet? How much like is it, for example, if it is Air India, whether it is a retrofitted, like a lot of these things that come in. So that entire experience is anxiety inducing even today.

Ankit Sawant (Co-founder & CEO, OnArrival): What is the probability of your flight will get canceled? Is that historically have been canceling for 50% of the time? Too many things. Then even today, there is a, I don't want to name the company, but it is a company that recently got listed. And it is in travel internationally. They have a 300 people team sitting in Philippines. Okay. Whose only job is globally when any hotel booking happens on their platform, they pick up the phone, call the front desk of the hotel and confirm whether the reservation has reached their system. Yeah. Because a lot of times the reservation would be stuck in some random integration somewhere. It will not reach the front desk. And when the customer walks in, he's like, I don't have your reservation. So those things are what are the biggest pain points today in travel. And everyone is trying to figure out how do I solve like even simple things. For example, you book an experience for which includes a pickup from your hotel. Now, where exactly do you stand in a hotel? Do I should I be in the lobby? Should I be at the curbside? Will they call me when they're at the reception? Do I get the number of the person do I call? There's so many of these small, small, small, small things.

Dhruv Sharma: So many moving parts. Yeah.

Ankit Sawant (Co-founder & CEO, OnArrival): Yeah. So what I like to say to people when I also start to when some people ask me, what exactly are you solving in travel? So my go to answer is key travel is an industry which is like, it's like you can die with thousand cuts. There are thousands of small things you have to solve to get that entire experience, right? Otherwise, you cannot be just sitting on one thing and saying, I'm just solving this. I don't care about what happens to the rest of the customer. Yeah.

Utsav Somani: Business, I think reminds me a lot of Madhu of M2P. Have you met him?

Ankit Sawant (Co-founder & CEO, OnArrival): Yeah. I've not met him. But the fund that has invested in who's our lead investor in seed, basically, they were also the seed investors of M2P.

Utsav Somani: Because first conversation with him, I'm an investor in his company as well. I mean, he just described how tough it was to digitize or sort of build around these APS for these legacy systems of these banks, right? And you've pretty much done that as well. So give us a look back into what that journey was like, I mean, these legacy systems of various travel providers and operators and B2B systems and convincing them to partner with you.

Dhruv Sharma: Can I add to that question? So you know the lore about Stripe, Pankit, which is that they basically turned what used to be the CFO's decision to a dev's decision by just making it so easy to integrate with the APS. Do you really see that happening in India when you go out and sell?

Ankit Sawant (Co-founder & CEO, OnArrival): So, sorry, can you repeat it? I just lost you for a second there.

Dhruv Sharma: Yes, I was saying, you know, the lore about Stripe that they made it so easy to integrate Stripe APIs that the developers would actually, who at the end of the day had to do the work would be like, yeah, they became the decision makers and they said, yes, we're only going to go ahead with Stripe. Do you see that happening in India?

Ankit Sawant (Co-founder & CEO, OnArrival): So what happens with us is largely, so everyone thinks that I want to reach to any customer largely in any part of the, in India, internationally. Everyone believes that I want to build our experience. Sorry, are you able to hear me?

Utsav Somani: Yeah, little bit. I think the internet is glitching.

Ankit Sawant (Co-founder & CEO, OnArrival): Yeah, just a second. I think, sorry about that.

Utsav Somani: I know you posted about you talking at top 2% whisper flow, 125 words per minute. So the system is trying to catch up with him. The model is playing.

Ankit Sawant (Co-founder & CEO, OnArrival): Yeah, yeah, sorry. So I'll come to that. So what happens is, so that text, so our 30 day guarantee in terms of integration, right? It involves a mix of, I would say, it's an SDK with some kind of flair where it basically behaves very natively. And every integration that we write from a host app to our app is basically slightly customized based on their tech stack. Okay. When I tell people that it is an embedded experience, the first thought that everyone has is this will be like a PWA opening in my app. I don't want this, right? Everyone is like, even I'm assuming when any of the apps you open and then you realize this is a PWA, it is opening up a web view inside ULA. I don't want to do this. When we actually demo the actual integration and show that people are not able to figure out when they actually went out of the app and came into the experience and I can create this in your app. That's when we had a magic moment where the CTO says, I want to rather use their tech stack, spent his entire effort into putting up engineers, they learning how to integrate, what are the nuances, what are the smaller things. So we have focused not on, I would say today, not just on a better API experience, because that is still doesn't take you to the final consumer product. I am saying, I focused on the end consumer product itself, that this is you install or you put it into the app as an SDK or whatever, we're calling it micro frontend. We've actually applied for a payment for it in terms of how the integration works. I'm hopeful that we'll get it. But the idea being that that experience can actually replace the ability what you can do with an API also. So it is slightly forward in terms of what you can achieve. And looking at that, then I think the CTOs then say that, okay, I would rather do just this because I'm able to whatever I can think of, I am able to recreate it in this. So why should I spend the effort of actually integrating the API? So they make that decision. Well, most of the business folks are largely driven from an experience point of view without the nuances of it. They understand that API experience means I control everything. But we have shown that I can do this and make you control everything. So that's where it basically helps us largely in terms of where the CTOs and the product people take the decision and tell the business folks, I don't want to spend time doing the actual integration. I will do this.

Utsav Somani: Ankit, as a final question before we let you go. So you've assembled these serious mods with your agentic AI, with new patents. And of course, I mean, digitizing these legacy systems as well. What is the scale of OnArrival right now? And where do you see this in the next five years?

Ankit Sawant (Co-founder & CEO, OnArrival): Today, I can tell you from an access point of view, today we have access to roughly 5 million credit card users through our customers. And so our goal largely is to get to maybe a 10% penetration in the market from an access to credit card point of view, which with the kind of customers that we have, which are supposed to go live over the next one month, two months, three months, we believe that we will hit that number. We will also very soon start hitting an annualized GMV of maybe in excess of 100 million. And the idea being that how do we... So the question right now is how do we reach? We are not trying to solve for how do I reach a billion dollar today, because I have a part to that. What we're trying to figure out is what it takes to become from a billion to 10 million. From an inventory point of view, from a customer point of view and what other use cases like agenting. We are actually at the end of this month, hopefully we will release a MCP, which will take care of flight bookings end to end in India. And that should... And the idea being that is that itself is a B2B product. So anyone can integrate and create that experience. So we are trying to do something in credit card travel. We're trying to do something in corporate travel. We're trying to do using the same platform. We're trying to, I would say, solve for the agenting or the MCP part of the travel. And the idea being to double down on all three use cases, because the fundamental building block of all these use cases is the platform that we have. And these are all expressions on top of it. Tomorrow, if you tell me that I want to launch an influencer-led travel where every influencer gets a URL, which is an OTA of their own, I can do that. But should I prioritize it? Probably not right now. But the idea is that a platform has a lot of ability to do a lot more things. And our customers, the kind of customers that come on board, will actually guide what use cases to build on top.

Utsav Somani: Awesome, Ankit. Thank you so much for coming on to educate us. Thank you so much and all the best.

Dhruv Sharma: Thank you, sir. YouTube travel looks exciting.

Utsav Somani: All right, listeners, let's welcome our next guest. We're going to switch gears up yet again. Nakul, who's going to educate us about the world of what's happening in the world of crypto. Nakul, welcome to the show.

Nakul Gupta (Co-founder & CEO, Multipli): Hey, guys, how are you? Congrats.

Utsav Somani: You recently announced a 20 million round as well, Pantera and Sequoia, and some really good names. What does it mean for the company and what does the company do?

Nakul Gupta (Co-founder & CEO, Multipli): Yeah, absolutely.

Utsav Somani: So, by the way, assume that we're all 10-year-olds on the show. So explain everything in the world of crypto, which can get, of course, very technical and you can get really passionate into the weeds of the whole thing. 10-year-olds on the show, including the hosts.

Nakul Gupta (Co-founder & CEO, Multipli): Think of multiply as like a smart interest engine, right, for crypto and real world assets. So we help money actually earn instead of just sitting there. So, you know, today, I think the thesis that we went about is that the billions in Bitcoin, gold, even, you know, as people are tokenizing things like Tesla stock, they're just parked, right? And the goal is we make them actually work for you and sweat those assets, right? Safely. And if you think about like FinTech, right, a lot of people got attracted. I worked in the, you know, in the space for about 10 years, you know, with PayPal, Venmo, Coinbase. And I think yield was at the center of it, right? Why people came to DeFi in the first place or went away from their banks like JP Morgan Chase, because FinTech started providing them a good yield, right? And essentially that's what the wedge is with DeFi as well, where essentially what we're doing is we are kind of taking these assets that are coming on chain in a big way. It started with Stable and Bitcoin. Now gold is coming on chain, tokenized assets are coming on chain. We're putting some wrappers around it and providing that yield, which is one of the main reasons why people should come on chain instead of just, you know, kind of accessing those assets traditionally.

Utsav Somani: And what does DeFi mean for our listeners?

Nakul Gupta (Co-founder & CEO, Multipli): Yeah, so DeFi is nothing but decentralized finance, right? What it really means is that traditionally we have had guardians who have taken care of our financial world, whether it's banks, custodians. Now, I think money has gone permissionless. And what I say permissionless is you can store your own money in your own wallets, right? And there's no government, there is no custodian, there's no bank that can take it away from you, right? And that essentially is at the core of DeFi. So it creates a lot of permissionless markets with smart contracts. Essentially what smart contracts are, you can just kind of deploy something and trust in the code, right? Instead of trust me, bro, with the financial institutions, you can say, hey, let the code do its magic. And that creates a lot of markets on top of money movement. So lending, borrowing, you know, what kind of Goldman Sachs did for the TradFi world, a lot of players in the crypto world are doing for DeFi and creating those kind of on-chain markets because now money is truly, you know, permissionless.

Dhruv Sharma: Nakul, your customers, are they mostly individuals? Are they mostly institutions?

Nakul Gupta (Co-founder & CEO, Multipli): So we're focused mostly on institutions, right? Because those are the ones who kind of understand crypto can move in size. So we're trying to target the family offices, the high frequency traders, a little bit more crypto native folks first before we go more wider.

Dhruv Sharma: And when they come to you to maximize yield, is it typically across a single asset, a bunch of assets or an entire portfolio?

Nakul Gupta (Co-founder & CEO, Multipli): So essentially, they start small, right? They dip their feet with stables, like everyone's got stable coins or USDC kind of lying around. So they start with that. But our real pitch here is actually getting them more and more involved in other tokenized assets, right? So one of the largest gold kind of reserve holders, you know, out of India, like they actually want to, they're sitting on a ton of gold and that's generating them no yield. So as Paxos, you know, brought kind of tokenized gold on chain, we are helping them with our partners, tokenize that gold, bring it on chain and use it as collateral for actually unlocking yield on it, right? What that means is that now these idle assets in the real world with these institutional partners, we can unlock them in this crypto world.

Utsav Somani: And stable coins have been top of the town, right? A lot of people are investing in it. A lot of people are talking about it. A lot of countries are looking into it. I mean, they've been existing for a while. What's the landscape like? What's the history like of the world of stable coins?

Nakul Gupta (Co-founder & CEO, Multipli): So stable coin is nothing but a digital dollar, right? It's the easiest way for someone to understand and onboard. I mean, with UPI or a card, your money hops through banks and fees, right? With stable coin, it goes directly. So there's no middlemen, there's no delays. So you can send like a thousand dollars from India to Singapore in like 5 or 10 seconds and maybe even earn interest while it's traveling, right? So I think the way I think about stable coin is nothing but UPI, but for the planet. You know, when I was in PayPal and other places, you know, everyone was busy building walled gardens, right? They were trying to move money from Venmo to Venmo. And they spent, I think, about four years trying to create rails that can move money from Venmo to PayPal, their own system, right? So this is what traditional fintech has been about, right? Because it makes sense, because the earlier Tractify system was even more broken with Zelle. Now, stables has truly liberated us, right? What this means is that you can move money around, whether it starts with retail, sending it to your family and friends without going through Western Union, it's like 6% interest, whether it is B2B for exporters, right? Who are today not able to do that trade because it takes two months for kind of the Swift or something else to get approved or gets caught up in compliances. And then at a larger scale as a hedge against inflation, if you're in Argentina, one of these countries, by the time you're trying to buy bread, you have to take a truckload of money. So I think all of these things, I think over time, stable coins is maybe a really elegant like design and answer for that. Has the Genius Act really propelled them, legitimized them? Not really. I think Genius Act has put maybe more of a regulatory framework that don't mess around by issuing stables that essentially can be rug pulled, right? So more than anything else, it's created this kind of more regulatory clarity around who the issuers will be. Who can actually be participating in the economy? Because as we see, there has been a Cambrian implosion of stable coin, right? Everyone and their mother wants a stable coin. Stripe, Coinbase, everyone kind of launched their own stable coin. And there'll be many companies who continue to do so. Now, the Genius Act introduces a way they can do that without many of these rug pulling people.

Utsav Somani: Some more protection, but there was an A16C report that came out which said crypto went mainstream in 2025. When does that moment happen in India? I mean, do we move up the chain in terms of becoming actual users and consumers of this or will all of our talent just move away because of regulations in India?

Nakul Gupta (Co-founder & CEO, Multipli): Yeah, I think India is in a tricky spot, right? Because the reality is like it doesn't make sense for them to open up crypto just because there will be a flight of money outside. And India's historically loved to kind of protect money movement, right? With all the FEMA and the RBI laws, which makes sense. Now, I think it's a double-edged sword, right? With India, I think they have to open up crypto in a very kind of slow and gradual way and understand crypto. I'm not saying like do CBDCs or something like that because the government is not equipped to do that. Payments as a use case is already solved in India, right? So clearly like crypto is not used for payments. But if you look at the broader participation, like for example, even mainstream use cases like polymarkets, right? Today, a lot of people are betting on the elections and things globally through polymarkets and no one even knows that crypto is powering the rails, right? If you do a survey, 80% people don't know there's crypto behind polymarket. So I think a lot of these things like globally, whether it's gambling, whether it's prediction markets, I think there are a lot of things that India can unlock. But at the same time, I think India has to take it more gradually because of the ability for people to actually, you know, misuse that. So I think India will adopt it, but in a very cautious, like kind of gradual way.

Dhruv Sharma: Sorry. I was wondering if now that Nakul is here, if we could do what maybe a two or three minute capsule on what it means to tokenize the real world asset. We hear this all the time, but in the simplest terms possible, Nakul, help all of us understand.

Utsav Somani: Like assume that Dhruv has one kg of gold at home. How does he tokenize that and earn a yield on that? Let me look for it. It's right here.

Nakul Gupta (Co-founder & CEO, Multipli): So maybe gold is a harder example because I'll...

Dhruv Sharma: Or private stock.

Nakul Gupta (Co-founder & CEO, Multipli): Let's talk about the easiest thing, right? Let's talk about T-bills. I mean, the US treasury T-bills are the gateway to tokenization right now because they're like trillions of tokenized like... And what are T-bills? These are treasury bills of 4% like yield typically that the US holds. Now, how the space started getting disrupted initially is because as a user outside of the US, like even in the US, by the way, I tried to onboard to treasury bills to, you know, my JP Morgan provides me a 0.1% yield. So I do need that 4% yield for my dollar. Now, how do I access that is the main question, right? What tokenization is all about. If I were to access that as a US consumer, I would have to onboard to an E-Trade account. And like it took me like a week or two to figure out where the exact button is and how to buy these T-bills. And it was not just complicated, but also it took a few weeks, even as a savvy user. Now, if you're outside the US, those T-bills are really hard to access, right? In terms of compliance, onboarding, in terms of everything, right? Now, what does tokenization do? So the companies like Ondo, which is now a $10 billion company, the companies, you know, many such tokenization companies. And what did they do? They essentially created a structure like an SPV or, you know, more regulated structure that can take in so-called investors, right? And what they did was they wrapped the tokenized, they wrapped the T-bills by saying, I will create the infrastructure where someone can come in and buy even $100 of a T-bill. I will aggregate all that demand and then go and create my own institutional account where I can buy the million dollars worth of treasury bills and distribute that through tokens to my holders, right? So something as simple as that, where I am someone sitting outside the US, I can go on this website, like just swap out my stables for $100 in treasury bills. And underneath it is this company, which is aggregating, tokenizing. And what I mean by tokenizing, it's doing nothing but very traditional work, right? Just create an account, aggregating that demand and issuing tokens on-chain, right? Which means there's a one-to-one dollar peg kind of, you know, very similar to stables, but essentially they created the infrastructure behind it, right? And herein lies the trust that, you know, these acts and regulations will come in and enforce so that, you know, there's no rug pull. But at the core of it, tokenization is nothing but simplification of A, a legal structure and B, allowing people to get one-to-one backed assets so they can hold it in their wallet and continue to earn interest on it, right? And now you can take that analogy and blow it up for gold and any other asset where there's a custodian at the end of the day, like, you know, doing that structure while issuing these tokens that represent that, you know, physical asset or financial asset.

Utsav Somani: And the world of crypto, I think, moves very fast, right? I think, I mean, NFTs were all the rage during the pandemic where everyone was buying boards, apes, monkeys, all pictures, all JPEGs stored somewhere else and this thing. And the hype failed to live up, right? And what do you think might be the next, or is the current like sort of pockets of excitement that's happening in the world of crypto, which might actually sustain as a trend?

Nakul Gupta (Co-founder & CEO, Multipli): Yeah, crypto is like any other industry, right? We'll see, like, if you also saw the A16Z report, like we're going through a 10-year cycle where, you know, many of these use cases like NFTs, it turns out that there's something really good and we borrow the tech, right? And the actual use case was just like, you know, just speculative in nature. But I see like the thing that's sweeping the world today is actually payments and programmability, right? So what people are doing is they're moving money very easily from point A to point B, right? And institutions are doing that in a very big way, especially around B2B, like cross-border payments. I think that has come about to be really easy because the big boys, right, the stripes of the world and Adyans and everyone's gotten in. So they've realized that this is a better way to move money. So I think consumers will see that very soon, whether it's an exporter or whether it's, you know, you doing cross-border remittance. I think that's the simplest way to understand kind of, you know, what's happening. The other use cases, I would say, are like I just spoke about polymarkets and prediction markets, right? I think betting, like essentially think about like there used to be a ton of these like local players, regulated, local TAMs. With crypto, everything becomes a global TAM, right? Like if you can imagine like a billion people, you know, which is maybe the penetration in the next couple of years will hold crypto, then any financial application, any time of gambling, any type of prediction markets, they can just link their wallet and do that activity, right? Whether it's poker, whether, you know, it's anything. So I think that'll be the second big use case where you can unlock speculation and you can unlock yield from those kind of financial assets that are held by anyone globally. And on top of that, now you can build like some amazing consumer applications. And obviously, trading will always be a part of it, right? I think another really cool innovation that crypto did was perpdexes. It's called perpetual dexes. You don't have to own an asset, but you can trade on it in terms of just longing and shorting it. So I think that's opened up really massive markets where today the perpdex market is 10 to 20x of spot markets, right? If we trade like trillion, like hundreds of billions in spot, the perpdex market is even bigger, right? And I think that's where a lot of, you know, innovation lies in crypto and creating like better trading engines and these consumer facing applications.

Utsav Somani: Awesome, Nakul. Thank you so much for educating us and all the best to you and the team. Really appreciate it.

Dhruv Sharma: You guys have a great day. Fun chatting with you, Nakul.

Utsav Somani: All right, folks. Thank you so much for tuning in. Hope you've had a nice Monday. We'll see you on Wednesday at 4pm. Thank you. Bye.