Full Transcript

Dhruv Sharma: Happy Friday, listeners. We do this live, which means we bring insights to you live and also bloopers to you live. You'll notice Utsav has a really nice t-shirt over there. Utsav, remove your hand. What do we see over there? And how can people get one?

Utsav Somani: AI Native, they write to us. Leave a comment on our social media, on our WhatsApp community or even on this live stream and we'll find you. Make sure the comment is marked. We're not distributing them freely. We have limited quantities, but they're all fun. Compound Daily, 996, Locked In and AI Native. I'm not an AI native. And we have four guests today. Four guests today. Deep Tech Special today. Crazy. Let's go. Yeah. So let's cover the news quickly. I think the major thing that happened and people are tearing this apart in both sides of the argument where somebody is saying this is because of AI productivity gains or this is just a bloatware being reduced from the company. So Jack Dorsey, who was the founder of X or Twitter previously, his new company that he still is an executive at called Block, they laid off 40% of their staff. So 10,000, they've come down to 6,000 people all in one go. And he did this all via a small caps announcement on X. So the truly Gen Z way of laying people off, it's a sad, sad thing that has happened. But people are debating, again, AI efficiency or is this truly just bloatware being reduced from companies? But the stock market reacted positively to this. The stock jumped 24%, making 600 million more for Jack Dorsey on paper gains at least. And the company is profitable, their Q4 gross profit was 2.8 billion. And they've spent on social gatherings within the company as well. So the company seems to be doing well and he wants to take this number of per head gross profit up significantly as per his recent expo.

Dhruv Sharma: That's the general wisdom they dish out, right? Which is if you have to do one of these things, do it from a position of strength and cut once and cut deep. And I think his post also says something in that regard.

Utsav Somani: 40%, I think that's just insane. So he makes this comment that within the next year, I believe the majority of companies will reach the same conclusion and make similar structural changes. And I mean, you'll be remembering the Elon Musk cut at Twitter as well, where, I mean, he just went pretty heavy in terms of these layoffs as well.

Dhruv Sharma: There's one counter to that, which is, you know, the Citrini piece. Was it earlier this week? God, everything moves so fast in AI. Like maybe it was earlier this week, but it caught everyone in a frenzy. Since then, a couple of rebuttals have come out, including one from Citadel, where one of the lines in their rebuttals is that, I mean, all of this, I'm just going to paraphrase it a little bit, but that doomsday scenario, I mean, it just presupposes that intelligence is going to be either cheaper or free. And that is far from, I mean, compute costs money, you have to pay for electricity and a dozen other things. And so only if the marginal cost of compute rises above the marginal cost for human labor is there going to be like large labor substitution. If that doesn't happen, then our concerns might be too ahead of its time. I have a lingering feeling Jack might be forcing some sort of regulation against like AI layoffs and mass centers, which is why he just, you know, preempted them and did like a preemptive strike in a sense.

Utsav Somani: Do you think they're coming? Like, do you think job cuts by AI will be regulated? It's speculation, but I personally think they're coming. And what happens, I mean, if you're going philosophically for a minute, like, do you think like these AI native companies will typically be low in terms of headcount or will people still, I mean, use this as an excuse to like just chop people off or low performers out of the team?

Dhruv Sharma: I mean, if you're AI native, then by design, you can just, you know, build a company around a lean team. But if you're, you know, you're AI adaptive, then you have to, again, then you have to restructure the organization. And my sense is that because jobs around the world, especially in developed economies are contentious political, you know, factor, they will step up and create some regulation around just laying people off at Marseilles or doing something to sort of redistribute any productivity gains and capital.

Utsav Somani: And let's talk about the world of releases. I mean, February has kept up, I mean, it's going at an absurd pace, like February 5th saw Cloud Opus 4.6, which is a big upgrade for coding and different use cases for tooling. Then came Gemini 3 DeepDrink, and then came 3.1, then Grok 4.2, then Cloud Co-Work had some upgrades as well, new connectors. As recently as yesterday, we saw Perplexity Computer and then Google Nano Banana 2. It is insane. Like all of these were supposedly an annual release cycle until last year, and now they're happening on a weekly basis.

Dhruv Sharma: At this point, I'm wondering if the models are dictating when they release the next version.

Utsav Somani: And it's just happening so fast. Like somebody actually made a joke on X, where they said that to keep up with all of this, you truly need to be unemployed. And I actually ended up posting that you just need to tune into TUN, where you'll get all the latest insights. All better ways of doing it, yes. Keeping up to date. Talking about Indian venture capital, there was an article that came out on Newcomer, guest written by Sriram, where they went deep or sort of like a peek behind the curtain of what went down at peak 15. But apparently Ashish Agarwal, who was one of the top performing MDs for leading the Grow investment, which became the second highest outcome from the tech ecosystem after Eternal, he demanded extra carried interest. So he was already getting partner level economics from the fund where Grow had invested, had been an investment. And he wanted more of it. And so, I mean, he wanted more than triple digit million returns. And that's caused some friction.

Dhruv Sharma: I have two things to say on this, by the way. Yeah, please. One is maybe one way of avoiding this would have been to be a solo GP like yourself in the first place or doing like a billion dollars. Unless you're Lockheed. Unless you're Elliot Gale. On a lighter note, one or two, of course, single asset deal by deal SPVs and no such conflict arises. But I think the second point is, and maybe this is the most serious one, which is, is venture really a meritocracy? Where, you know, the sponsor is entitled to disproportionate gains? I don't know how you factor this into the fund docs, right? I mean, you could say X percent of an unknown outcome and X percent could be five percent or 10 percent, but X percent of an unknown number is still an unknown number.

Utsav Somani: And all of this is penciled in or discussed when, I mean, it's 10 years back. So a decade, you're looking a decade into the future. And it's very hard to imagine when all of these funds were being created, India had not seen a big liquidity cycle as well. Very true. So you, when you're investing in Grow, when stockbroking existed in many different forms, like Sharecon, even Zerodha at that point of time, you cannot imagine that Grow would become the second largest outcome. And so, I mean, of course, with hindsight, it's easy to say that, hey, wanted more of the economics, but in the future, of course, and somebody putting you in business as well. But that being said, I think economics and, I mean, venture returns and all of this is a very opaque industry. So I think people do want to study this a little bit more closely. All right. I think we've covered what we had to discuss this Friday, but I think we've got an exciting lineup of guests today. Let's welcome our first one, Gautami from Aqua Air X. Gautami, welcome to the show. Hi there, Gautami.

Gouthami T S — AquaAirX⁣: Hi, Dhruv. Hi, Utsav. Thank you for having me here.

Utsav Somani: Congrats on raising from Zerodha. We were talking about the stock market, but Zerodha runs Brain Matter. So Brain Matter has been awesome and supporting awesome companies like yours. So for our listeners who are hearing about Aqua Air X for the first time, can you explain what you do and what are amphibious drones?

Gouthami T S — AquaAirX⁣: Yeah, Aqua Air X is a deep tech maritime robotics company. We make underwater intelligence accessible, faster, reliable, and make things move quicker in oceans. So that's what we do. And amphibious drones are the future of maritime segment where you don't need to have multiple systems to do your inspection surveillance. You could combine them in a single system which can seamlessly fly and dive underwater like a science fiction.

Dhruv Sharma: Gautami, maybe it'll help our listeners if you could explain how maintenance is conducted when a ship is at sea and traditionally what roles have divers played and what do products like yours have to, what role do products like yours have to play over there?

Gouthami T S — AquaAirX⁣: Sure. Today, if you look at any inspection maintenance repair in offshore, typically this undergoes a long cycles because first you need availability of the ships, then you need an ROV which you will further deploy through the ships. And then you have a lineup of very skilled, highly skilled divers who goes underwater even today to do certain inspections and maintenance. Now, how Aquarex could solve this is let's imagine, imagine this as a picture that you don't need to wait for the ship. You don't need to wait for the skilled people. All you have to do is fly our vehicle from the shore, send it to the location where you need to do the inspection. Bird all the way from shore to the location flies in the air, dives underwater, does the inspection, collects the data that you want and delays back through satellite or through the RF and comes back to your location. So the entire life cycle of you doing the inspection in the offshore is cut down by 80% and bringing this intelligence is our vision and that's where we are working towards.

Dhruv Sharma: Can you talk to us about the aerodynamic design and how you've managed propulsion in the air and then however you navigate underwater?

Gouthami T S — AquaAirX⁣: Oh yeah. So the very critical part of amphibious is this aerodynamics and hydrodynamics. If you look at our design, the first version that we came out with a drone, you know, quadcopters type of a design, but we understood that, you know, using the same sort of thrusters in both aerial and underwater is going to, is basically being inefficient in either of the domain. So, and also will not be able to achieve our endurance and range that we require. So we came up with this flying wing design and optimize our both aerodynamics and hydrodynamics for underwater. And we use the physics in its own capacity to, you know, help us in underwater because it's a flying wing. You might have seen Manta Ray is, which is basically a wing sort of a design done by US Navy and something similar was an inspiration. And we said, why not try a fixed wing type of a bird underwater, which could also fly. And that's how we have optimized. We have a dedicated propulsion system, which are optimized for their independent domains.

Utsav Somani: And as a company that was founded in 2024, I read one of your achievements was tracking the TRL 6 milestone. So talk to us about that and maybe other engineering challenges that you found were not solved before, but you guys are the first ones to do it.

Gouthami T S — AquaAirX⁣: Oh yeah. So we started officially in 2024, but me and Jitendra, we have been putting work towards amphibious drones since college. So that's how, you know, we were able to prototype things faster, fail faster and find out some of the problems very early in our journey. Even today, there are a lot of problems, a lot of things that are still yet to crack. But some of the critical parameters or critical tech that we have solved is our transition. In the entire mission profile of amphibious drone, the critical part is exit from water to the air. Rest, everything is solvable, but that's the critical part. And that's what we have been failing over the last, I would say, 12 months. And we happened to achieve and crack that in the last three months. And you know that the same was also demoed to IAI. So that way, the very critical component we have solved and there's still a lot to solve in the next 12 months and bring it to the life.

Utsav Somani: Talk to us about the Israeli defense partnership that you just mentioned.

Gouthami T S — AquaAirX⁣: So, IAI Newspeer is a global accelerator program. This is the first time that because, of course, the India-Israel partnership, they have come to India, they've started with this accelerator program to support deep tech startup. And this was their first cohort. And they handpicked only four startups throughout India in different domains. They don't, they don't basically like let's say if Aquarex is in maritime and they won't pick another three, four companies in maritime in that cohort. So they picked Amphibious Drone because they saw a clear use case for their harbor and port security integration. And we happened to demonstrate our technology. That was a very good time for the team to, you know, like get pumped up and everybody was on, you know, fast track mode to make sure that we demo it to the IAI ecosystem. They brought in a lot of clarity. They brought in a lot of user and operational knowledge and experience. So that partnership has been good. We are looking for a longer collaboration with them. Let's see how it goes.

Dhruv Sharma: Gautami, talk to maybe, you know, other deep tech founders who are listening to this about how you guys have managed to marshal the resources you're going to need to sort of bring your ambitions to life. My guess is even the prototype to test it, you need advanced facilities like wind tunnels, etc. They're not available everywhere. So how do you go about doing all of these things?

Gouthami T S — AquaAirX⁣: Yeah, I think for early stage startups, it's always difficult, especially hardware startups. You so in our journey, there are a couple of things that worked out. One is our college, Nittay Meenakshi Institute of Technology, where we graduated from, have been very kind to us to provide us the space, provide us the basic infrastructure to start off with. You know, a lot of time, basic infrastructure is also a problem. So the college has been kind for last, you know, two, three years, they've been giving us a free space to utilize. And we are very, I mean, on this, you know, television, I would want to say thank you for them. And second, yes, so IAC has a wind tunnel facility. And moreover, that we extensively do simulations, because in hardware-centric designs, every time when you fail from a hardware perspective, it's very expensive. So we do a lot of simulations, both on, you know, hydrodynamics, aerodynamics, structural simulations, to make sure, you know, when we are actually in the field testing, we minimize that iteration. Being said that, we run things parallelly, we don't wait for simulations result to completely come to us. But, you know, because of our experience, and because of some amount of knowledge that we have, we try to take a shot and test it before the simulations result come in, especially in the controls. We don't wait for our PID tunings, or we don't wait for our controls to actually tell this is what the numbers say you test to go for PID. In simulation, the problem is, it's endless, you can keep doing it forever. But you need to get the actual, yeah, real world numbers coming in between.

Dhruv Sharma: I understand. Do you also have a founder community of deep tech founders where you guys exchange best practices or figure out what to get there? Is there such a thing?

Gouthami T S — AquaAirX⁣: Yeah, absolutely. Without founder network, you can't really do anything like not just tech, right? In every aspect, founders talk to each other, in hiring, in legal, in fundraise, in teams, how do you handle teams? More than tech, everything else we talk because we all are tech founders. But what we need to learn from each other is how they have managed to scale teams. That's some of the, we have a robotics community where we exchange notes.

Utsav Somani: And what does commercialization look like for you?

Gouthami T S — AquaAirX⁣: We have just started our commercialization. But, you know, being said that we are still six months away to actually bring our products to commercialization ready. By this year end is where we are looking at fast tracking our commercialization. By then, products, Avatar and Nexa would take a good shape and be ready for actual world.

Utsav Somani: Amazing. Thank you so much. All the best on this amazing mission of yours. Thank you so much for coming on the show.

Gouthami T S — AquaAirX⁣: Thank you, sir. Thanks for having me here. Wonderful talking to both of you.

Utsav Somani: All the very best. All right, now let's shift focus. I mean, we're doing this rapid fire thing today, right? We're shifting focus, shifting context. So now we're welcoming our next one, Madhav. And he's going to talk to us about US healthcare. Madhav, welcome to the show.

Madhav Garg — Jindal Healthcare⁣: Thank you for having me. Thanks, Art. I love the background.

Utsav Somani: What does it say? Is there an infinity symbol?

Madhav Garg — Jindal Healthcare⁣: Yeah, it's an infinity symbol. It's something that we have in our company, which sort of symbolizes together exponential is what we, you know, go after. So that's, that's truly what it means.

Utsav Somani: Awesome. So let's introduce the company and the work that you do.

Madhav Garg — Jindal Healthcare⁣: Yeah, so my company's called General Healthcare. We essentially help hospitals and providers get paid by insurance companies. Our primary market is the US, which is where we operate out of. And, you know, in essence, our goal is to help providers focus on quality of care so that they, you know, and for us to focus on reimbursement so that they can focus on quality of care.

Utsav Somani: So US healthcare system is known for a lot of wastage, right? And you're using AI to improve that. So the revenue cycle management that you focus on, I mean, India, I mean, you're building, I mean, I understand that you're based in India solving for the US healthcare. And I think a lot of this play revolved around cost arbitrage, maybe. Do you think AI is coming for that as well? Or, I mean, we've been having this discussion on the show where, I mean, just recently we were discussing on the news segment today where AI is taking away jobs as well as in companies like Block or at least so they're saying. And a lot of chatter about AI jobs going away, all of that stuff. So how does that affect your industry?

Madhav Garg — Jindal Healthcare⁣: Yeah, very philosophical today, because like you said, lots of news coming out. I think every day is very different. You know, you had Jack Dorsey come out and say that he's slashing like half of his workforce or whatever that number was. And, you know, you had at least in our space, Andreessen Horowitz is a pretty big investor and they came out in this article that was talking about infinite healthcare, right? So, you know, broadly, 100%. I think AI is definitely coming for jobs. But philosophical, I am philosophical today because, you know, I've been obviously in this space for a while. There are two concepts that are very important to me. One is Jevin's paradox, which I'm not sure if you guys are familiar with that. But in essence, what that means is, you know, as technology automates, there's actually an abundance of things to do and an abundance of people to do those things, rather than lesser amount of sort of things to be done, right? So technology leads to growth rather than to scarcity. And you have to sort of emerge out of scarcity and build those jobs up, right? So the view that I have and that we have as a company is, you know, how can we convert the human knowledge worker into a machine operator, rather than for them to just focus on sort of manual, tedious tasks that they've been doing for a long time? And the second is that, you know, which is something that we are trying to contend with, which is probably what everybody's contending with is something called the bitter lesson, where, you know, people truly believe that, you know, generalized sort of model of intelligence is probably going to trump a very specialized model of intelligence, right? And so, you know, we are a very vertical specialized solution. We want to make sure that we remain relevant. And some of that is harnessing the subject matter that we've developed over the last six years in operating for clients and so on and so forth.

Dhruv Sharma: And Madhav, to understand what the company does a little better, can you walk us through the patient journey and the role RCM plays in it? And is RCM also what you might call a system of record?

Madhav Garg — Jindal Healthcare⁣: Yeah. So I'll answer the second question first. We are the layer on top of the system of record. We're the system of action. The system of record is what in at least the healthcare world is called the EHR, which is the electronic health record, right? So if you're a patient, you go to a doctor.

Utsav Somani: That's more patient focused and RCM might be more hospital revenue focused, right? Absolutely, right.

Madhav Garg — Jindal Healthcare⁣: So the focus that we have, like you said, itself is to layer on top of the system of record, right? And we're focused on making sure that we're billing out those claims and that we are getting paid on those claims, you know? So we take whatever information that is in the system of record and then we action that and make sure that we sort of put in intelligence to get paid on those claims, right? For the first question, you know, the process, at least in US healthcare, is a little different to India in that, you know, in India, the patient has the liability of getting the claim paid, right? So if you were to go to a hospital in an outpatient setting, you have a, you know, you stay overnight, etc. Then you have to go to the TPA desk and, you know, file your documents and your claims, etc, etc. And make sure that that's approved before you're checked out, right? And before they discharge you. In the US, actually, what happens is that the hospital and the providers take on the liability of getting the insurance companies, or sorry, the hospital takes on the liability of getting paid from the insurance company, right? So, as a patient, all you're doing is just handing over your documents and you're, you know, you're telling the hospital, this is, you know, who I'm insured with, BCBS, Medicare, whatever. And then the hospital takes on that liability, right? It's also more sort of wider network. In India, you know, usually, you know, the, you're in, you have to be in an outpatient setting, right? In order to have a claim, right? In the US, insurance coverage is for primary care, for dental, for eyesight, etc, etc. So it covers a more generic list of services. So the things that we do, and that's sort of how the models are different, but the thing that we do is make sure before the patient arrives to after the patient leaves, that we're sort of following the process that's needed to get paid, right? So right from the appointment to the payment, that entire process is what we handle for the hospital so that they can get paid on the claims that they need to. So that's typically how it works.

Utsav Somani: And are we talking about massive hospitals or are we talking about smaller care providers as well?

Madhav Garg — Jindal Healthcare⁣: Our target segment is sort of mid-sized physician groups, specialty physician groups and rural hospitals, actually. So we're quite mission-oriented in that, you know, the large health systems, you know, they will get paid by the payers, right? By the insurance companies. But what tends to happen with the smaller companies, the independent groups and the rural hospitals is that, you know, you're dealing against a counterpart, right? And so you have to sort of have that negotiating power, right? So the smaller players are, you know, they're getting squeezed for the margin, right? And typically, they don't have the experience or the expertise in order to work the long tail of claims, you know, to identify underpayments, you know, the things that you need in order to get the last 5-7% of your top line so that you can actually be a viable business. So our target segment is truly those groups that I wouldn't say only are at risk but need the help in order to operate as an organization. So that's typically who we go after.

Dhruv Sharma: And this is very interesting because, I mean, we don't always come across people who found a way to crack what we typically call tier 2, tier 3 markets here in India. And how have you built an org that can do that? Especially your sales org.

Madhav Garg — Jindal Healthcare⁣: Right. It was tough. Listen, I started this company in 2019. It took me a while to get up to speed. You know, I built another company, which is still running, called Auctionate that operates out of India, which maybe we can talk about another time. You know, building from India to the US, penetrating rural hospitals, tier 2 markets like you said, Dhruv, it's a tough thing. I think credibility, trust, relationship-based selling is all that's worked. And, you know, we've built that trust in specific segments and certain markets. So, for example, we work a lot in Texas, right? And, you know, there's an association of rural hospitals, right? Now that we've been sort of ingrained in that area for a while, people recognize us, right? And it's been a slow burn. So, from an org structure penetration perspective, it's very tough, especially in the healthcare market. You have to have localization.

Dhruv Sharma: Have you hired locally a lot?

Madhav Garg — Jindal Healthcare⁣: Yes, yes. So, our team is based out of the US. You know, we have a sales organization based out of the US. We have a customer service organization based out of the US. Word of mouth is the best way to get sales. Maybe not the exponential growth, but, you know, it compounds pretty quickly. And as AI is sort of evolving, you know, we've done more demos in the last three, four months than we did in the previous two years, right? We signed as many contracts in the last six weeks as we have all of last year. So, I think things are evolving pretty quickly. I think people are getting used to an outcome-based model, which is what we offer quite quickly as well. And they want to go after that because they see the inefficiencies in that model. So, there's some tailwind behind that. But, honestly, it was a pretty slow burn for the first couple of years trying to figure out how to deal with the sales side of the market, you know?

Utsav Somani: You mentioned a couple of interesting things in passing. Because of AI, your sales cycle is reduced and shortened. Talk to us a little bit more about that. Like Dhruv mentioned AI adaptive versus AI native organizations. I mean, since you started the company in 2019, but you're clearly adapting to AI pretty fast. So, what's enabling all of this?

Madhav Garg — Jindal Healthcare⁣: Right. So, I think two things, right? One is we actually started as a services-based company, tech-enabled services, to really understand the market. My background is technology. The other company that I started is a technology company, Auction8. So, my goal was to figure out how to build technology in order to be a better company, right? That's obviously become a lot different in the last year. Where we are actually transforming into an applied AI company, right? We already used AI in order to be better at the things that we were doing. But like you said, adaptive versus native, right? In the next four or five months, we want to transform to native, right? And I think that, you know, that's very important for us as an organization, right? From a sales cycle perspective, I think the market is very, at least the market that I'm in, you know, can't speak generally, but the market that I'm in, you know, the sales cycle has been six to nine months, typically. It's been quite long to build that trust. And we're seeing that reduce to, you know, three to four months, right? And people are taking decisions a lot quicker because earlier the models were also, how can I handle RCM end-to-end? Because there are a couple of different processes within RCM. And now the market is a lot more in tune with how can I have a point solution for this specific task that I can do, because I can sort of implement very quickly. I can implement a system of action pretty quickly for that specific task, right? So the sales cycle and the experimentation for that is increasing quite quickly. So seeing somewhat of a crunch, you know, the peers, again, counterparty risk, you know, the peers are becoming more intelligent as well. They're using their own AI. They're increasing denial rates, you know, that you have to work essentially and get paid on, right? So people are seeing that go up, and therefore they're looking for solutions to that problem statement. So, yeah, very interesting time, you know, lots of change happening.

Utsav Somani: Utsav, do we have one final question for Madhav? One final question. Do you use this as a wedge to, I mean, you cracked, I mean, like you mentioned that trust and word of mouth and those things working in your favor, and especially in an industry like healthcare. I mean, I'm guessing it's a crowded, crowded market. Like you hear about different healthcare systems operating in silos. Sometimes you have, I mean, companies like Kaiser as well, doing the full stack thing as well. So where do you think your advantage in RCM helps you expand into this healthcare pie?

Dhruv Sharma: I have a question to add to that, which is, hey, Madhav, is Maha real? Like the RFK make America healthy again?

Madhav Garg — Jindal Healthcare⁣: I mean, listen, I think it's real in that people are listening to him. How much will come out of it, I think is a question mark.

Dhruv Sharma: Are the burgers getting smaller?

Madhav Garg — Jindal Healthcare⁣: They're not. The burgers are becoming bigger and they're becoming more expensive. Yeah, exactly. I mean, there's always a counter to the food that you eat, you know. But to answer your question, Utsav, listen, man, there's like 400 companies that we compete with. That's how fragmented our market is. The market is like $350 billion. It's a huge market, right? And that's why it's so fragmented. Certainly it's not a winner-take-all type of market. And certainly there's going to be solutions that come up. But the model that we started with in 2019 is something that we've evolved from, you know, in the last two years itself, right? And I think that market is going to go down from 400 companies to probably, you know, 60, 80 companies most, right? But I do think that there's a fair amount of room. And there's a lot of innovation that's happening in the models as well, right? So you have companies that are now talking about, you know, at the point of care, being able to recommend, not only from a clinical perspective, but also from a reimbursement perspective as to how to think about, you know, what you should be documenting, how you should be doing that. How you should be billing out to the insurance companies, et cetera, right? Because the latency and the compute capabilities have already emerged. But you can do that, you know? So I think there's some level of innovation that's going to happen that will change the way that we operate from an RCM perspective. But I do think that the landscape from a competitive perspective is going to boil down and crunch down. I think the pie is going to become bigger. I think there's, you know, 30, 35, 40% penetration from an agency perspective. I think that's probably going to go massively up at least in our target segment. And models will evolve. And certainly there will be companies that die or get bought out for their contracts and so on. And, you know, hopefully we'll be one of the winners that emerge out of that.

Utsav Somani: And that US market pie is large enough. So all the best. And thank you so much for joining us on the show.

Madhav Garg — Jindal Healthcare⁣: Thanks for having me, guys.

Utsav Somani: Thank you. All right, listeners, we're context switching again. We've got Pratap from Verve Semi, our first guest from the semiconductor industry. So we're truly keeping in line with the deep tech special today. Pratap, welcome to the show.

Pratap Narayan Singh — Vervesemi⁣: Thanks, Utsav. Hello, Dhruv.

Utsav Somani: You're on a very solid mission. And thanks for driving this forward. Everyone's talking about the ISM 2.0. We'll cover that in a bit. And you've raised a $10 million round as well recently. So congrats on that. Let's break all of that down with you. Can you explain what Verve Semi does for our listeners?

Pratap Narayan Singh — Vervesemi⁣: So Verve Semi is a fabless semiconductor company. Fabless means we don't have captive fabrication plants. Actually, there are no fabrication plants doing commercial jobs as of now in India. There is one which is very exclusive to government. So fabless semiconductor companies are like NVIDIA, like Qualcomm, like Apple, who don't own the fab. So fab is owned by other companies which are, it's like TSMC, UMC you might have heard of. So these companies, they just specialize the fabrication. And the companies who are actually controlling the intellectual property and products like NVIDIA or Apple or Qualcomm or Broadcom, you name, these are all fabless. So they don't own the fab. They use these fabs as a subcontractor to do the job. So Verve Semi is a fabless semiconductor company which is working in the in the field of analog signal chains. So analog signal chain is like if you consider your human body as a chip. So in this chip, the brain is the CPU. The senses, eye, nose, lips, ears, they are all sensors. And hands, legs, they are actuators. So this is how a machine or anything works. So what we do is the analog signal chains are that anything happening in the physical world, how it is transmitted to the CPU for processing. So it's like your hand, leg, mouth, eyes, ears. And this is what we specialize in. So we are building many products which are focused on our speciality of analog signal chains.

Dhruv Sharma: Can you help us understand with some examples, Pratap, the products you build, which systems are they embedded in further?

Pratap Narayan Singh — Vervesemi⁣: So the products actually Verve Semi owns more than 140 IPs in different markets. But just to give you an example, so many customers doing mobile base stations, these cellular towers, they use our IPs for receiving the data you create, the data we are transmitting right now. So all this data goes to the mobile tower and then sent back to the network and then streamed just like this one is happening.

Dhruv Sharma: And in order to build the products, I'm assuming you also have partnerships with foundries?

Pratap Narayan Singh — Vervesemi⁣: Yes, so basically it's not a partnership, it's like more buyer-supplier kind of relationship. But for intellectual property, there is a partnership because this intellectual property components we are building, they are very critical to the fab. Like you want to build a chip and for building that chip, you need some function. If it is not tested in that fab, it is very risky to use the fab. So analog components, they are very much dependent upon the quality of fabrication. It's not like digital, digital is very easy. In terms of fabrication, the digital leads are always far, far better than what you can analog.

Utsav Somani: So if we were to talk about the value chain, maybe a little bit more about, I mean, the India semiconductor mission 2.0 was announced, a lot of incentives were put behind this mission as well. But for founders who are listening to the show, hearing about the semiconductor industry who want to study what the value chain looks like, can you spend maybe two, three minutes talking about that? Where the opportunity lies for India specifically and for founders in India?

Pratap Narayan Singh — Vervesemi⁣: So semiconductor, see, semiconductor value chain is actually pretty simple to judge with. One is the fab. Predominantly all this semiconductor industry is like you are using the sand and making the gold. This is what happens. Some of the chips even we build, if you compare them by weight, they are more expensive than gold. So what happens is there is one thing which is fabrication plant. Then second part is the packaging plant. It is called ATMP, packaging plant. And third one is the fabless manufacturer. So fabless manufacturers are subcontracting the packaging, subcontracting the manufacturing. So value chain like any other product, it varies the same way. Who is more close to the customer makes most money, who is most far from the customer makes, the profitability is like that.

Utsav Somani: In terms of capex as well, I mean, that must vary.

Pratap Narayan Singh — Vervesemi⁣: Yes, yes, capex is very variable for fabrication plant, for ATMP and for fabless company. Overall, this whole industry is very capital demanding. So fabs create something like if a product is probably one rupee, the fabs create revenue of something like 30 paisa. And depending upon what they are doing, ATMPs make something like 10 to 15 paisa. And rest all is made by the fabless manufacturers.

Dhruv Sharma: Can you help us understand where the national priorities are at the moment? You mentioned that there is only one fabrication facility, mostly for, I'd imagine, strategic use or something like that. And also in terms of the nodes, like at what nanometer level are we very, very comfortable and where does the future lie for us? And also, I think in the value chain, one thing we could perhaps cover is OSAT and if there's an opportunity for India over there.

Pratap Narayan Singh — Vervesemi⁣: See, let's answer your questions one by one because these are too many. So, first question is, where is our national priority? So, you know, when supply chain was disrupted during COVID, there were many components which were not available at all. And you are building a car which is worth maybe, you know, 20 lakhs rupees. And the component which was actually stopping it to be sold was costing maybe 100 rupees. So, it is very critical industry, like, you know, you are maybe operating chips manufacturing plant or, you know, biscuit manufacturing plant and imagine there is one chip or one motherboard which gets broken. So, what happens? The whole plant gets disrupted, right? And you can continue only after you receive the replacement. So, the criticality for semiconductors is like you have nothing around you which is fully mechanical. So, those days are gone. Even for automotive, which was automotive and some of the industries are very slow adapter of this kind of automation. And some industries are very fast adapters like your mobile phone will adapt AI faster than your probably your engine of your car. So, the basically the national priority is that you should not be zero when you are, you know, one of the probably third or fourth largest consumer. You should not be completely dependent upon imports for things. Imports can be disrupted by many ways. You know, it could happen because of, you know, some kind of, you know, pandemic or something, but it could also happen because of some, you know, strategic non-compliance or non-alliance between two countries or some leaders, some politics for whatever reason. So, for our, you know, national priority is very simple that, you know, we cannot do 100 percent and nobody does right now. Even, you know, most of the countries which are, you know, much ahead from us, they also do not do 100 percent of semiconductors in their own country inside. So, you do not do 100 percent, it is very difficult to do 100 percent of it. But for some of the critical industries, we should have possibility to produce semiconductors inside and like you ask Dhruv, the technology nodes. So, let me tell you one thing. There is, there are few technology nodes which are called golden nodes for semiconductor development. And the technology nodes are not decided because everybody wants to do very high-end compute. It is decided based on how many applications it can actually fulfill. So, the most, there are two golden nodes. One is the 55 nanometer, one is 28 nanometer. Most of the analog which is around you except your mobile phones beyond 55 nanometer. You do not use anything below 55. But if, even if you want to do some, digitally sophisticated things, then 28 is the another golden nodes. Most of the FinFET nodes which are below 60, 16 nanometer.

Utsav Somani: But the mobile phones like being TSMC is using 3nm, right? 3nm or 3nm.

Pratap Narayan Singh — Vervesemi⁣: This is, this is for application processor. This is for application processors. But the application processors are high-end mobiles. If you exclude them, then most of your needs are probably 55 nanometer or something around that. They are fulfilled. There are three main applications. Actually, if you, if you want to distinguish between two, one is the compute. So, all compute intensive jobs, they need high-end digital processing. And most of the smaller nodes are very efficient for doing digital or improving density. So, up to 22 nanometer, the cost per transistor is getting lower. Okay. So, if you go lower, like 22, your same chip, if you just move the node, it will be cheaper in manufacturing. But beyond 22, the cost increase per transistor. So, the same chip, if you take it to 16 or 12 or 8, the cost will be more.

Dhruv Sharma: Out of curiosity, what is the highest transistor density in any, any chip at the moment, Pratap?

Pratap Narayan Singh — Vervesemi⁣: The, this is, you know, there are two companies in the world. One is NVIDIA and one is Cerebras. So, they have complete wafer as one single chip. It is like this. And one wafer is 12 inch. So, transistor density is trillions of transistor.

Utsav Somani: And before we let you go, can you tell us the scale of the business? You mentioned the number of IPs that you have, but the scale of the business, the partners that you are working with, anything that you can share?

Pratap Narayan Singh — Vervesemi⁣: See, most of the Indian companies, they are, they are all starting up. We, we are only company which is exporting IPs across the world. And it is, the company is profitable since inception. And these IPs are very popular. Some of the biggest customers of some particular industries across the world, they belong to us. They use our IPs. They are dependent upon our IPs. The, the other business is semiconductor products, where we are just, you know, we just raised our series A. We have just started building some, you know, sampling some products to customers, building some new products. But we are just starting up. For IP business, we are a 28-year-old company, and we have been pretty active in the IP licensing area.

Utsav Somani: All right, Pritam, thank you so much. I hope every device that is sold in India in the near future has a made in India or a designed in India chip coming out of Bobsemi. Thank you so much for coming on the show.

Pratap Narayan Singh — Vervesemi⁣: Thanks. Thanks, Utsav. Thanks, Rohit. Thanks, everyone.

Utsav Somani: All right. Final set of guests today, Shubham and Sushant from Prava. They're building Stripe for Agente Commerce, something pretty exciting, pretty topical. So please.

Dhruv Sharma: Are they AI native like you, Utsav?

Utsav Somani: Guys, do you want this T-shirt or not? For sure. For sure. Let's introduce Prava for our listeners, and then we'll ask you some questions around it.

Sushant Pandey — Prava Payments⁣: Thanks, Utsav, and thanks, Rohit, for having us here today. And about Prava, we are trying to make a payment stack for AI application and AI agents to make payments on your behalf using your card and wallet. What Prava does is that it integrates with your AI assistant and AI agents, and you have to just pre-approve the transaction. And whenever you ask your AI to actually go and do something for you on your behalf, it is able to do it securely using your card and wallet. And that's what a recent announcement with Visa was all about. Yep.

Utsav Somani: And when did you realize this was a problem to be solved? Like, what was the aha moment for you?

Sushant Pandey — Prava Payments⁣: Yeah, so for us, I mean, we were building kind of in the fintech space only. And we were trying to create like small AI agents as a GTM tool to launch on Reddit for people. And that's where Shubham and me, we both realized that what if AI could actually like go and make payments because AI is able to schedule a call for us. It is able to do a lot of stuff for us because we all have grown up like watching Iron Man movies. Like Jarvis is there, you're talking to it, it is going out and doing everything for you. So we're like, why not have something like that? AI is already there. It can act on our behalf. If it had to make a payment for us, maybe it can do it as well. And that's where we dig into a problem. We went to a hackathon in Singapore and there we built like a smaller version of it using stable coin infrastructure. And it got, I would say, a lot of traction from developers. And people started giving this idea that, okay, this can fit here, this can fit there. And that's where we're like, okay, cool. I think so there's some potential in this. And then around January of last year, we both went.

Dhruv Sharma: I read somewhere, Sushant, that for a while you guys were like, let's just take a punt and send stable coins becoming a thing in India. And then that wasn't happening. And so you decided to take matters into your own hands and then build coins. And now, tell us about that story.

Sushant Pandey — Prava Payments⁣: Yeah. So I mean, what happened was initially we started working with stable coins. And we were like, we're trying to create like a remittance startup for India where people from outside India would be able to send money here. Because I was working with some of the U.S. clients. Shubham was working with a Japanese startup. We both were working with clients outside of India. And we felt like getting money in a bank account is like a huge asset. Like even if there is a current market rate of 85, you'll get 81 from the bank. And the money will hit you after a week. So we were like, okay, this is very clumsy. We had experienced stable coin. We saw that it can settle up in seconds. It can hit your bank account very soon as well. So we were like, why not just make it happen? But obviously, we were running into a lot of hurdles with different regulators and there was no clarity in India as well on stable coins. So we kind of like lost hope on that. And that's why we were just doing some small experiment with launching small tools for people to try out, to get some traction over there. And that's when we realized like, I think so we need to work on something where we'll be able to deliver right away. We will be able to get something in the hands of people and they will be able to experiment. And that's how we came about this.

Utsav Somani: Dude, there's a lot of love for Prava in our YouTube comments. There's like a Prava army going on. Like people are sending like emojis and stuff. So we've got a lot of fans already. Yeah, I mean, I mean, so yeah, I mean, I'm sorry, forgetting the question.

Dhruv Sharma: So I think there's one question that looks familiar to me from I think a Nikhil Kamath video, if I'm not mistaken. We were on there.

Utsav Somani: Yeah, yeah.

Sushant Pandey — Prava Payments⁣: And I think Shubham can talk about a bit better because he went viral with his bit on Nikhil's podcast.

Shubham Kukreti — Prava Payments⁣: Nothing much. It was just like about an AI assistant. So I was just showing it to folks at WTF, the other founders from the same cohort. And apparently it was termed as, so it had like a anime persona. So wasn't it an AI girlfriend? Yeah, so later on it was termed as AI girlfriend. And so that bit got like a bit viral and people were just recognizing me from more like an AI girlfriend guy than Prava. So that was like all the funniest things going on.

Dhruv Sharma: Let's change that right now. Let's have Prava associated with your identity.

Utsav Somani: For sure. Yeah, please do that. I mean, will people trust an AI with their credit card? Like, how do you build guardrails into this whole system? This can go wild, right? I mean, we saw OpenClaw bot, which was running around and like, I don't know, burning tokens, spending money, sending random replies, like safety will be a big concern in this.

Sushant Pandey — Prava Payments⁣: Exactly. I mean, when we started, I mean, safety was one of the major points that was being asked by all the people around us. And we had to like find an answer to it because you can't like, you can ask your AI to actually go schedule a call for us. Maybe you'd be able to willing to give your email access to AI, but card and money in your bank account is something no one would actually do that because it can drain your account at any given point. So how do you put the guardrails in? How do you put the security in? That becomes a major question.

Utsav Somani: So how we thought will come up like in India, payments is typically extremely regulated, right? For stablecoins, of course, but this is just a whole different level. Autonomous payments. Yes.

Sushant Pandey — Prava Payments⁣: 100%. But the thing is, like, whatever is happening right now, we can make it work with the existing infrastructure and all of it is regulated. So we'll talk about what we have done with Visa is that when you approve your transaction, when in a partnership with Visa, what happens is on an AI application or AI agent, when you're approving a transaction, we are taking an approval from you through a simple passkey verification, or like a face ID or touch ID, where you approve that, hey, I want to buy this bomber jacket, a black bomber jacket from Zara for let's say $28. So once you approve that, what happens is that a new tokenized card is created for AI behind the scenes, and which is given to AI to go and shop on your behalf. Now the catch here is that this whole card...

Utsav Somani: With a particular cap, I'm guessing, yeah.

Sushant Pandey — Prava Payments⁣: Yes. So the amount is capped to $28. The merchant is also capped and can only be spent on Zara and can only be spent on the black bomber jacket that you approve. So this is the infrastructure change that card networks like Visa and MasterCard are bringing in. That is what this partnership is all about, where we're working with them in implementing this for AI application and AI agents at large. So any AI agent and AI application can integrate Prowl in just four lines of code with SDK and then get started. They need not to go and talk to card networks and go through the hassle of compliance. We take care of everything for them. They just need to get started with us with our SDK, and they'll have everything live and working for them.

Dhruv Sharma: Maybe there's an opportunity for someone to build like a personal finance school for AI agents that they don't become spendthrift.

Utsav Somani: Just one thing. So there were these commerce protocols, I think, Dhruv, we discussed right on the show, like Sprite was launched and Shopify was launched. New primitives altogether. Yeah.

Dhruv Sharma: In fact, that's a great point that you bring up, Utsav. And even speaking of regulations, even today, we can save our own card details, right? I mean, except for CVV.

Utsav Somani: Tokenized, yeah.

Dhruv Sharma: Yeah. And so, yeah.

Utsav Somani: Even Apple Pay is coming to India, apparently. There were news around that also. But talk to us, is this built or adjacent to these commerce protocols with Stripe, Shopify? I think all of these players are launching.

Shubham Kukreti — Prava Payments⁣: Yeah, sure. So commerce protocol, so how they are, they are more like, first of all, protocol is something like in the paper, you say theory. If we do this, this is something that can happen. And this is what everyone is kind of releasing. And it's the job of the giants to actually do it because they have that distribution and say, to bend everyone and say, hey, accept this kind of structure. This is the protocol you have to accept and we can do this. Of course. So also these protocols brought more adoption to the space. Earlier, I can remember when we were talking about when agentic payments, people were not even getting it and were like, how can AI even start making payment? And this is, and also like AI tools are not that mature, but now people are using AI to take actions actually. So these protocols are helpful and also like how Prava is operating. So end merchant and applications, they don't understand what protocol is. They just care. Like, can you enable agentic payments for us? And this is what exactly we are doing. So UCP, ACP, AP2, whatever the protocol is, we are making it available for the end developers as well as for the merchant. So they don't have to do the heavy working of understanding the protocol, changing the catalog, exposing their feed in a certain way. We do that for them. Also for the AI apps, we don't make them install tens of SDK, integrate with each of them. We just ask them to install our SDK. We make them available over any protocol they want to use. Also, we started with Visa. We are bringing more payment methods. A few of them you already talked about. So we are bringing all the payment methods so that AI apps just has to integrate Prava and they are go to go for any methods they want to enable for the end users. And same goes for the merchant.

Dhruv Sharma: Shubham, who's most enthusiastic about agentic commerce? Is it the card networks? Is it the merchants? Is it issuers?

Shubham Kukreti — Prava Payments⁣: Yeah, I think I'll say it started with the indie developers and AI apps. But a big thing happened when Charjipati brought their results. So I think it was something like 15% of the search query was about searching about a product. So this was a huge number. Charjipati has a huge user base and so many queries. So 15% of them were related to shopping. So I think these kind of numbers became exciting for merchants as well as the card network. Because if they don't do it, then something else will come up and take the pie over there. So now I'll say everyone is kind of excited. And also India is not behind. India NPCI is also doing a lot of things in that space, experimenting a few more things over there. So now I'll say everyone is kind of excited. No one's behind.

Dhruv Sharma: Are you expecting to add Rupee as well?

Sushant Pandey — Prava Payments⁣: I mean, we're soon going to have like a UPI integration that we are working on. So we have also partnered with a bank in India. So we'll start working on it. But our focus started with US because of this partnership with Visa, different card network, and because of the AI application that we're working with are largely from there. So it made more sense for us to focus on that market right now. And adjacently, we'll also start working on India because we've got a really action in there as well from a lot of people and applications.

Utsav Somani: I mean, just for us to understand what the scale of agentic commerce is right now, as things stand in the US, like, do you have any numbers you can share?

Sushant Pandey — Prava Payments⁣: Yeah, so I mean, like, overall, what's happening right now is that a lot of transactions are yet to pass through agentic payments yet, because the thing is like, these, these transactions, these implementation, a lot happening like at a beta stage. So they're a bunch of AI application, which are doing 1000s of 1000s of transactions every day with these kind of integration, but it is still to be going live in production with a bigger application. So the only thing that comes to my mind is like the integration that PayPal did with publicity. So that kind of integration have come out in the market where people are actually doing it. But those are still very traditional and regressive in terms of how the interface comes out for the user. Because when you're asking the user in an AI interface to fill up a form for you, that is not the best case. So so but yeah, the kind of market that we're looking at, it's predicted to be go at around like around 100s of billions of dollars, as we're looking at it, because every interaction in AI application where a user actually wants even a microtransaction or any kind of shopping interfaces, these transactions are going to go up in US alone, the whole ecom market is around at 250 billion. And even if even if a small part of it comes out in the market with agentic payments, it's going to be a huge, huge market.

Dhruv Sharma: Because I'm wondering if, if ever there's a payment failure, if the agent makes an immediate reattempt.

Shubham Kukreti — Prava Payments⁣: Right. So everything is like, so what we have built with the card network is basically when agent fails to make a transaction, of course, there can be retries, but it can't go beyond what is being told. And also if there is a deduction of money, but not fulfillment of order, for example, you agent went on a merchant website, it placed an order for a minute, money got deducted, but now merchant has canceled the order, maybe because of a number of reasons. So in that case, best part is that reconciliation is easy and refund is easy because it goes to the source of payment, which was the original card of the user, because it is directly. So issuer acquirers and card networks are in the confidence and in the process. So this makes it like whole easier and owners of this whole does not come to the AI apps as well. So the moment merchants cancel the transaction, the refund goes directly back to the original source of the payment, which was the end user's card.

Utsav Somani: All right. I think we're almost at the end of time. Dhruv, any final question or should I ask a fun one? Yeah, please.

Dhruv Sharma: No, I mean, I'm just asking them. So, I mean, if the card companies get too aggressive with their sales tactics and, you know, we have agents calling us to set up a set of stuff that we don't want, are you guys going to put agents in customers' hands so they can push back and just?

Sushant Pandey — Prava Payments⁣: I think so. If that saves you.

Utsav Somani: I have another boomer question or millennial question or whatever you want to call it. You are not a boomer, I think. But the internal philosophy has become boomer. I mean, I never thought that agents would be buying and paying for all this. So, I mean, I'm guessing like whenever you speak to a VC or an investor, they'll ask you this question that why can't Stripe do it? Are they structurally not positioned for this? Or I mean, even in India, Stripe globally where you're targeting US.

Sushant Pandey — Prava Payments⁣: Yeah, I think the structure plays a major role here because the thing is the way it works is that when you're on an AI application, you're not integrated with hundreds and thousands of merchants outside of your interface. So, there are like roughly around like on Shopify alone, there are 20 million merchants over there. And you're not integrated with all of those merchants in one single place. So, how do you actually allow your agent to go and make payments on different kinds of merchant? Each of these merchants are going to use different kinds of payment gateways. It could be Stripe, it could be Checkout.com, it could be any other network or any other PSPs over there. So, if Stripe really has to enable agentic payment, they'll have to build an infrastructure which supports payment on all of the payment gateways, which is something which is not in the favor of their business. So, they would try to build something which would always favor their own business, which is the ACP protocol, which is the agentic commerce protocol, where they have allowed like a shared payment token with OpenAI and ChatGPT. So, people over only be able to shop with merchants which are on Stripe already. But what happens to 80% of merchants which are not on Stripe? So, that is where we come into picture. We become kind of an orchestrator of payments, agentic payments over there. And we are able to switch between and route between multiple payment gateways and payment service providers. So, that's something which any of these existing PSPs are not positioned to do well.

Utsav Somani: Awesome. You have earned an AI-native T. Thank you so much. And whenever in the future, my agent goes and buys out something for me, I'll always remember this chat and recall this. For sure.

Sushant Pandey — Prava Payments⁣: Make sure it's Grava doing it for you.

Utsav Somani: Thank you so much for coming on the show. Wishing you all the best.

Sushant Pandey — Prava Payments⁣: Thank you so much for having us. Have a great day. Bye-bye.

Utsav Somani: Yeah. All right, listeners, that was it from Dhruv and me for today. We will see you on Monday. Have a wonderful, wonderful weekend. Stay safe. Bye-bye.