Hemant Mohapatra of Lightspeed India joins Nick to discuss Investing in the Gen AI Extraction Layer, Value Accrual in New Tech Waves, and India’s Digital Currency & Identity Economy. In this episode we cover:
- AI Investment Opportunities and Challenges
- Healthcare and Legal Implications of AI
- Defensibility in AI and Long-term Investment Strategies
- Vertical vs. Horizontal AI Opportunities
- India’s Digital Currency and Blockchain Ecosystem
- Investment in Indian Innovation and Infrastructure
Guest Links:
The host of The Full Ratchet is Nick Moran of New Stack Ventures, a venture capital firm committed to investing in founders outside of the Bay Area.
Want to keep up to date with The Full Ratchet? Follow us on social.
You can learn more about New Stack Ventures by visiting our LinkedIn and Twitter.
0:17
Hemant Mohapatra joins us today from Bangalore, India. He’s a partner at lightspeed India. Before Lightspeed, he invested in software and infrastructure at a 16 Z earlier, what and earlier was a pm and engineer at Google and AMD. Hey month has invested in companies including supabase, pixel, dot space, sarvin.ai airbound and Pintu, amongst others, in addition to investing he’s an award winning poet whose work has been featured in major anthologies. Hey month. Welcome to the show.
0:50
Thank you for having me. Nick, yes, it’s a pleasure
0:54
to have you. I’ve been following you for some time and watching your content, so it’s really fun to have you here today. But can you tell us your quick, sort of two minute backstory and path to becoming an investor at lightspeed?
1:06
Yeah. I mean, look, maybe just really sort of a big pullback. Grew up in India, went to the US in oh three, right after 911 and it was an interesting time to be in the US. Spent about 15 years there, like most Indians, before we really figured out what we want to do with our lives. You want to be an engineer, so that’s what I tried to do. I worked for AMD as an engineer. I should have kept my stock, frankly, wouldn’t have required to work anymore, but sold my stock at the bottom, moved out of the US to go to the UK for an MBA, came back to the US and joined Google. And then I was doing a bunch of this interesting you know, early stage work for startups in the at Google and Google Cloud, because, you know, our charter was to basically grow the business by doing anything and everything. So I did pretty much everything except writing code. I did write some code which never really got productionized. But, you know, we did some small m and a deals that did not hit the co op dev team. We did some small investments, we did our product launches and BD work and all that partnerships. And then I ended up sort of spending a bunch of time with early stage startups, and began really liking it. Did a few of my personal investments, and I was actually quite happy at Google five, six years in, and one day I get a call from a friend of mine from a 16 Zn. She said, why are you wasting your life at Google? And if you want to start a company, come talk to us. If you want to join a company of ours, you know, talk to us if you want to make investments, come talk to us. And I actually remember looking up what Andreessen Horowitz is, because I was that unfamiliar with the venture landscape. And like, Okay, well, this is kind of a cool job, and I like my job, but this looks really hard and tough and something I never done. So I ended up getting in a process there and really liking it, and I want to move back to India for many years. So I ended up taking the opportunity with Lightspeed in 2018 and moved back, and I’ve been here since then.
2:48
Perfect. And tell us a bit about the thesis at lightspeed and your focus there. So
2:53
yeah. So we are a global venture capital firm. We manage about 30 odd billion dollars globally, of which about one and a half to 2 billion is in India in early stage. And look at this point in time, I think there is far there are far fewer categories that we do not invest in than we do. So we are fairly horizontally. Invest in consumer companies, and we were one of the first investors in Snapchat, in the US. We’ve invested in infrastructure companies, hardware companies, deep tech companies, Frontier tech company. So fairly broad charter. I personally focus a lot on global AI and software opportunities coming out of this part of the world. And maybe about 30% of my time, give or take, goes into what we could call frontier tech, which at least I define as companies and technologies that are absolutely net new to the world anywhere or net new to India. So I spend time in crypto, some climate I’ve been investments in space tech companies. Keep looking at, you know, robotics and quantum and variety of other things. So 80% of my time goes into AI and the rest in everything else.
3:54
And is it entirely India focused, or do you focus around the region of Asia?
4:00
Yeah. So we cover from the India fund. We cover all of APAC, and we loosely define ideas, yeah, all of Singapore, Indonesia, Bangladesh, and variety of other countries where entrepreneurship is and, you know, we dabble in other regions, like Japan a little bit. We spent some time there. We spent some time in South Korea, but not very active in those regions. Yet,
4:18
perfect. I saw why Bob Gupta in your portfolio, who’s a he was a classmate of mine.
4:24
Are you serious? Wow, yeah. He’s a founder of a company called Ron, and one of the fastest, most critical company in the B to B infrastructure space in India, in the B to B commerce space in India.
4:33
Oh, I think I’m talking about a founder of udon and Flipkart.
4:37
Yeah, that’s the guy.
4:40
Yes, perfect. Well, you know, I was reading in the information recently, and they were talking about Lightspeed as one of the most exposed firms to AI. You know, most of us are, are optimists, and we, we talk about the opportunity in AI, but the information kind of played both sides of it right. Is this a boom? Is this a buzz? Asked, Is this over hypes? Because we’ve seen these waves before. We’ve seen lots of capital plow into a wave, and we’ve seen corrections right along this innovation curve. So talk about what makes you bullish on the firm’s current positioning and exposure to AI,
5:17
yeah, look, I mean, I think this will not surprise anybody that, as most VCs in the world, we look for transformational changes, and we try to time it as best as we can. Nobody really gets it right all the time, but the change in our internet, the change in our mobile, change in our AI, I think these are transformational changes, and once they start to disperse in society, changes in pretty dramatic ways that we cannot predict today. How people date each other, our relationships are built or buying and selling, and those things happen, and our travel happens, and so on and so forth. All these things were very different 20 years ago, 10 years ago, and will be very different 10 to 20 years from now. So we do think of us as people that try to live as much at the frontier with the founders who are trying to create and push the frontier forward. And in many cases where we have, you know, the right opportunity, the right founder, the right kind of business and growth coming together then, like we deploy capital into those right in AI, is there a hype? There absolutely is a hype. Try building a company without the hype cycle. Everything is 10 times harder. Nobody wants to join your company, nobody wants to buy your product, nobody wants to invest in your business. So I actually, personally love hype cycles, as long as you’re discerning of where the hype ends and not drinking your Kool Aid too much, which all the best founders do, and hopefully the smarter investors would, I think you can make a lot of money and you can create a lot of change. So I absolutely love hype cycles, as long as you are careful about which founders you put money into, and what kind of segments in the wave you put what amount of money into. And we can talk a bit more about in AI what kind of segments we see, and why, how, why have we deployed what kind of dollars into which of these segments,
7:21
and before we jump into those specific segments, you know which of the trends do feel overhyped to you right now, and where do you think are some opportunities that you know may may be hiding or may be under hyped, but will be coming in a big way, in specific to AI across the spectrum,
7:40
across the spectrum, Hmm, let me, let me peel it one by one. Maybe I’ll pick up a few few, few big categories in AI. I think there is a pretty good deal of hype around the possibility of AGI and what it would mean to us. My personal view is that when AGI, quote, unquote, AGI comes, it would actually feel like any other day. An AGI would be a human of 98 IQ, which is the average IQ of the humans. It’ll do a few things that would really surprise us. It would do nothing really, really well. It’s almost like a human bond with almost the basic intelligence that a brain provides. A human brain provides, which is a function of your brain size, your neural capacity, your neural networks, and all the things that are kind of combined with each other, and how they are, how dense they are, and all of that. And then you train that human with different things. The same human becomes a cricketer. The same human becomes a mathematician. The same human becomes a poet, after years and years of training, and that’s where the weights of your neurons play a role. You may have a lean towards the artistic, and you may have a lean towards the mathematic. And similarly, the AGI probably would look a lot like this gooey mesh of neurons tied together in different ways that are impossible to really separate out. But in a question comes an answer comes out. And sometimes the answers are obvious, sometimes the answers are really surprising. That the AlphaGo moment right, and then you train that, that, that, that neural net to be an expert. So at one point, the same AGI model would be an Einstein, and the same AGI model would be inside a robot who can really navigate the environment in a very fluid way, and so on and so forth. So for me, the AGI wave seems like, I think it would look like, Okay, well, whatever. What do we do now and then the real work begins to train it, to optimize it, to standardize it, to make sure the outputs are reliable, safe, unbiased. That’s where the real world begins. So I think that’s one thing, that one thing that sounds like it’s over hype. What I think is under hype, frankly, is the I think. People are still too skeptical of when it works, just how powerful the output can be, and they are almost like this fulcrum that can almost lift the world up. It is that powerful of fulcrum just today, I don’t know if yesterday. Maybe there is a paper that came out of China, Alibaba research, and what they have been able to figure out is the possibility of a patient having neuronal intestinal or pancreatic cancer two years before it becomes a stage one cancer two years and if you know anything about pancreatic cancer, the survival rate of a patient is 5% with the operation with the pancreas is removed, and without the removal of the pancreas, it’s 0% but if you catch it at stage zero, almost two years before it actually becomes even stage one or two, the chances of level becomes 90% and above Wow, and specificity of the model is 96% or 92% which means if it says you may have stage zero cancer of this kind, 92% of the time, you actually do
11:08
have it. And that’s testable at stage zero. That’s it’s
11:12
a paper in Nature bio, so right here peer peer reviewed, I’m guessing they’ll go into lab tests and so on and so forth. But it’s a peer reviewed nature bio paper. So imagine the implications of that China loses 250, to 300,000 people to just pancreatic cancer every year, given the meat consumption and all of that stuff. And this could save a lot of lives, and all it took was some GPUs, some data and a transformer model. Remarkable, remarkable.
11:40
So, so the healthcare implications vastly under estimated at this point, healthcare
11:47
implications vastly underrated, vastly and there are so many segments like that where a small change could change the lives of many in India, for example, they are the population is so large the number of qualified GPS to take care of that many number of patients is actually minuscule. I think the ratio is roughly one to 2000 maybe, versus one to 20 or 50 in most of the advanced nations. And many of these people do not live in cities. They live in villages where doctors don’t want to go and imagine a world where you could take a photograph and get a diagnosis done, or a cell phone, and it is so accurate that, over time, legally, you can prescribe medication through that diagnosis. You don’t have to go travel two days in a train to come to a hospital to get a diagnosis, to get medicines, prescriptions, and then medicines. Imagine
12:39
you’ve seen the breadcrumbs of that already, like in the pandemic, you know, scripts written over the internet and, yeah, exactly,
12:46
all legal cases in India. India has decades of backlogs in legal cases, and they’re all tiny, tiny legal cases, some personal dispute, some civil dispute, some consumer dispute. And if you could actually train and model on indian legal data well enough so that the dispute resolution is final, and, you know, unassailable. In a court of law, you essentially have a lawyer, LLM that has all, all the legal right to pass judgment on small, small issues, like Harvey is doing for trouble tickets, so for traffic tickets in the US and so on that were really clear. I mean, imagine the amount of GDP you could value, you could create. But this is one single authorized LLM to pass judgment on people’s conflicts. I mean, stuff like that is so powerful, people are highly underestimating what it could actually mean to all of us, you know,
13:41
as as AI, capabilities become commoditized. Where do you think the long term defensibility comes from? You know, is, is it data, ownership, distribution, vertical focus, we read a lot about context lately, in the workflow, or is it something else entirely? It’s
13:59
a great question. Let me, let me sort of go back, give you a structural response, and we can talk about AI specifically. I think most of these super technology, super cycles come in 5030, 5070, year waves. What does that really mean? The wave begins when 2030, years of research and development has gone into a certain area. Let’s go back few 100 years, or 100 years oil and gas. Right? Where do you go dig for oil? So under most probability oil is there and will come out. Took decades of research, so much so that when US figured out how to do it, they ended up occupying a variety of countries because they knew they were to dig oil there. It will be oil, right? But that took 3040, years, and probably billions of dollars of R and D expense, and in the beginning of the oil revolution, oil being the commodity that is the resource that has been now unlocked and will be distributed over time to everybody that that first phase of that. That super cycle, that technology super cycle, is what I call extractive phase of the cycle, where the industries that are actually extracting that resource from, wherever the resource exists, are the ones that have the most momentum, that will attract the most capital, that will grow the fastest and will have the highest alpha, because you are in the part of the extractive cycle. If you are close to the ground digging for oil, you are the Exxon, the chevrons, the BP, you create the most value, which is why, if you look at the stock prices for these companies from the mid 90s to 2000s extractive cycle of oil, it was expensive and it got cheaper, cheaper, cheaper, cheaper, cheaper. Nobody had it. Then everybody had it. They were the most value creative. You look at their stock prices from 1990 to 2020 they’re kind of flat. Even the price of oil had actually gone up quite dramatically over that time. The actual Alpha was in software and internet. So the extractive cycle is where people should invest when a super cycle is very early in its stage, and the extractive cycle takes a lot of capital, oftentimes billions and more, and it actually works purely on price. The differentiation is price. You’ll have five or 10 companies that are going to be pretty big, and the big ones will become bigger. Why? Because they had the most capital to drive the prices down.
16:22
Is this the case for the AMD of the world and Nvidia? Yeah, this.
16:27
This played out in the CPU and the GPU super cycle. In the 60s. It began. There were actually five or 10, you know, Chip architectures. HP had its own. Sun spark was there, and there were a few others. The next 86 came, and IBM had its own and so on and so on and so forth. But eventually it all compressed to just x86 and maybe arm to an extent, right again. These are all lower down the layers. The job was to extract, extract, you know, compute, from silicon. They were the extractive phase. And then over time, that extractive phase became more commodity, and the value moved to what you build with that resource. You build cars with oil, and you build engines with oil. That’s where the car companies start to grow very quickly. And the Googles and the Facebooks took that, you know, CPU cycle, and then built on the internet. And the stock prices for these companies kind of really was flat. Then the GPU cycle came, and the same thing played out for the first 10 years, it was kind of slow then, then you could extract the AI. AI token is the next resource. So now we’re in the extractive phase of the AI resource super cycle. The first few years was the close to the ground, was Nvidia, AMD, and Nvidia obviously has taken up, you know, majority of that value, because these extractive cycles are capital intensive, already intensive, and they won’t have many winners. They’ll have only few. There’ll never be hundreds of CPU, GPU companies will be a dozen or half, a dozen or less. That’s what’s played out. Above that are the foundation model companies that take the AI tokens and make it intelligent, make it useful. Above that are data and infrastructure companies, middleware companies. Above that are application vertical, horizontal. Above that are solutions, companies and services on top of that. So that’s where the wave is going to go, right? So the first few years was absolutely dominated by the, what I would call all the commodity resources, the GPU cycles and the foundation models. There’ll be many foundation models in the beginning, and it’ll be few over time, because they only survive on price differentiation and some performance differentiation, but primarily as as you get more and more commodity prices, though, what is what drives differentiation and who can drive prices down? The big corporations like Google, meta and Amazon, who have a large balance sheet to put prices down, or companies like open AI anthropic Mistral that have raised billions to be able to sustain long price wars, and the and the price to win is that you are able to stake your flag on a piece of land that is so vast and so important that for the next generation of companies that are using that resource, the AI token, the oil drop cycle, the CPU cycle, will pay you rent because there is no other land worth Building on. This is it is that, why
19:02
is there a case to be made that this extraction cycle will be more condensed, it will happen much more already than it
19:09
already is. Already is, how many foundation model companies can you name on your fingertips now? And I can guarantee you two years ago, there were more, yeah, less than two hands, yeah. And then over time, you’ll have the excitement. Will move a layer of Bob, a layer of Bob, a layer of Bob, developers will move a layer above, and value will move a layer above, and investments will move a layer above. And that’s what we’re starting to see now, over time, it ends at the application layer, and that’s what we can see, and that’s what we are seeing.
19:37
So is Lightspeed actively investing, let’s say a couple tiers up in the middleware and infrastructure as well as the application layer. Or are those still somewhat nascent? Do you believe
19:50
I would say we? Are we? Are we are above the AI layer, below the high layer, around the AI layer. We invest across the stack. We’ve invested in hardware companies. We. We’ve invested in foundation model companies, data infrastructure companies, middleware companies and application companies, both in consumer and in enterprise. So just circle back, yeah,
20:13
just to circle back to the question then, so where does if we’re not talking about the extraction layer, or, you know, the LLM layer, where does long term defensibility come? You know, from? It
20:25
depends on who you are with. Yeah, it depends on who you are talking about. For the foundation model layer, the long term defensibility is essentially in getting adopted as the de facto piece of land people want to build their houses on. For some it is Google, for instance, you have to be Google, anthropic, Mistral, open AI in India, sarvam and so on. You could dominate that space and own it. And some of it would be improvement performance of a certain kind, like you could be much better at image generation, you could be much better at code generation, you could be much better at synthesis and reasoning. You could much better write something else. And there are different models for that. You could also build specific models for bio. It could be the best bio model to generate insights on cancer ligand binding molecules that you could put into test and so on and so forth. And there are models like that, and they will be absolutely price, you know, as differentiation, and just to an to a to a diminishing order of importance performance, because performance will asymptotically match each other over time. The first few years will be a large gradation of performance improvement, but the final few years would be, you know, we are kind of fighting for the last millimeter or an inch, not the last mile and then, so that’s a defensibility for the first few years. And these, these layers, are already solidifying and and becoming, you know, dominant, right? And that wouldn’t change much, because, as I said before, in the more lower down the stack, but it’s hot, it’s highly defensible, because it just takes the big ones become bigger. There is no reason for you to start another anthropic now, what would you differentiate on performance, price, how so? And then when you talk about the middleware, right? This is the one that stitches together the application with the foundation model. It is the hot spot of the stack, right? Now, why? It is very hard to know what part of the middleware will be eaten up by the Foundation model layer. Two years ago, things that were trying to do the job of a database or converting unstructured to structured data and so on, those were great opportunities. But now foundation models, you can throw a PDF file at it, audio file at it, video it’ll just figure it out internally on its own, so they don’t really know it’s harder to really invest there, mostly because they don’t really know how far the foundation model layer will come up. And there is not enough direction being given to them today from the application layer. I want this API, I want this throughput, I want this spec and so on, because there not that many large applications that have been built out. The only big one today, outside of a few other companies, is chatgpt, and it’s a full stack app. They use their own middleware and their own foundation model over time, over their own, you know, vertically integrated data centers, right? So maybe more on silicon. So the other hardest piece, and application layer, is where the value would eventually move. Like I said, that’s where the alpha will go. But then, obviously it is much harder to call winners of the application layer. So there we really orient towards the founder quality their vision. How clear is the vision? Can they build a solid team? And in some extent, where relevant, growth, product, all those things, we really orient a lot towards long term nature of their vision, versus leaning into the short term growth that will go away equally quickly.
23:43
What are your thoughts on application layer businesses that are in, you know, a very specific vertical, yeah, with very specific workflows, specific context, yeah, you know, can can winners emerge there by kind of developing, you know, their experience around those workflows, building their, their data sets, their their broad, deep data sets, and just having an edge that you know, other entrants that are well capitalized with really good teams can’t, can’t catch up to,
24:13
yeah, so I’m sure you will hear a lot of people saying that. You know, vertical is, is great. Love vertical businesses and all of that, but I’m trying to unpack why that is the case. It is a good answer. It is the right answer for the most part. But let me try to unpack why that is the case. If you know the scaling laws of AI, that is the algorithmic advantages, you can build a better algorithm, and it’ll converge faster to the right answer with less compute, less data and all that. But that’s one, one layer of improvement. You can make an AI. My AI is better than yours because I’m using a better model than transformer models. Second is just compute. You know, I have the same data as you, but I just have more GPUs. I’ll just run towards. The answer faster. Third is, I have just better data than you. You can throw as much compute if the data is completely random, you will never converge to the right answer. And we know that at the scale that these foundation models are operating in, it does look like the bitter lesson applies where you can throw a lot more compute to generic enough data, broad enough data, it starts to converge beyond a certain point, you can actually brute force your weight of the answer without having specific data. But But what that actually implies for us is that the convergence to the right answer is a combination of algorithm, compute and data. And enterprises have a ton of data. Now there are two kinds of workflows we see one is what I call the Open, open ended workflow, where you say, do X, X gets done, but you don’t know how it was done and toward quality. Let me give you an example of say, black box, yeah, yeah, yeah. It’s like, it’s an open ended loop. So let me give an example to make it real. Let’s say you have a company that is taking data from an Apollo or a or a, you know, bombarda or Zoom info, and, you know, it goes online, scrapes the web, and says, Oh, hey, month is so and so, age, so and so, gender, so and so work profile. And then it goes into a foundation models, and writes me a beautiful, powerful email. Then it says, Go to HubSpot. And there are million Hey, months they have different IDs. Go run a campaign against all of them and come back to me, and I’ll probably generate new leads. Now, this layer that sits in the middle between the the name database and the workflow that sends personally crafted customized emails to millions of these names and email addresses. This layer does not talk to HubSpot which is running the campaign on the other side, HubSpot knows which of these emails are opened. HubSpot knows what subject line worked the best. HubSpot knows whether this person opened an email between four to six or WhatsApp was better than SMS, was better than emails or better than HTML, was better than text or video. HubSpot knows that HubSpot has no reason to push this back to this company. Okay, HubSpot can integrate directly with the database and say, if the only value this person is adding in the middle is writing beautiful, customized emails, I can do that today myself. I can build a model so they will compress this middle and this middle piece does not improve, because the next time it gets another million haymans, it does not know what to do differently from the last campaign, because last campaign results never came back to it. It’s an open loop AI. It does not converge very quickly, and the quality does not improve very quickly. So we really care about in the application layer. We really care about founders who know what the open loops in their businesses are and how to so they may not all be open or closed today, but they have to have a view of how to close them. Will require more data, more control over workflow? Would it require more complicated workflow? Will it require more integrations, partnerships? They have to have a view of that. Okay? And it just so happens that in vertical companies, let’s say healthcare, the workflows are much more well defined. You go to a hospital, you are a male, you are a female, you are a geriatric, you are a newborn. It’s the same form, different names, different age, different you know, phenotype, different numbers, weight, but everything else is the same. It’s a step by step legally defined process. You cannot skip anything in the middle. So just because vertical software is a lot more defined, it is easier to close the loops. To know what the open loops are. It is easier to close the loops, which is why the AI that is being built for vertical software is converging faster to high fidelity, low ROS loss rates, high reproducibility, and so on and so forth, which is why the value that is perceived by the customers today seems to be just higher on some of these companies versus something that’s Much more horizontal. It could change.
29:01
Yeah, will the horizontal opportunities catch up as the models in the amount of compute? Yeah,
29:08
absolutely. They will absolutely catch up. And we’ve invested in both of these categories, but that’s the underlying reason why some of these things today seem like they are converging faster, having better ROI today, but over time. It’s anybody’s guess whether the horizontal you can or cannot catch up. As long as they’re able to close the loops, they absolutely will, theoretically, hey, month is there
29:28
risk? You know, you talked about before that the foundation models, in some cases, have gobbled up the middleware, right? They’ve decided they can do that themselves and put many companies out of business. There’s some famous YC lists, you know, floating around of companies that were put out of business overnight. Is that same risk the case at the application layer, as you know, the models get better and the compute increases,
29:51
you know, it. This is a very interesting question. The railroad companies do not run the No, the they don’t. Run the passenger business, right? Boeing does not have an airline. It does not have the muscle to brand itself as an airline, even though, without Boeing, there is no airline business. So it is unusual. The people who own the telecom wires could have built a content business, but they did not. I mean the net, the Netflix versus Comcast lawsuit in the US is is a great example where Comcast was claiming that, you know, you are essentially using my wires that I’ve spent 20 years digging into the ground, fiber on over, over the net, or whatever, and you are making ton of money on every single customer that I do, I should get a cut of that. And the lawsuit kind of went pretty deep, and at one point Comcast really throttled the Netflix traffic to its end consumers because they said that we own the end consumer. You cannot service content and charge them 50 bucks, and I’m charging them 20. And then Netflix won the lawsuit, and then the unthrottling happened. So
31:01
and we’ve we’ve also seen the CounterPoint. We’ve seen businesses like SpaceX vertically integrate and build like, you know, full stack Yes, instead of just occupying one one layer, yes.
31:12
So we are now starting to see in India, Jio is a good example of company that owns the fiber network, and they also own a Geo, TV and content business. So we’re starting to see some of these businesses becoming fully vertically integrated and have the muscle and the culture to build a vibrant consumer brand, while also having had the muscle and the right kind of, you know, talent to build the infrastructure part of this business. It is rare. It is happening with chat GPD, which is building the infrastructure piece. But it’s not an API company only. It also has a consumer brand called chatgpt. Sorry, OpenAI and chatgpt, and we are seeing SpaceX that may end up building a consumer space travel business, tourism business along the way, as they’re building the pipes to the space so I think it’s rare, but we back to your question, am I worried about it Only in rare instances, but more often than not, those who build the the extractive part of the business have not shown, historically, the metal and the pizzazz and the you know, ability to really rebrand, reinvent themselves to also build what needs the resources and then gets distributed to the consumers. The oil companies do not build the cars.
32:28
Talk to us for a minute. Hey, month, while I have you here, talk to us a minute about India’s sort of approach to medium of exchange and currency. And, you know, I’ve heard some about this, so you may, you may have to, you know, give it to us in layman’s term for the audience, but I’ve heard about this, you know, this structure of of currency and the way it’s being handled, and incorporation of blockchain and tokens. And you know, I’d love to hear kind of how you’re viewing that. And if you think this is, you know, I a positive opportunity for the country, or if it presents a bunch of downstream, you know, implications that are going to be difficult.
33:15
So in India, I mean, this, TLDR, is that, you know, 1.4 odd billion people, 900 to a billion odd people online, I would say, perpetually online. Average age of the country is under 3025. To 30 years of age, and very digitally native. India has one of the world’s lowest cost of internet and some of the highest bandwidths in the world, I pay about $2 a month, give or take, maybe $3 a month, give or take, and I get a four GB download free per day. And I get 4g to 5g speeds pretty much wherever I am in the country. So that’s where India is. India spends also some of the largest amounts of time online, per capita, per online capital. So that’s where, that’s where the country is from the currency’s point of view, you know, we have something so the currency, actually, we had this moment in India called demonetization, where a lot of the physical currency was kind of removed and taken back by the banks under the RBI, which is the, you know, the the Federal Bank for for India, and then that was the moment where India started to really digitize its its its currency, so to speak, where I don’t even actually, right now I’m sitting in my office. I don’t, I don’t have my wallet with me. When I travel to the US. I keep forgetting my wallet in the hotel, because I’m just not used to carrying it. Basically, you can go to the streets and buy, you know, vegetables with a street vendor and use what is called UPI, universal payment interface. Everybody has a QR code next to their offices, next to their hotel reception desks, next to their shops and so on, and just pay with your phone and are your identity is tied to your phone number, yeah,
34:55
so is your identification also digital as well. We have
34:58
something called Digital. Locker, which is the Digital Locker, and it’s the, it is the legally, legally approved way to store your your authentication. So we have something called the Aadhaar card, which is essentially the biometric card for each Indian. And everybody stores a digital identity on there. And we can get into a airport, and we have a face scanner at the airport, which you have to go through once to get approved, and then you can just pass through the airports without any entity. But even if you don’t have their entity figured out, the airports and airports don’t have it, then you can use the digital pass, and they are legally bound to let that, you know, let you in. So, yeah, all that exists. So in terms of your, you know, the currency, the India does not have that much of a vibrant crypto ecosystem because of regulatory pressures in India, because a lot of the stuff was, you know, essentially, for lack of a better phrasing, it felt like money tax saving, you know, like tax evasion was happening in India for the longest time. And I think the baby got thrown out with the bath water. Because there was a lot of great companies building crypto companies in this part of the world, and because of some bad players, all of them had to be, you know, exited out of the country, and then they all building there, but not seeing in India, Blockchain, on the other hand, is something that the government has actually shown a lot of interest in. We don’t have a CB, cbdc currency yet, so we haven’t really digitized the Indian rupee part in their work, there was work in motion to digitize, you know, your land records, digitize the currency and digitize, you know, your healthcare records and so on a blockchain. But ongoing things not not done
36:28
yet, and in your opinion, on, I mean, is this, are these the type of infrastructure capabilities that are going to position India really well for investment in the future? You know, a lot of people just talk about India as a cost arbitrage market, but it feels like it’s not that. It feels like real invention, real innovation, come out can come out of India, and with some of the factors you mentioned, it presents a lot of tailwind for innovation.
36:57
Yeah, absolutely. I mean, we invest in in in many companies, in what you could call net new innovation for the world. So we invested in a company called pixel space. It is building hyper spectral satellites from India, and they are the most advanced hyper spectral constellation in the orbit today, right now, highest resolution, highest visit rate and the highest throughput. And there are four or five companies in the world that do this, private companies as a resolution, but they are not in space yet. They’re still in the build phase. So we are the first and we are the most advanced. So India does have, you know, a fairly vibrant ecosystem around space tech. It has a very vibrant ecosystem around semiconductors. In fact, 20 to 25% of AMD, Intel, Texas Instruments and so on. Design teams sit out of India. Quarter of these companies are actually in India, and a variety of things like that. So what we do need to really focus a lot more on is to attract the best quality India talent back to India. What China did with their, I think it was called the 1000 mind, or the Million Mind project, where they pulled back high quality researchers that had exited out of China to the US in the 90s. They brought they targeted individually. Each of them, paid them high salaries, gave them a lab to run, and they brought them back. India needs to do a lot of that. That’s one that I think the government is thinking about, and we’re trying to shape that thinking a little bit from our side. Second thing is, India needs to spend a lot more on infrastructure development. India’s infrastructure spending over the last seven years has been as high as we have spent over the last 70 and will be equal to what we will spend in the next five so we are spending a lot more infrastructure. This is the roads networks, the airways, the railways, the, you know, waterways and all of that, but also variety of other things like a semiconductor plant or a solar farm or so on and so forth. But I think we still need to do a lot of work in improving India’s r, d contribution to the world. We are sitting in one to one and a half percent of AI papers that get published in top publications come out of India. It has to be at least five or 10% in the next few years, if not more. And China has done a fantastic job of really pulling ahead over the last few years. If you look at ICML, one of the top conferences in I think it’s happening in Canada this year, 50% of papers have a primary author from China, 14% from the US a year ago, I think it was kind of equal, if not more, in the US. So they have pulled up very quickly. India needs to really do a lot more there.
39:29
Perfect on today’s special segment we have Hemanth Mohapatra of lightspeed. Hey, month, can you tell us a story about a startup that you passed on? I
39:38
think it made us think about it. Yeah. Passed on that actually worked really well.
39:41
Usually people give an anti portfolio. Every once in a while they give, you know, a pass that gave them an insight about something, yeah,
39:49
so, okay, I have one that’s very it’s my part of my anti portfolio. It’s a company called postman. This is a company that is, I would say, very central. To the API economy in the world. It’s last round they raised was roughly at 5 billion. I saw them five six years ago. Actually, you didn’t really pass on the company. But this is an example in our business. If you blink and you have conviction, and you don’t really act on your conviction, it really haunts you for the rest of your career. So we had done a lot of work on the company the founder was not really raising, so we didn’t have access to the primary data, data room deck, none of those things, but we really, really had a lot of respect for the founder and the business they had built and how quickly they were growing. And we spoke to about 50 people across all the top companies that were users of this product, and we heard great things. So he went there, you know, since he was not fundraising, it took us a couple of months to really build that level of conviction. And then one day, one fine day, I heard that, you know, he is, he’s raised, and I was in India at the time, in the office, I heard it kind of skipped a beat, went back home, picked up my passport on the way, my EA book ticket to the US on the flight, we designed what kind of strategy we’ll do to really, you know, be in business with him. Landed with a term sheet and an offer. And unfortunately, just about eight to 10 hours before that, just the night before, in front of his office, when I was there, he said that I already signed with a partner that you know, that we like. And unfortunately, there’s no more room. And then I realized that, you know, it’s so rare in our business to find that tip of the spear company that actually is scaling has the highest quality founder. Really excites you, right? Tailwinds, all things coming together. And if you have done the work to build conviction, do not sit on your conviction, act on your conviction. Put a ring on it, otherwise somebody else with
41:41
I love that lesson, right? Because even off cycle businesses, there’s, there’s always an opportunity, right? Like this, this thing isn’t just done in phases and gates. Like, yes, you build a true relationship. Things can be worked out often. Yeah, hey, month, can you tell us about an exceptional founder you’ve worked with and what specific thing they do that’s unique to any other founder.
42:04
I’ll mention. Aves Ahmed from pixel space. This is a founder. When we met him and his co founder, they were right out of college, so they were 2122 years old, and they were trying to raise money for a hyperspectral satellite company to against, to go against Planet Labs, and every part of my brain was saying, shit, this stuff is just like impossible to do when you’re sitting out of here in a tiny College in India, and then you’re trying to, how would this be possible? You haven’t even built a satellite yet. But then, every time I sat with him and his co founder, shit edge the way my body reacted to the way these guys were responding to questions, just how hyper aware they were. Aware the world of hyper spectral imaging is, who else is at what stage of development, what would their technical advantage be? Who would they want to hire if they were able to raise Who are they going to they have a list of LinkedIn profiles. They really want to go after super senior people. They already had hired a couple of really senior people that were that were working for really large scale companies in the US, and they were because they had spent time with this founder. They were so impressed that they actually began to spend time and joined full time before he had even raised his seed round. So just the clarity of thinking, the resourcefulness, the ambition and aggression and the product and the technology depth. In deep tech companies, we really look for not just technical depth, but commercial nose. Otherwise it becomes an R D project. So we came away like against all of our natural instincts to wait another round and see what happens. We said, let’s go get in business with you, and we put that company in business. You were the first institutional fund to invest in that in a company in India, in space tech. And they’ve gone to raise about 90 $95 million in the last two years and become the largest company in the category. They have won the Fast Company, top 10 space companies award two years back, last year, they won the time 100 Innovation Award and so on and so forth.
44:06
Incredible. We’ve got about four minutes left, left here. Hemanth, do you have a few extra minutes? Yeah, sure do. Okay. Hemanth, how has your philosophy or approach to investing changed over the course of your career? You
44:19
know, one of the things that I have learned the hard way, frankly, is that great businesses take time. And when I was young, new to this business, I would see all these companies scaling quickly and raising a ton of money, and you would feel this that why not this company? Why not this founder and many of those companies did not survive. In fact, what did survive in scale were the ones that built the foundational layers, patiently, aggressively, quickly, patiently, and really were focused on what the couple of North Star things were for their business. For some it is technology superiority. Some, for some, it’s team and go to market. Some it’s. Brand and positioning, but they all knew what their North Star was. They’re really focused on just that and ignored the noise. Over time, I realized that being patient is a virtue, balancing a bit aggression is an art, and I did not see it coming in, you know, four or five years back before when I began to invest.
45:17
It’s a great quote. Heyman, if you could share one piece of advice with a young, new investor. What would you tell them?
45:23
Somebody asked Michelangelo, how do you paint so perfectly? And his response was, first you be perfect, and then you paint naturally. And the most important lesson for me is that work at building your instinct, and once you have gotten a good sense of your sphere of knowledge and your sphere of competence back yourself. Because, frankly, if you have spent time with the greats, if you have studied the greats, and you feel somebody who is not a great yet, but you feel instinctually that it is pointed in the right direction, you kind of feel the same things that you feel and you hear about the greats, and you have spent time with the greats, more likely than not, this person is going to surprise you on the upside, and at that point, it is so hard to anyway, find these people. You don’t exist. Literally don’t exist. It’s so hard to find them and not backing yourself. Then is a criminal mistake in our job. Then go back yourself. Have the confidence and back yourself. Don’t look for conviction. Don’t borrow conviction. Don’t look for consensus. Back yourself.
46:29
Love it. Hey, man, if we could feature anyone here on the show, who do you think we should interview and what topic would you like to hear them speak
46:34
about? You know, I I’ve learned a lot from Is he the investor friend of mine. His name is samesh Das. He’s a MD at IVP. I think his clarity of thinking, his approach to investing, and just what the founders say about him, is this exceptional. I think he’ll be a fantastic audience. He’ll be fantastic for your audience. His name is
46:56
good friend. Good friend of the show. He’s been out a couple times. Oh, wow, yeah. Hemanth, what book, article or video would you recommend to listeners? It’s a book
47:05
of poetry that I’ve really very close to my heart. It’s a book by a poet called Jack Gilbert. It’s called the great fires. And I love it because I think all great companies and all great founders at the very kernel of their heart, they are great stories. And my insulation towards poetry and reading actually allows me to really shuffle through the marketing sheen and really get to what the person’s story is, what the company’s story is. So I recommend that to really just understand what a human thought affair is all about, and what I how to really get to the heart of the human that you are interacting with that you’re getting in business for 1015, 20 years, because you’re going to grow old with this human that you’re backing. He or she’s going to grow old with you. They’re going to have kids, people are going to die, they’re going to have accidents and suffer grief. To Know Who are you getting in business with requires a certain kind of taste. So this book of poems, I keep going back to it to really cut through the marketing and the polish to the heart of who they really are.
48:13
Perfect. Hey, month, do you have any habits, tactics or behaviors that are a secret weapon?
48:18
I just don’t give up. I don’t know how to give up. It’s not a habit tactic. It’s just my nature. My father was in the military. He grew up in the military family, and for me, my entire career has been this. Brenna ferociter, fiercely, one step forward, forward, forward. So it’s not really a habit, but it’s something that I keep going back to when things are very tough. I go back to I cannot give up. I do not give up. I will take that one extra step tomorrow and day after, and we’ll put distance between this. And then
48:51
perfect. And then finally, here, hemuth, what’s the best way for listeners to connect with you and follow along with the firm?
48:58
Yeah, so Lightspeed India has a Twitter account at lightspeed India. Lightspeed ventures us has a Twitter account at lightspeed BP. My Twitter is very active. It’s Mohapatra Heyman. I’m very active there and open DM, so feel free to reach out if you know an interesting founder, or you are a founder yourself.
49:15
All right, he is Hemanth Mohapatra, and the firm is Lightspeed Heyman. Thanks so much for your your time today and all your wise insight and helping us understand what this AI wave means for the future. So
49:27
appreciate Nick, the real pleasure. Nick, thank you for your time. Thank you.
49:36
All right, that’ll wrap up today’s interview. If you enjoyed the episode or a previous one, let the guests know about it. Share your thoughts on social or shoot them an email. Let them know what particularly resonated with you. I can’t tell you how much I appreciate that some of the smartest folks in venture are willing to take the time and share their insights with us. If you feel the same, a compliment. Was a long way. Okay, that’s a wrap for today until next time, remember to over, prepare, choose carefully and invest confidently. Thanks so much for listening.
