And How Matic Sold 6000 Robots with Zero Marketing | Navneet Dalal & Mehul Nariyawala

What does it actually take to build a robot that cleans your home when everyone before has failed?

Matic is a home robot that sweeps and mops your floors, navigating entirely with cameras, no LIDAR. It shipped its first unit in 2024 and has since sold 6,000 units at ~2,000 a month, almost entirely by word of mouth. And is now the largest consumer robotics company shipping in the United States.

Navneet Dalal (a computer-vision pioneer who co-invented HOG) and Mehul Nariyawala met building Flutter, a gesture-recognition app that became #1 in 72 countries and was acquired by Google, where they then worked on Nest cameras and shipped one of the first deep learning algorithms in the wild. Matic is the company they decided would be their last: they wrote "Not For Sale" on the wall on day one and built it to last 20 to 30 years.

Their bet was deliberately contrarian. They chose the "unsexy" floor-cleaning market, a category with a net promoter score of -1 that people keep buying anyway (21 million robot vacuums sold in 2024), because entering an existing market beats creating a new one and because it's the foundation for true indoor autonomy. Then they put roughly $35 million of their own money in, about 70% of their net worth, with no plan B.

Along the way they lay out a full worldview: why robotics is 100x harder than software (the demo is only the first 20% of the work); why humanoids doing your chores are still 5 to 20 years away (the data problem), why no consumer hardware sells above $2,000; and the skin-in-the-game philosophy captured by his late father's advice: "Sell your home if you have to, but keep the company alive."

If you're excited about how home robots actually get built and what it really takes to bet everything on hard tech, this episode is for you.


00:00 - Trailer
01:08 - When they quit Google to start Matic
03:30 - Solving home cleaning with cameras only — no LIDAR
04:46 - The $35M bet: funding Matic themselves
07:36 - What a "level 5" robot in your home really means
08:00 - Why they started with floor cleaning — on purpose
09:45 - The rule: never create a new market with your first product
10:02 - iPod, iPhone, Tesla — all entered existing markets
11:38 - Why new hardware gives you only one shot
13:02 - "Make something people NEED, not want"
16:25 - Why the demo is only 20% of robotics
18:50 - Teaching a robot like raising a child
21:30 - How far are humanoids from real homes?
22:22 - The data problem: "500 years of driving data a day"
22:56 - 90% in the lab, 60% in the real world
26:49 - Why no consumer device sells above $2,000
27:38 - Would you buy a $10,000 humanoid — for what?
28:47 - "History rhymes": General Magic to the iPhone
29:38 - Earning trust after 20 years of broken robot promises
30:31 - Shipping the first robot
30:45 - 6,000 units, all word of mouth, zero marketing
31:10 - Why they're US-only for now 
31:50 - The investors: Sutter Hill to the Collison brothers
33:20 - Two companies, both acquired by Google
35:00 - The Flutter story: #1 app in 72 countries
35:25 - Why nobody believed machine learning worked in 2011
38:40 - Microsoft Kinect: 8 million units in 60 days
40:30 - The Google acquisition — and the $35M number
44:40 - The near-death moment: switching to NVIDIA
53:10 - iRobot's bankruptcy and what it means for Matic
53:55 - The real scale of robotics: 21M robot vacuums a year
57:45 - Putting 70% of their net worth on the line
58:40 - His father's advice: "sell your home, keep the company"
-------------
India’s talent has built the world’s tech—now it’s time to lead it.

This mission goes beyond startups. It’s about shifting the center of gravity in global tech to include the brilliance rising from India.
What is Neon Fund?

We invest in seed and early-stage founders from India and the diaspora building world-class Enterprise AI companies. We bring capital, conviction, and a community that’s done it before.

Subscribe for real founder stories, investor perspectives, economist breakdowns, and a behind-the-scenes look at how we’re doing it all at Neon.
-------------
Check us out on:
Website: https://neon.fund/
Instagram:   / theneonshoww  
LinkedIn:   / beneon  
Twitter: https://x.com/TheNeonShoww

Connect with Siddhartha on:
LinkedIn:   / siddharthaahluwalia  
Twitter: https://x.com/siddharthaa7
-------------
This video is for informational purposes only. The views expressed are those of the individuals quoted and do not constitute professional advice.

Transcript

SPEAKER_00 0:00

Wait a minute, so there are 200 plus teams who think they can build cars that wouldn't get into accident and be safe, but there is not a single team that thinks that can build robots that does not require bumping in our homes.

SPEAKER_01 0:09

You quit your job that Google to start Matex.

SPEAKER_02 0:12

You have only one shot in the hardware business because by the time you build the product, you have already spent billions of dollars. And if the market doesn't take off, what do you do?

SPEAKER_00 0:20

Cloud driving cars has two amazing advantages. One is that it has Google Maps, and second one is GPS. No matter how smart the car is, if it didn't know these two things, it would be lost. In an indoor world, how does a robot know whether it's on the right side of this table or the left side of this table? And that was the key piece of the puzzle that wasn't solved yet. We were struggling to raise funding and all of a sudden Mark and Andrews and Horowitz Investor, and I think week later we had.

SPEAKER_01 1:02

So glad to be doing podcasts with you, Mehul and Daniel. Thank you for being here. Oh, thank you today. When did you quit your jobs at Google to start Matic?

SPEAKER_02 1:11

So we we left in early 2017. Uh Behul left, I think it was March or April in 2017. I left in the in the July of 2017. Uh, but to and the whole point was that Matic for us is our last company that we want to work on. We envisioned that we want to work on it for 20, 30 years and not get bored. Uh so we picked a very ambitious uh problem that, hey, we want to make sure that we can genuinely add, uh save people time and save people effort through autonomous robots in homes. In homes, right indoor environment. And that's the that's the overall theme. And we knew this is the having worked at uh like you know, so we were after Google, after acquisition of Flutter uh in Google Research. We spent a year or so in Google Research and from there moved to Nest to work on Nest cameras. And Mehu led the Nest camera portfolio there. He worked on the uh like Nest camera, three generations of Nest camera. I worked on Nest Camera algorithms. And through this experience, we did ship the first deep learning algorithm in the wild in 2016 on Nest cameras. I also got an opportunity to work on the Coral TPU chipset from a specification point of view for Nest purposes. And we saw that okay, the camera, the computes are there, the algorithms are happening, and now would be a very good time to go and create a different kind of an opportunity because nest was in the IoT devices at home. And uh, but homes have so many other problems. Like, you know, if you start doing the chores at home with the young kids at home, doing dishwasher, washer, like cleaning floors, then they're trying uh like washing clothes, etc., then keep tidying up after the kids, all of these become huge, gigantic efforts. And so that was our notion that hey, we're gonna save people time and effort. The the idea to put our own money was also that look, if this is our company which we want to put for 20 to 30 years, uh, like you know, this notion of skin in the game from uh Nicholas uh Taleb uh that like hey, you need to have make sure that if you do fail, it hurts. So we this is this is back in 2017. The whole idea was that we're gonna solve the problem using cameras. And if you remember, self-driving cars, they were all putting all kinds of LiDARs, time of flight, radars, etc. kind of sensors. So it wasn't really obvious if you look at what we met does today, that you can indeed solve the problem using cameras alone. So we spent the first year and a half just building the technology to prove to us, forget anyone else, prove to us that we can solve the problem using cameras. And it's only when we got it working enough, we have done all the demos of the core critical vision that this is how it should work, this is the kind of experience we want to provide. That's when we went out and raced our angel round. So first Nehul mentioned earlier, million and a half. We just first uh 18 months just to bootstrap to build the demos. Then we raised our kind of uh you can say uh angel round, and then we build a prototype of the product, that this is what the product would look like, this is what it would do, and then we did the next funding round. And after that, this long slot of uh just getting the product working, uh, which was perhaps also the hardest time in terms of like, you know, it's you have to just keep your head down and keep continue working on the product. Yeah, robotics is hard.

SPEAKER_01 4:40

And you know, we you mentioned in an earlier conversation that you put 35 million in matic, combine both of you. Yeah.

SPEAKER_00 4:49

Yeah, um, that one we got lucky. So, so you know, uh part of the reason is I think we you touched we touched about it a little bit earlier that we heard from a lot of second-time founders as they're thinking about doing it second time, that it's always harder to raise funding, but second time it's slightly easier because based on your predicory, they might give you at least angel investors will come in slightly easily. So that kind of gives you a feeling that okay, your idea is really vetted out. But it's usually not the case. So we heard that, okay, before you do anything, make sure you've thought through your point of view. So then we just said, okay, let's make sure one of the criteria was let's make sure there is skin in the game. That just like Flutter here, also, we created a um situation where there is no plan B, that this is only planning, like put all the all the chips on the center of the table, so to speak. So we bootstrapped initially and invested our million and a half ourselves before um before and and built a demo and showcased the technology potential in 2019 and then went out and raised our change around. No, but I I do remember on day one, we so we we didn't have an office. We were gonna uh I had a spare room on uh sort of a loft in my home. And then it came in, and first thing he wrote down on a piece of paper was that whatever we do next, we're not gonna sell it. So not for sale. And then we pasted it up there. So that was the idea that this is the last job, this isn't something we're never gonna sell. And then we believed in a few different things. So this time around with Flutter, we we started working on it, but we hadn't really thought through what the long-term potential was. So we always knew what we were gonna achieve next for in the next 12 months, but not beyond that. Versus with Matic on day one, we had a plan for like next 30 years. So we had thought through everything. Uh, and then there were sort of rules of engagement. One of them was queen of skin in the game. The other one is we had been in computer vision, so let's make sure we do computer vision as a core uh sort of strength. So that also worked very well in our favor because you know, one of the theses was that indoor world is built by humans for humans for uh uh eyes to eyes and brain-based perception systems, so showing um, you know, indoor robots have it. And this was also around the time uh, I don't know if you had visited back then, but if you came in 2017, you would probably find 200 plus self-driving car startups in Silicon Valley.

SPEAKER_01 7:12

Yeah.

SPEAKER_00 7:12

So self-driving cars was all the rage, right? So it's like at some point Namnit and I are like, wait a minute. So there are 200 plus teams who think they can build cars that wouldn't get into accident and be safe, but there is not a single team that thinks that can build robots that does not require bumping in our homes. That doesn't make sense. Like there has to be something uh that we do. And then that sort of led us to this idea that wait, if level five robots can exist on the roads, why can't they exist in an indoor environment? And why can't they exist in homes? And what does it even mean? And then we're like, okay, if level five robots for cars means that cars drive like humans, then level five robots for indoor world and homes must be that they behave like humans, uh, navigate like humans, avoid obstacles like humans, clean like humans, manipulate like humans, do all chores like humans. So how do we go solve that? And this is probably the right time to go do it. So that's how it's starting out.

SPEAKER_01 8:01

But right now, cleaning is the use case that you have got into production.

SPEAKER_00 8:05

That was a very deliberate choice, yeah.

SPEAKER_02 8:07

Yeah. So the the notion and idea is that so we knew that robotics wave hasn't really come to like you know, yes, there are robots, like you know, iRobot was started in 2002. It's the kind of the only consumer robotics company which will do this initial algorithm was it's kind of a random walk. So the robot is in a home, you start it, it will just go in a random direction, bump into an obstacle. It's it's uh it's literally like you have a blindfold around your home and you fold your hand and keep bumping into things. Yeah. And when you bump into something, you turn in a random direction, go do it. So you can, in principle, clean an entire home if you don't like you know get stuck or something. But you have to also need an infinite amount of time to make sure, guarantee that you have done coverage all over the places. So it was, and that was kind of the what could have been done with the technology back then, because the algorithms were not there, the sensing suit was not cheap enough, like the like the computes were not really even feasible. Even if you have all of these things, you can't really do all these computers on device in real time. And our work at Nest gave us this like this opportunity, like, okay, things are starting to fall in place, but it will still require years of work to develop the core stack in robotics because it's not like today you can go download something on GitHub and say, like, hey, the operating system for the robotics is working, perception stack is working, things are plugged in, and you have a robot moving in going from point A to point B seamlessly. And we we realized that okay, it will take time. And one of our core beliefs and theses is that when building consumer hardware, you do not want to, with your first product, create a new market.

SPEAKER_01 9:54

Why is that? Can you can you elaborate?

SPEAKER_02 9:56

Yeah, so the the the let's let's take some historical examples that would be perhaps very good example. For when Apple worked on the iPods back in uh like you know 2001 and 2, we were already using Walkmans. Okay, Sony also had a network Walkman, which is kind of an iPod but a small disk. And Apple said, like, oh, people are using Walkmans, let's give them amazing Walkman experience, right? And they built an iPod. When they did the iPhones, again, like, oh, people are using Nokia phones, blackberries, etc. Let's combine the browser capability, the mobile capabil, the phone capability, the emails capability, and the music capability in one device, and that's your phone now, right? And similarly, if you look at like, you know, the watch, like you know, people wear watches, it's a digital watch.

SPEAKER_00 10:49

So again, even if you go for personal computers, right? PCs, you can say, oh, PC was in your market, but it wasn't. Computers existed. There were a lot of professionals who had used computers, whether it was uh mainframes or the mini ones uh that they used to have. So this idea that computers could do things was already obvious, and there was a desire for a lot of these people to bring that computing home from universities over there. So there was a need that was obvious. Yeah.

SPEAKER_02 11:14

And so again and again, you even Tesla, like you know, electric cars, like cars, people are already using cars. Tesla said that, hey, we're gonna give you electric cars with 200 miles range, super fast acceleration, things like that. And so in at least in the last 25 years, like you know, which I would say as my professional uh journey or career, I haven't seen a new hardware like a business created with the first new product. Like, you know, for example, VR AR is a new category, and uh a lot of billions of dollars have been spent, but we still don't know what to do with these devices, right? And iPad was a new device, and it wasn't clear where to exactly fit the iPad, and it took almost 10 years for various experimentations till it got figured out. It's primarily for students taking notes in their schools and universities. So when doing a new product in a new market, it's an extremely risky because you are taking a lot of market risk. And we were already taking computer vision, computer risk sector, and then you add all of the market risk on top of it, and you have only one shot in the hardware business because by the time you build a product, you have already spent 200, 300 million of dollars. Like, you know, you have done all of that engineering work, and then you get one shot. And if the market doesn't take off, what do you do?

SPEAKER_00 12:42

I think I think that's the key problem that in software you can really go and iterate even within a startup and you can pivot a little bit with hardware. Um, you will almost always get one shot because just uh the time it takes to iterate is so much faster that you probably do not have a time for uh whatever version two or product number two that you can think of. And and so, so like the motto, one of the things we learned at YC was that make something people want. In hardware case, you almost always want to say make something people need. Okay, that's one. And then the second part is there were, you know, there were a couple of uh companies, like you know, you can say, okay, Fitbit created a new market. So there have been some successes, but we what we also observed is that when you create a new market, uh incumbents come in it really, really fast, and then it's really, really hard to create a mode or really compete in a long term. Uh so so, and then everything about floor cleaning category was so unsexy, and that was that's why it was a deliberate choice that one market existed. I had never met, and Namita had never met anyone who said Roomba changed my life. Yet category was growing 16% year over year, which means there was a more demand. Category net promoter score is literally negative one even today. So no one really likes their disc robots, no one really recommends them, yet people keep buying there, so need is intense. And it's unsexy enough that Alphabet is not coming or Google's not coming, Amazon's not coming, Apple's not coming. And if you graduate from Stanford with a PhD in computer vision and tell your mom that I'm doing self-driving cars, she'll say, Cool, my beta is you know amazing. If you tell her that, hey, I'm doing robotic vacuum, she'll probably say, What the F is wrong with you. So that was the point that that really this is the place where we'll be able to build, get a chance to build this foundation of indoor navigation and indoor mapping and indoor obstacle avoidance kind of uh that you need for robots or true autonomy at a at a at length. And then and that was really that was really the priority. That was the that was the point of view. And and to go back to the earlier self-driving car analogy, um, self-driving cars has two amazing advantages. One is that it has Google Maps, so it allows cars to know where the road is going. And second one is GPS, right? So it knows where the where that allows cars to know where it's located. No matter how smart the car is, if it didn't know these two things, it would be lost. We would be lost if it didn't know where the road is going or where we are located. So in an indoor road, how does a robot know whether it's on the right side of this table or the left side of this table? And that was the key piece of the puzzle that wasn't solved yet. And that is required to do precise navigation and understanding.

SPEAKER_02 15:20

The the the another there is one more fundamental piece. If you take from a pure technology point of view, okay, so we talked about that, hey, this is an existing market and customers want it. So if you build a product, you can sell to customers. From a pure technology point of view, you fast forward, let's say Matic currently does floor sweeping and mopping, and you say, like, okay, let's say there is a robot which is in your home and it can do other things, maybe tidy up, maybe do dishwashing, maybe do laundry, etc. And ask what part of the current stack on Matic is not useful for that robot. So, yes, there are skills which would be needed to be added extra. So this new robot has to have a super set of matic skills, but there is no technology stack of matic that you would say that, hey, I don't need it for the next product. So the way we imagine it is that irrespective of what robotics come, they have to solve all the same kind of problems that we have gotten working on the device, and it's not simply a function of demo. So in robotics, almost always so in software, let's say you can do build a demo, you have done 80% of the work, and you then spend 20% of the work productizing it.

SPEAKER_00 16:37

Okay. Great example is before Novnet goes on is you have you have GPD 3, and then six months later you have chat GPD.

SPEAKER_03 16:44

Yeah.

SPEAKER_02 16:45

Yeah. And in robotics, it's almost always the other way around, which is the demo is only about first 20% of the work, and the rest of making sure it is working is the 80% of the work. And it takes quite a long amount of time to get to the perfection to make sure that robot can do the task in an autonomous fashion consistently.

SPEAKER_00 17:09

Yeah. And then this is one other thing as well that with today's AI, we collaborate, which is, you know, if you're using coding AI, let's say cloud code, uh cloud code, um, if it does 90% of the job, right, you're mesmerized because it takes years for us to master that, right? Uh and we're happy to take it to the last 10%. But we don't necessarily go to school to figure out how to navigate without bumping. We don't go to school to figure out how to vacuum without breaking things or chewing wires. These are things we just know. So it's trivial task for us. These are the tasks we hate doing. So we don't want to collaborate with robot, we want to delegate. And and the imagination we had is imagine, you know, we grew up in India as well. In India, most of us can afford domestic help even today, even if you're a middle-class family. If you had domestic help that came and only cleaned one, cleaned everything except one corner of the room left dirty, you'd probably tell them, hey, make sure you clean that tomorrow. They leave that corner dirty twice, they're fired.

SPEAKER_03 18:05

Yeah.

SPEAKER_00 18:05

Because we don't want to be doing that. We just want to give it uh give it up. So that was the point that look, as a homeowner and as a family man, I don't I want to live in a perpetually clean home with perpetually clean floors. I don't want to do it. I don't want anyone in my family to do it ideally. But then there's a third requirement, which is I don't even want to think about it. It should just work. And that just work part is actually quite hard.

SPEAKER_02 18:29

Yeah. In fact, I would say it this way that if we were to restart, let's say, Matic today, we would go still solve the same problem.

SPEAKER_01 18:38

Okay, so so there's nobody uh else who has solved making a map of your home kind of problems.

SPEAKER_00 18:45

It's a I I would even so the analogy, and then I like to do it in a little bit of analogy way because that's how I think uh you know, I might I joke that my knowledge on the technical side is enough to be dangerous. So, so um we are as a human not born as a full-blown humanoid. If you actually think about it, between zero to five-year-old human child, all they learn is their home and how to navigate and perceive it. And even they learn how to not fall kill themselves by uh falling down the stairs or or really going down the water or a swimming pool or bumping into things or getting their fingers stubbed in joints uh uh in uh in uh different places in furniture and stuff. So it's really just about perceiving. So in the context of floor cleaning robot, we are really teaching how to perceive the world just as zero to five year old. Then five to ten is when we finally start learning dexterity, when we first start learning language, so we can add semantics. Yeah, that's when we start to understand the world, and then we do basic tasks, like you know, kids can all of a sudden put their shoes back in their place or organize maybe their bookcase or make their backpack, but it's still basic tasks. We are not gonna give this an eight-year-old where we usually do not give a you know plate full of grass or or knives, right? Or tray full of glass. It's only between 10 to 15 is when we start doing both long horizon planning as well as complex tasks of cooking and baking. So our thesis was always that it's a three-step process that lets actually solve perception first to the nth level, and then actually, and which is just simply that does robot know what home looks like? What are the items inside it? So today, Matic only knows as much as cats and dogs, which is it can navigate without bumping, but it still doesn't, I still can't tell it to go hang out by the couch in a living room. But can we add that information inside map and can we teach it to semantics as well? And if we get to that level, then once it has a semantic understanding, now you can start telling it to manipulate and move it. So that was the idea behind it. So it's really that step process. And we, and I think I mentioned, right, uh early on that that when we were talking of the of the camera, that even if you have humanoid, you still don't want humanoid to get tangled in your wires inside hole. Yeah, you still don't want them to fall down the stairs, you still don't want them to fall, uh, you know, uh walk into food spills or spaghetti spills or or even you know duck hoop uh things along those lines. So every single thing that you want your humanoid to do, we are already solving. Matic is already doing that. And if you can manipulate, like you'll see some of these demos where they manipulate, but they don't usually walk as much. They don't navigate in a complex environment around complex obstacles that you find in homes. So that still remains a challenge, and even for them, they will have to solve it. So we're just doing bottoms up.

SPEAKER_01 21:34

So how far do you think, you know, companies like Figure are working on like full autonomous robots? How far do you think we are from fully robotic house help that can do everything?

SPEAKER_00 21:46

I think I think uh uh everyone will say that uh there are two parts of it. So one is if you go to academic academia and ask them about indoor mapping and navigation, they will tell you. That it's a solved problem. Now, if you are trying to publish a paper, it is a solved problem. But if you're trying to uh but uh but the way we thought about it is if it's a solved problem, well, where are all the robots? You know, why isn't our world swarming with them from airports to grocery stores to hotels to homes to uh uh um restaurants? We don't have that. So clearly this is not a solved problem, and we need it, number one. Number two, jumping directly to humanoid thing, the challenge there, as everyone knows, is the data problem. So even to build self-driving cars, Tesla has, I think there was some stat that they're collecting 500 years worth of driving data in a day. So if you just want to build a self-driving car, you need that much data. Now, humanoids and home environment are quite complex, quite unstructured, quite dynamic. So, how much data do you need? So it's a function of who solves that data problem and how do you get that data and how do you acquire that precision? I think there was a recent article that said that even if you see a 90% demo in a lab environment, the moment you go in a real world, it drops down to 60%. And closing that gap is really hard.

SPEAKER_02 23:06

Yeah. So another analogy you can think of is that in case of a self-driving car, it has in practice, like you know, three degrees of freedom. It needs to move in the two-dimensional floor, basically roads, etc., and uh direction it needs to go. And um you have to avoid all the objects on the way. So you can be in case of a car a good kind of uh three feet or four feet away, or like you know, kind of close to a meter away from the nearest obstacles, etc., and you can make it work. And you we all know the amount of data and the effort it took to get Waymo and the Tesla FST working. Now, within robotics and when it comes to like in a humanoid kind of robotics that we are talking about, you first have the the robot itself has multiple degrees of freedom. Even if we say it's a robot which is just on the wheels, etc., moving three degrees of freedom, arm has seven, then the fingers have a few more, and then the object you are picking, like if it is, let's say, a water bottle, where do you want to pick it from? Is it on the top at the bottom? If you squeeze too hard, maybe the water can come out. So it has its own degrees. The object themselves have one degree of freedom. If there is another obstacle, let's say maybe there's a book or something, then grabbing an object and pushing in this sense may not make sense. Grabbing it, push, pulling it in this sense may make sense. So the world has very many degrees of freedom. And uh, so I would easily say it's 20 plus degrees of freedom, like you know, depending upon how people count. So it's just going to take time. Now, is that time like you know, hey, five years? We don't know. Maybe there is a new advent, new technology comes up which makes it easier to learn faster with an existing data than yeah, maybe. But it still requires solving a like you know, currently uh unlike the the problem solution still requires that, hey, first you absolutely need data with the current way of doing things, tons and tons of data, in fact, bigger than internet scale data. And even if you if you don't say like hey, you can do it faster, then requires some advantage in the algorithm improvement and things like that. So it can happen. And the thing is that our take is that, hey, whatever we are doing currently is anyway needed to build those kind of algorithms and data. So we are not necessarily losing uh an edge over any of the existing robots by going and building selling matic. In fact, we are building muscles on how to sell to customers, how to make products cheap, how to make product reliable. And all these things, even after we have the technology working, they still have to solve. You still have to build, you build a robot, like you know, you still have to solve, make it cheap, make it reliable, like you know, have it feature rich, make the industrial design look nice, and then do customer service, brand, reverse logistics, supply chain, and scale it up. So these are all the things you still have to do. And I think we are just flexing those muscles also while building autonomy.

SPEAKER_00 26:10

It the other way to think about this, and this is something we knew. So, so by the way, first robot vacuum was not Rumba, it was Electrolux. It came out in 2011. I mean, sorry, 2001. It was priced at $1,400 and it failed. It was packed with sensors and it failed. Then iRobot did this clever redesign of it and shipped it below $200. Now, we had a chance to connect with Rodney Brooks, who was the original algorithm person and one of the co-founders, and what we learned from him was that they were hell-bent on keeping it below $200 because that's the price point at which you don't have to ask permission of your wife to make that purchase. So price point makes a big difference. Now, if you extend that, then you realize that there is literally zero ubiquitous consumer electronics device that's priced higher than $2,000. Beyond $2,000, you pay for professional equipment. So either in a prosumer space or only thing consumers spend higher than that is cars.

SPEAKER_01 27:08

Yeah.

SPEAKER_00 27:08

Right? And cars, after 100 years of proven utility or vehicles, right? And so let's assume cars, at least in the United States, $10,000, $20,000 is a car price point. After 100 years of utility, still remains considered purchase. And there is this gigantic amount of infrastructure built from financial infrastructure to enable those purchases, also repair, oil change, tires, all kinds of infrastructure available for you to maintain this as well. So now if you take that analogy to say, okay, there is a $10,000 humanoid, great. Why do I buy it today? Do I buy it for laundry? Do I buy it for dishwasher? Do I buy it for babysitting? How what is the lifetime? Is it do I expect it to work two years, five years, ten years? Uh, what is the maintenance requirement? If I want to maintain it, how do I take it? Do customer, you know, is the maintenance person going to come to my home and am I taking it? These are all the questions that customers care about beyond a few early adopters. Every one of the customers cares about these things. So there is a lot of that to be figured out. And then to go back to the earlier point I'm made, I think first DARPA self-driving car was 2006, if I remember correctly. So it's been 20 years, and we're still working on self-driving cars, right? Now, few things have changed. Computes have gotten better, algorithms have gotten better. So let's assume that the complexity of a humanoid, which is much more complex than self-driving cars, can be uh sort of a shrunk or can can be countered by speed and computes and improvement in algorithm. Even then, you're probably looking at five years, maybe ten years at least, maybe twenty. So we think it's far away. And there is a there is a uh there is an uh analogy for this. So history doesn't uh uh repeat, but it definitely rhymes, and consumers have to be incrementally led there. Consumers typically don't make a step jump if you think about how long it took us to adopt cars or adopt planes. It took for a long, long time. So, in the same way, um, you know, in 1995, there was a company called General Magic started by these two guys named Bill Atkinson and Andy Hertzfeld, who were the um one of the most distinguished engineers in the Apple Macintosh team. An entire ex-Apple Macintosh team tried to build iPhone. That was Tony Fidel, Nest co-founder's first company, and they massively failed. And what we eventually got was Nokia feature phones, and then we got the PDAs, and then we got the BlackBerries, and then we got the iPods, and then we were carrying three devices, and we're like, ah, I don't want to carry three devices. So then the need for iPhone was obvious. An iPhone came out and everyone was ready for it. On the flip side, today people don't even have a trust that the robot vacuum would do the job, right? They've been overpromised and undelivered for 20 years. So our thesis has always been that next five years, seven years, we have a job to go earn trust of these customers, these homeowners, to say, hey, I can sell you a robot and it's gonna do exactly what I think, what I tell promise you that it will do as a product, uh, and earn that trust. And if you have a three, four, five, let's assume, thousand dollar robots, all of a sudden if humanoid comes and says I can do all five things at 10,000, you might just consider it. So I think there is an element of building that need and under uh creating that awareness inside customers that we are completely skipping over, not just the technology itself.

SPEAKER_01 30:27

And when did you guys ship the first uh robot to the customer?

SPEAKER_00 30:32

November of 2024, Thanksgiving 2024.

SPEAKER_01 30:35

And you have shipped 6,000 units still now?

SPEAKER_00 30:37

Correct.

SPEAKER_01 30:38

And what's the current pace of shipping per month?

SPEAKER_00 30:41

About 2,000 units.

SPEAKER_01 30:42

Go. And how are you generating demand for it?

SPEAKER_00 30:45

Uh luckily for us, it's at the moment, it's all word of mouth. Uh, and and um, yeah, we haven't done any marketing whatsoever except the reviews and people on the internet, uh, especially on Twitter and X, just a lot of people who have genuinely good amount of followers, uh, just talking about how they love Matic, and and that has been really the great way to sell it. And it's only available in the US right now? That's correct. That's correct. So that's that's a deliberate choice as well.

SPEAKER_02 31:12

Yeah, yeah, we haven't expanded internationally. It's not that we don't want to do it. We absolutely want to ship to like we have so many customers pushing us, or like you know, potential customers pushing us to ship in Europe, uh, even in Asia, and even in India. It's just that you in a hardware business, you have to think of all these things we talked about, like you know, logistics, supply chain, get it working, make doing it properly, repairability, repairability. So we tariff.

SPEAKER_03 31:39

Yeah.

SPEAKER_02 31:40

So so we are just focused on US for like, you know, at least this year. And then you would see that as we start scaling up, we would want to go out in another international market.

SPEAKER_01 31:49

And uh, you know, you got some amazing set of investors. How did you get them from Sutter Hill to Collison Brothers?

SPEAKER_00 31:56

A lot of luck. Uh I think being in Silicon Valley gives you an opportunity. So we've been here since 2005, both of us, so about 20 years, no meet in two since 2006. Um and uh Collison Brothers, for example, when we were doing Flutter in 2011, we randomly ended up in their office. There was like I think Middle YC startup day, and one of the startups was Stripe. So we ended up there, and uh he uh John Collison was actually there, and and we just ended up showing him a demo, and then he helped us get into YC. So that's where the connection was formed. And at that time, we didn't know Stripe was going to be as big. Uh, you know, we thought we were doing something very, very cooler, honestly, than Stripe, but then very quickly we realized that they were doing something useful. We were just doing something cool, and there is an academy there. Um, but but that's where the relationship was formed, and we kept in touch with him. So in 2019, when we were raising our seed round, actually. Sorry, Angel Round, the five million one. Uh, that's when we initially got them to invest in us. Um, and uh they were they had already uh begun angel investing back then, and that's really how we got it. So a lot of it is through connection, a lot of it is through just getting to know people. Uh, I think Naval Rabikan the same way. Uh it was pure dumb luck that his office or angel list was literally a cross flowers office. So we got to know him that way as well. So uh on San Francisco.

SPEAKER_01 33:21

Your first company, you know, that saw large success like dot com, you were among the early team members, right? Got acquired by Google. And the second company that you were founder of, you know, Flutter, again got acquired by Google. So I see a common pattern there.

SPEAKER_00 33:38

Back to mothership. Um no, I think uh uh I don't need feel really jump in, but I think uh at that point in time, between 2005 to I want to say all the way till 2014, Google was really the only company really pushing the envelope and AI and envelope on machine learning. Um they were really ahead of the curve and looking into it. When we were doing Flutter, unlike today, where you just expect AI and everyone expected to work, back in 2011-12, no one believed machine learning worked. Um Need had actually so so the story is that um Lite.com came after I mean sorry, Google came after Lite.com acquisition was about to happen. I stayed through the acquisition and joined Google for a bit. Now Need actually left way before that. Yeah, and started working on Flutter.

SPEAKER_01 34:32

Okay.

SPEAKER_02 34:33

And and the whole look, so my background is in computer vision, and the whole idea was that uh I really believe that this is a space and domain, computer vision machine learning, which has tremendous potential. And the idea was to like, hey, if you start the company on your own, you're not gonna give midway, but you're gonna see it through in terms of technology working and shipping it to customers. And that's what we did at uh like at Flutter and got it working. It was the number one app in 72 countries for good six months of 2012. Um, like, you know, Apple even ranked it on their app store as one of the best apps of the year. Uh, and we got about like, you know, million downloads on the desktop app, which is almost about, like you can say, a factor of 10x when compared to the mobile apps, and uh got about like you know, close to 77 million gestures in a very short amount of time. But entirely honestly, outside of Google research, nobody really practically believed that machine learning and AI can work. And they were uh kind of at a forefront. Uh they saw the potential in the technology. We were state-of-heart in terms of gesture recognition and object detection back then, even after the launch of AlexNet in 2012. And what we built as a technology was eventually superseded in 2014 and 2015 with the with the mask, CNN, etc. So uh I think they saw the potential of what we were doing, and uh yeah, so just ended up um kind of uh you can say events just lining up properly.

SPEAKER_00 36:17

Right, plus at the right time. But but this was a different world to give to give a little bit of a credit to venture capitalists and Silicon Valley. Machine learning was popular in the 80s, never worked out. Machine learning was also popular in the early 2000s, so never worked out. So it wasn't necessarily clear that what changed now that would make it work in reality was that we were getting data, we were getting uh internet and iPhone cameras, we're solving a bunch of problems, but it wasn't obvious to you. And then when Apple picked us as one of the best stuffs of 2012, they actually approached us initially for acquisition. And then Google got interested. And by that time, uh we knew that uh this was really, really good technology, really, really good UX, but no one wakes up in the morning and says, I'm gonna buy gestures. So there was no business model, there was no way for us.

SPEAKER_01 37:04

What were you solving for consumers?

SPEAKER_02 37:07

So the the the what we were solving for consumers is that this is also the time when a lot of folks were moving to their personal devices for consumption of media, like you know, laptops, uh, iPads just came out, and people were switching it for watching videos. And our whole point was that can we prove and get working a technology where you can control your devices from a distance? So if you're watching a YouTube uh or a Netflix, etc., you can shush with the mute, you can give a thumbs up if you like something. Okay, you can just raise your hand, pause a video, take a call from your friend. Once done, you again raise your hand, start playing the video again. So it was our view of looking at like, you know, how we can bridge the the if if machine learning truly works, the the interactions would be multi-model. They would be audio, they would be natural interactions, and the video would all be merged together. And what so we got the technology working, but we also knew that this is eventually would be better solved in a setup box under your TV, so that if let's say Super Bowl is happening and like you know, someone is calling or something, you can raise your hand, pause the online streaming media, just take a call and then raise your hand again. And we build amazing technology, but we were way ahead in terms of an adoption. The setup boxes, the cameras still haven't happened even after 12 years, in some sense. And we realized that we build amazing technology, but not solved the business problem.

SPEAKER_00 38:41

Yeah. There was there was a period in 2011, uh, around 10 when Kinect launched. So Kinect still probably remains the fastest-selling consumer electronics device. In the first 60 days, they sold 8 million units. And what was the device? And it was just device to create a gesture detection, right? So if your Microsoft Kinect was, it came with uh uh their uh game box uh game gaming console. Xbox Xbox and you could use Kinect as a gesture detection, but what they did was they turned your hand or finger into a mouse. This is typically how gestures are done, yeah, right. And we had this insight, and there were two insights actually. Namnid, with his computer vision background, was always saying that look, we shouldn't need time of flight sensors. We should be able to just doing do things with the RGV cameras, which is in sort of when you add a time of flight sensors, what you're doing is trying to build better eyes. And Namit's point of view was let's build better brains, better algorithms. And then the second part was as Namit alluded earlier, that whenever people do to do this, uh turn your finger into a mouse, we just felt like it was really cumbersome that if you have a TV and you want to mute it, raising your hand up in the air, trying to find the mouse pointer and then air clicking a mute button is really cumbersome. So why can't we just say shush? So the idea was as computers computers get smarter, they should understand us, yeah, not the other way around. So we were relatively ahead of that curve in a technology curve. Um there was a period, if you remember, where Samsung TVs and other TVs used to come with cameras as well. So there was this idea that maybe gestures will be ubiquitous, but then it just changed. Mobile took over.

SPEAKER_01 40:14

Got it. And uh if you can share roughly how how big was the acquisition during that time?

SPEAKER_00 40:20

Yeah, so so we raised about one and a half million there. Uh so it was very early on back then. There were no pre-sale funds, there were no near funds that we could raise from. So it was actually a very different world. Um, and then we were acquired for about 35.

SPEAKER_01 40:33

Okay, amazing.

SPEAKER_00 40:34

Yeah.

SPEAKER_01 40:35

And were Collison Brothers and Novas are we can't invest in Flutter also?

SPEAKER_00 40:39

Uh no, no. Flutter was very different. Flutter uh uh, we had good investors back then too, but honest answer is Flutter was very different, meaning that if you kind of roll back time to 2011, I mentioned there were no angel investing, maybe there were 50 angel investors, there weren't that many big things.

SPEAKER_01 40:57

Because people hadn't made their money through angel investing. I think it's only through Twitter, Uber, and all these kind of and they didn't have a deals access. Yeah.

SPEAKER_00 41:04

So angel investor investment also wasn't a thing because there was no angel list or there were no platforms, or this idea that you can invest in the angel wasn't obvious to a lot of uh entrepreneurs who had made money as well, right? I mean, there were Google Angel investors for Google things as well, but this this risk appetite hadn't sort of uh come in as well. There weren't as many pre-seed or seed funds as well. So we had to go to big companies, and we were actually really, really struggling to raise for a long time for Flutter. Okay. Uh it was really hard. And it and Namita had done a phenomenal job of actually building a demo before I even joined him. Uh so there was an amazing demo, but even then we were struggling, and what happened is sometimes you get lucky. So I my first job was at Salesforce.com. One of the colleagues of Salesforce from Salesforce.com had gone and uh started this company called Opta. Uh in yeah, and uh Frederick Christ, right? So he's he's very popular, value and loan. And Andresen Horowitz had just started and they had funded him. So I reached out to him and he made an intro to Mark. And Mark Andreessen at that time had a thesis that gesture would be ubiquitous, similar to us. So we were just at the right place at the right time. And then and once we got them to come in and they didn't really put in a lot of money, it was just hundred thousand dollars. That's it. That's it. Uh they don't do a lot of seed round. And out of their seed round, like you know, they had this seed round, like a seed fund, yeah. So it was just a small amount, but the fact that Andreson invested just just yeah, we were as I said, we were struggling to raise funding, and all of a sudden Mark and Andrews and Horowitz investor, and I think week later we had million and a half raised.

SPEAKER_01 42:46

So before Sato Hill came in in the 60 million round that you're Matic, yeah. Formatic, yeah. Uh how was your journey of fundraising for Matic?

SPEAKER_00 42:55

So for Matic, okay, so so you know, both Namneet and I talk about this. We are not necessarily the best fundraisers in the world. We have not been able to ever raise just based on a deck. For us, funding has always been an output, not an input. So, first time we built this uh uh amazing demo, as Namneet was telling you in 2019 and raised our angel round, and we could show everything that we are doing in Magic today, we could show back showing back. But there was no deck, there was just a demo. There was a deck, but demo was the key piece of the puzzle that it's not just this vision that hey, we have proof of concept. So 2019 was a proof of concept. By 2021, we had working prototype. So it was 90% there, I want to say. And but in in robotics, 90 to 99 is 2x the work. Maybe even 99.9 is 3x the work. So, so uh we had that demo and we had the white version, the exact in 2021, Matic looked exactly the way it looks today. So we had industrial design, a lot of these things figured out, and we built this demo um uh just a 30 second wait, it was maybe a 90 second video. And Showed it to a bunch of our angel investors, and luckily for us, uh John Pat and Patrick Collison and then Ned Friedman and Daniel Groves, they were just really interested in investing, so they just came in and took the whole round. Well, that was a five million. That was that was no so the first 2019 was five million, uh uh 2021 was 23 and a half million dollars. And all all by angels. All by angels. And then next?

SPEAKER_02 44:24

Next was uh next was the bridge round, yeah. Yeah, so next basically uh we next we there was a time between 2023 and 2024. Uh like 2024, I'll say, when we struggled, uh was meant to raise again to raise. And in this case, it's like, hey, we were on a different chipset before switching to Nvidia, and we were realized that the software for the chips was not really good, and we are getting bottleneck that even if we try to ship the product, we will die very, very soon because we can't iterate fast enough from a software and algorithm point of view, because the compiler for the chip for the NN accelerators was very bad. And we made a decision to switch away to NVIDIA. So there was a time when the whole product is kind of ready, and in 2023, we said we're gonna redo it. So we changed our song, we changed all of the memory layout, etc., because we changed the psalm, we changed the cameras, we had to redo all of the platform work, the firmware work, we had to how the stuff is laid out in the memory that has to be redone because now you have a different system which has a different kind of APIs, different preferences on how the stuff should run on the memory. We do entirety of the EE, you can say, take the guts out and put the new guts back in, entirety of the platform software, the compilation of the code, etc. Put the new guts back in. Got on it done in like kind of a summer 2024, and then we knew still then, like, look in robotics, where we are, when you're so close to the product, you got to ship the product and get it working. We focused on just shipping the product at that stage. And this is also a time, luckily for us, uh, to some extent, to Mehul and I, we had uh capacity to continue to put our own money in. So that's when you went from one and a half million to 35 million. Well, we did we did five million in the previous round also. So we did one and a half first, then we did five million with the with the when John, Patrick, Nat and Daniel came in of our own money, again, like skin in the game thing. And then we like, okay, this is we're gonna have to we have to see it through till we ship the product. So we kept putting money in till we ship the product, and even after that, because just not that the product you ship the product every next day, everything is gonna work. You have bugs, you have issues, you have features that you have to still ship, and we kept pushing all the way on that front. And the what it happens is that in the valley is like you know, people bought the product, they used the product, and slowly and slowly it started taking off.

SPEAKER_00 47:07

And that's yeah, and and then the way to think about this is this is not the struggle, it wasn't just like you know, it's not investors' fault or anything. We just took longer to ship. Um, we ourselves did not think we were gonna take this ship. Like the joke inside Matic is that oh, we were only off by one digit because we're gonna ship in 2020 and we were already initially at least in 2017, we thought that we'll ship in 2020 and we were shipping in 2024. So it just took way longer. In hindsight, looking at everything we've done, like kind of go back in time, and aside from this nine months where we sort of made a mistake of picking airline sort of NVD and had to switch, I can't find any time where I could shrink our timeline. So just robotics just takes a long time, and and we understood that. Like I said that if hardware is 10x harder than robotics, I mean sorry, software, robotics is 100x harder than software because you're doing hardware, software, platform algorithms, everything. And autonomy. And autonomy. So full stack takes a long, long time. Um, so it just took longer than expected, and that's where we had to put in our own money and we had to ship, and that was really it. That hey, can you ship? And everyone was asking, can you ship as well? Um, and luckily for us, because of the patient capital, we knew that zero to one was going too hard. It turned out to be even harder than that. So we lucked out, and luckily we had an opportunity to do it. But we were seeing good signs. So you mentioned that during that round, Ashish came in, right? So one of the first persons to you get uh our an NVIDIA-based robot was Ashish. He had gotten one of the very, very early units, used it, and he saw the potential. So there were always glimpses of that, hey, we are heading in the right direction, that customers are loving it. And we had actually built about 50 umbrella-based robots and put it in customers' home, and they barely worked in any sort of low lighting. They were just okay. And they were if it was bright light like here, it would work fine. But it was even slightly, even like a dust kind of lighting, it wouldn't work at all. And there were a set of people who kept using even that kind of robot for six months, eight months. So we knew that we were heading in the right direction. If we could just fix it, there is a genuine utility that people want to have.

SPEAKER_02 49:17

So before we shipped, like you know, in uh like uh just after Thanksgiving 2024, we had internally used the robot for more than two years. Yeah. We were iterating on the product all along. It's just the stack took a long time to build, like, you know, everything, just making sure there are no bugs, the crashes, etc., build tools around them, and then of course redoing it. But yeah, it just took time to make sure it matures up. So one of the downsides, which is like, you know, is like when you are coming in an existing market, we talked about existing market, new markets, right? The downside of an existing market is that you have to come with a minimum lovable product, which is you have to mature your product to a certain degree because customers have expectations after having used existing products. Like a very crude example would be, but very like you know, if Tesla has to launch Model S in 2012, they didn't put electric engine on four-cart wheels and ship like how initial cars were shipped in the 1880s and 1899. Like, you know, like no roof, like it's like a horse cart, but you add a motor engine. They had to make sure the cart looks extremely beautiful, like you know, industrial design is top-notch, everyone just looks at it whenever it crosses three, that it has 200 miles of range, has it like you know, the acceleration which is through the roof, the inside looks modern, has uh the safest safety rating of a car and the crash rating of a car. So the seat belt works, the paint is looking nice, the things, basic things are there. So essentially the downside of an existing market is, and that's why perhaps very very many people try to avoid it because you it just takes longer time to mature a product.

SPEAKER_00 51:04

That's correct, yeah. So so that's what we did, and then really the last round came together again. We shipped uh um a lot of unbiness to us. Some people, uh, one of the managing directors and partners at Sutter Hill Ventures uh named Pete Schlam got the robot and he fell in love with it and he started talking about it internally.

SPEAKER_01 51:25

So it wasn't inbound from Sutter Hill?

SPEAKER_00 51:27

No, uh yeah, so so so not well both. So then I also knew another managing director at uh in Sampular, and we got him a robot as well, and he liked it. And then they just started talking about that, hey, if someone is interested, we'll let you know. And turns out Vick Miller, who ended up leading over round, and he was one of the managing directors there as well, was interested in Matic. He came in and spent about three and a half, four hours just walking around, looking at everything we did. Uh, we didn't have any deck or anything, but just based on the product and how we were making progress, it sort of came together.

SPEAKER_01 51:58

And how how fast was this process of you know getting the first Matic to the to the first managing director and then closing the round?

SPEAKER_00 52:07

Well, I don't know when Peach Schlam got it, but from moment Vic came to I guess close was maybe four weeks, three weeks, four weeks, yeah. Very fast. I I think uh we we we knew Sarah Hill's reputation. Yeah, we knew that those guys are like the best way to describe Sarah Hill Ventures, and and I think I had mentioned that they win they are one of the first and the original Evergreen Fund in Silicon Valley. They were become they became Evergreen Fund in 1978. And since 1963, they've returned 35% IRR every year. So that sort of track record is just mind-boggling. And then they incubated Pure Storage, they incubated Snowflake. I mean, they owned 23 to 25% of Snowflake, I think, when it went public. So very well reputed fund, very well respected guy. And and because they mostly incubate, uh, they are sort of builders disguised as investors. So we knew their reputation, and their thesis is exactly like ours, where they were they tend to take technical risk, not market risk. And that's precisely what we had done. So it just worked out.

SPEAKER_01 53:10

But but let's say iRobot filing for bankruptcy, did it you know, change any of your plans?

SPEAKER_00 53:17

If it doesn't change the plans, it increased the demand, number one, and then it in it sort of increased interest in us as well, because for all practical purposes, we are the only consumer robotics company that is shipping in the United States, and and ironically, the largest consumer robotics company in the United States, even at this scale. So there isn't anything else out there. And in fact, I would actually uh uh venture, this is one of the other pieces of the puzzle that a lot of people miss about robotics. There is only the largest category in robotics is still just floor cleaning robots, robot vacuums worldwide. Um, any guesses, you know, I'm gonna ask you, any guesses how many Roomba's iRobot has sold in their entire lifetime? Maybe 10 million. 55 million. Okay. Um 20 million, 21 million robots were sold uh in 2024. We don't I don't have a 2025 number yet, but 221 million robot vacuums were sold in worldwide in in 2024. So sheer scale there is generous. After robot vacuums, the next highest category is Amazon Kiva Robotics, which is really just warehouse robots uh that can take pallets, also very much like Roomba's flat robots, right? Those are 1 million. Okay. Then if you kind of go down, all the picks and uh uh uh sort of uh uh uh packaging robots or any sort of you know humanoids or anything, right? That at least in the US, none of those uh robotics companies have scaled beyond maybe hundreds or maybe a few few thousands here and there. I think Boston Dynamics had shipped shipped about three to five thousand robots in their entire lifetime, if I'm not mistaken. So the scale for robotics is not there anywhere except this category.

SPEAKER_01 55:06

And when you both combine together to put 35 million in the company, why your family is worried that like a large part of net worth Well, we're both a little crazy.

SPEAKER_00 55:16

Everybody thinks we are a little crazy, and wives are also has given up on us, so that works out.

SPEAKER_02 55:21

But no, it was it was I think I think the honest way of saying it is that look, um, just like how each company is just not us. Like, you know, Matic is not us. There are very many capable people which are working at Matic to help us succeed. Okay. And we rely day in and day out on them to take the right decisions, working through things. Similarly, I think perhaps we both got lucky because in the sense that our wives were very they knew this is not won't be an easy journey, but they were very supportive of us doing the effort, right? Of us saying that, hey, you have worked on it for years, you have, if there is a clarity of the vision, etc., and kept supporting us throughout this endeavor. Would we say that, hey, it was the easy journey? No. Uh, but at the same time, I think honestly, we were not here if without their support and kind of unwavering support throughout this time.

SPEAKER_00 56:15

And if you kind of put the investor head on from our perspective, we had we if you have money, you're gonna invest somewhere. Whether you put in a bank account, maybe in stocks somewhere. Well, it turns out we have the most information about Matic, right? Even our wives were using that robot. Like I would, even in that before we ever shipped, my wife was already relying on it. And every time if I tested it and put it on some uh beta release where it stopped working, she would get mad at me. So even she saw that there is a utility that we are actually heading in the right direction, so that conviction was there. We had the most amount of information. So it was very much this, you know, obviously we want to have a skin in the game, but we also knew that we were heading in the right direction, that this is some what we are building is correct, and at some point you have to put money where your mouth is. So we did. Yeah.

SPEAKER_02 57:03

And look, I want to put put some emphasis here in the same thing. It's not that we generally like, oh, every single founder has to put money in the company. It's not like that. I think it's just that we we did certain things because we felt like this it will take time to mature optimic. And it requires this long-term viewpoint. And we genuinely believe that we need to have a skin in the game. But as the our capital requirements go up, is we it's not that we have like you know billions of dollars in the bank that we can continue putting money in, right? So it's just we internally felt this is the right thing for us to do at that stage in the company, and that's what we did.

SPEAKER_00 57:46

And also we've still raised relatively much lower amount of money than many of the human art companies or many of the human home robotics companies that have been funded lately. Or I would say hardware companies, or even hardware companies, like I think Humane had raised a lot more. Uh 400, 500, like yeah. Yeah. So so one of the things that does happen when you put your own money is capital efficiency becomes a priority. And that does force you to do hopefully very good decisions. I think we generally say that constraints drive innovation, right? So this is also one of the constraints.

SPEAKER_01 58:18

Yeah. And they say outside of the matic equity that you hold, the money that you put in, how much percentage of your net worth had gone into matic?

SPEAKER_02 58:26

Oh, like I would I would say it's kind of uh pretty high. Like I would say, yeah, 70, 70 percent. If it is, in fact, it's always there is no plan B.

SPEAKER_00 58:35

There's only plan A.

SPEAKER_01 58:36

Except the home and matic.

SPEAKER_00 58:38

That's really it, except for the home and matic. Um, yeah, it's actually uh, you know, um so about eight months ago, uh my my dad passed away, but he's a businessman. And throughout the journey, he kept saying that if you had to sell your home, sell your home. But you have to do keep Madic alive, and you have to keep going Madic, which is when you have a conviction, go for it, because homes you can buy again, companies you can't do it again. So there was that, you know, I I'm very lucky to have him in my life, but there was this mentality that hey, and we never had to get to that stage where we had to sell sell homes, but but that was the mentality that there are in any business you're gonna have a near-death situation, any business journey, adventure journey, you're gonna have a tough time. And how do you survive? Uh, which is how do you find Nemo? You just keep swimming. How do you succeed? You refuse to fail. I think there's a lot of those cliches are a little bit true in some ways, and we just we just knew that we had to keep going.

SPEAKER_01 59:31

So today in a $1,200 product, and this is without any subscription, right? Is it just one time for $100? Uh, what is your cost that you have to incur, like in terms of parts majorly?

SPEAKER_00 59:44

Yeah, um, so we would be gross margin positive if it if it wasn't for tariffs. So we've been pretty deliberate about doing that. And I think if we keep doing what we are doing, we should be potentially cash flow positive very, very soon. So there's a path to get there. And a lot of that involves how do you do it. So one of the choices we made, for example, is we sell it, you mentioned earlier, right? Like we sell it directly from our website, uh, and we just do built-to-hour model that helps us with working capital, number one. Uh, so turnover in Indian Indian parlance. Uh, so we have you know neutral to reverse working capital if we can keep doing that. And then second thing is is um if you go through any retailers, you do give 30 to 40% of your margins away, right? Now, one of the most underrated innovation of Tesla is actually the direct uh sales model. They have their they don't have a dealership, and that saves about 30%. And if you look at electric cars as an industry, no one except Tesla makes money on electric cars. And Tesla's margin is about 17%. So if Tesla was going through dealership, even they wouldn't make money. Yeah, right. So so you sort of have to look at uh what has succeeded, what has worked out, and you have to replicate that model. So we knew that built to order is the only way to go do it.

SPEAKER_01 1:01:05

And how did you price it at $1,200?

SPEAKER_00 1:01:09

Both. We did some research. Um, I mean, and initially, uh I think we made yeah, but we made a hard price. So okay. I'll take a step back. Marketing, pricing, all that stuff we expect to get right on day one. Learning is that every product is different, every category is different. Most of the pricing theory that you may learn in business school is sort of narrative policy. It's a reverse work there. Ultimately, you have to go out and test it out. There is a range, we couldn't have priced it at 2500. So obviously that's a wrong move. But we did have this ambition, and at least I would say this food I would put on me. I had this dream uh that we would be able to do robotics as a service model. That was uh so we launched that way in trying to put it in. That was interesting. Yeah, I had I was hellbent on trying to do it, and this was my wishful thinking, if you if I were to say it, but we launched that way, and that was like an organ rejection.

SPEAKER_03 1:01:59

Why?

SPEAKER_00 1:02:00

I think there is just this people one, at least in America, people are just tired of subscription. Um, I don't know about the engineering. They don't want to pay for subscription. It's just it's just another another thing I have to pay monthly service for, and that was one. And then most of the people, this idea that you instead of buying a robot, you subscribe to it. Most of the people came to our website and said, wait, why subscription? We don't feel ownership in that way. Yeah. And they weren't asking why Matic. They were saying why subscription. That's a wrong question to answer, number one. And number two, I think at least for homes, there is this innate desire to own. Like most people in the world don't lease their cars, don't lease their scooters.

SPEAKER_01 1:02:41

Prove that subscription model works and healthcare.

SPEAKER_00 1:02:44

It does, but those are very different things than owning for your home or personal devices. Like we don't we don't like renting our phones, we don't like renting our cars, we don't like renting our scooters. Uh, generally, people do not like renting. So there is this innate psychological desire to do it. So part of the thing that we kept hearing from customers that, like, even if I start initially as a subscription, why can't I just own it after two years or three years? Why isn't it mine? Uh and and we know this now a little bit more than even before, because now that we are inside home, we just see the sheer amount of connection that people are making with it, right? So Lenny Rycie just tweeted that his son always writes four names uh on or from his family, and Madic robots is one of them. So Lenny? Lenny's son, yeah. So it's like my kid, he's just saying he's posted a picture saying that Maddox robots is always part of our family. And is he an investor also? He came in and became an investor after he bought the robot or after he got the robot. So, so uh uh he came in last round. So that's where we realize so there's just this connection that people want to make with the robot, and people want it to become part of the family. And if you're renting things, it's not really part of the family.

SPEAKER_02 1:03:59

The point is that uh you we are we took technology risk, we took product risk. Product risk in the sense it's a white robot, bigger, taller, etc., rectangular shape, right? So we're there are a lot of modes which we broke with the product. Okay. We we we went in the existing market, so you can say we didn't take a market risk, but with the subscription, now you added one more thing. So instead of the questions coming up like why all these other things that like why cameras, why this kind of a size factor, etc., the first question becomes a pricing question. And I think that is not really the good first thing to do with the new product. Would there perhaps be a time and opportunity in two years from now when the product is scaling well, people understand it, they have to gone through the word of mouth at mass, etc. They would say that hey, I would rather think of it as a subscription, maybe we would want to explore it again at that stage. But uh that didn't work out. Then we changed the pricing and we struggled, like you know, how to price it. We were like, hey, is it a 1500, 1600? We also wanted to make sure we don't um like you know, we were still trying to figure out what the product would, what would be our bomb costs, like you know, what would be our actual cost of the cost of the goods sold. So we had an early estimate. So, like, okay, we don't want to be net negative, we went with that thing, and then as we mature, continue maturing it up, we're like, okay, we think we can, we know where the cost of goods sold will end up, etc. And that's when we put the price up of about like you know, uh 995 because the features were missing as the features introductory price, like you know, when the features matured up a bit, then we increased the price to 1095. And then when we're like, okay, now we have to do more work uh next year to do more pricing, so then we increase the price again and and tariffs and everything that plays a role.

SPEAKER_00 1:05:56

But but you know, I'll I'll go back to this point of view. I think that if you we are going in the existing market, in the existing market, whether it was vacuums or robot vacuums, you were just used to owning it. So changing that behavior actually is really, really hard. Um, and I don't there are not that many items that I can think of aside from maybe set up boxes or cable boxes that we actually rent.

SPEAKER_03 1:06:20

Yeah.

SPEAKER_00 1:06:20

Right. So there is just this desire for people to own. Um, it's a bit more innate for for whatever reason. I think, especially in connection to homes.

SPEAKER_01 1:06:30

Yeah, and any plans of launching in India, like or a person can take this product to India and you can you can you can take the product and it will work. So how do you manage the electricity and that that so it's already dual core?

SPEAKER_00 1:06:43

So so so so our doc is already dual charge capable. You just have to put the connect, like the plug, like you just have a simple outlet, simple adapter.

SPEAKER_01 1:06:51

Just so cool, you know. I'm I'm glad we did this podcast. Thank you again for being sure.

SPEAKER_00 1:06:56

You should come by to our office if you if you're off.

SPEAKER_02 1:06:58

If you're around, you would actually like the hardware, what it means to actually get the hardware done, you would see it in office.

SPEAKER_00 1:07:03

Yeah, yeah.