The Power of DePin: Revolutionizing Infrastructure Networks - TOKEN2049 Singapore 2024

all right hello everyone thanks for having us here uh my name is tashar a managing partner at multicoin Capital and just to kick it off uh would you each like to introduce yourselves and say you know a few words about your project starting in order sure uh hey everyone I'm abai Kumar I've been a core contributor on helium since 2019 I started off as a hotspot operator and I uh have been building the core protocol since um uh helium is a decentralized wireless network uh it's a network of networks that supports both iot and mobile and uh if you can find somebody with a backpack with a helium mobile hotspot you might even be able to connect to it if you have helium mobile uh I'm Jules urbach I am the uh founder of the render Network and uh render was uh kicked off in 2017 it's a um decentralized GPU Network we um supply gpus for doing 3D rendering and of course now ai uh it's evolved to the point where we do have major movies and television shows and nfts that are created on the system there's a lot of other layers with provenance and uh we're pretty excited about the future for a decentralized compute Ariel Sidman uh CEO and co-founder of Hive mapper so we're building this decentralized global map so the basic idea is that we've built this device it looks and smells like a dash cam but inside of the dash cam is technology that is purpose built for mapping um and there's many many thousands of people all over the world just driving around with this device and building this map we've in the last uh little less than two years map 25% of the global Road network uh and it's growing even faster today than it was even 6 months ago uh and then on the other side we're building data products for businesses and then ultimately consumers right so think Automotive companies Logistic Company transportation companies all utilizing and integrating that data into their systems and into their products hi I'm Mike Horton project creator of geonet geonet is a geospatial network that um consists of it's a deepend decentralized physical infrastructure Network that consists of people setting up satellite mining stations and these stations collectively provide a global hyper precise location service so that location service can pinpoint you to about a centimeter in XY and this is an incredible utility for everyday navigation but also a fundamental piece of infrastructure for devices like drones and robots and the next generation of uh artificially intelligent autonomous devices all right so deepin has been getting a lot of attention recently as a sector and a lot of people are are really excited about it so I just want to start from the beginning of uh you know why is deepin interesting what does it enable that you could not do if you were trying to build your company in a centralized way um and uh you know maybe we can start off with helium here since you were the first dpin Network second I think live PE was before us but yeah um yeah like we're definitely the largest at this point you know at scale we've uh been around since 2019 is sort of when we launched the the network uh as a blockchain network um you know I think the the real unlock for us is that it enables uh sort of capital formation and distribution of this kind of network in a in a way that's really never seen before um the typical way of building telecoms is to uh get some real estate build a tower on it put a radio on it and then every time you need to update it someone you know from Verizon or in the US has to go uh call a local cable operator to go climb up that Tower and update it that's that's the sort of typical model and you know one of the things that we when we started our iot Network we needed a way to rapidly deploy this network there was no way that someone was going to use a new kind of iot protocol uh without uh having youbi equitus coverage and this was a way for us between you know 2019 and and I would say 2022 or so to very very quickly deploy a network of hundreds of thousands of notes um and now we're starting to do the same thing in Mobile in our case you know my um my other business besides starting render is is a computer Graphics company that sells software to artists and we launched um a product that ran on AWS with gpus that were um you know there you there's about 10,000 of those we would run out of those gpus in 2015 2016 because the prices for data center gpus were so high and the costs you know for Amazon Were Always On Demand pricing right you couldn't get to the spot instance pricing which is on10th of that because there wasn't enough gpus so we actually launched render to solve a business case need which was there's hundreds of millions of gpus gpus that are out there even our own customers use them for rendering locally and turning that on and flipping a switch and paying them even a fraction of what we paid atws or what would charge for on demand pricing meant that we could have 10 times the GPU capacity we could offer that even faster because there's way more gpus on the network and the cost was significantly lower but on the Node operator side they were making more money than at the time it was 1080 TI mining ethereum so we paid more than that to this day if you're on the render Network and you have a GPU typically takes about a year for that to pay itself off so I think that now there's a lot of attention on GPU shortages and and all of that around AI but we were starting even before that with just normal you know generative um you know media CG rendering and that's of course exploding even more so with the demand for generative AI on top of that um but in our case we actually had a business reason to introduce that and there is no way that any of the centralized Services could offer that to this day we have AWS asor Google as partners they contribute their gpus to the network you know with a high security level um but almost none of our customers use that they're they're rendering fullon you know high-end tentpole movies on average users gpus on the network with endend encryption so it's been an interesting journey to go from uh from that business use case to where we are today um so in hyp mapper case I think there's really two unique things that deep in enable so one is location and then the other one is coordination so look if you're mapping the entire world who is better to map your neighborhood than you right the people who live in that neighborhood and so what this enables us to do is put the technology in the hands of the people who live locally in that neighborhood and say hey look you can actually participate in building this neighborhood versus like us sitting in California and like sending a driver out to the Philippines or to Soul Korea or here to Singapore so that that's really powerful the second thing is coordination right because we have this incentive layer we can basically say look you know mapping the same freaking neighborhood a thousand times that's actually not very valuable but if we can say okay we need more coverage here or we need less coverage here we need more frequency here we can do that because of this incentive mechanism which is baked into the token that we have um the last part is I think is pretty interesting for those of you who probably have used ways in your in your in your life there's really two types of participants so first of all you just driving you are actually giving ways a tremendous amount of data for free but then there's this other group of people we call the map editors in the industry and they effectively sit in front of a computer and literally edit the map right so they draw lines effectively and there's about 30,000 of them in case of ways when ways was acquired by Google those 30,000 people that effectively made ways right they got nothing right because of this lack of coordination lack of incentives lack of sharing economics that something like Hive mapper embeds inside of it from day hero so those are the three things that I think are pretty unique to us yeah so I think that it's really underappreciated how incredibly Capital efficient deepen based infrastructure is and then once that infrastructure is in placed what a massive Economic Opportunity it creates for the downstream users of that Network when we look at geonet people have been trying to improve GPS accuracy for 20 years and the major barrier to it has been the cost of deploying physical infrastructure and traditionally to deploy High Precision location infrastructure was sort of on the order of $100,000 per node I mean it was a very significant maybe 50 to 100K per node so the fact that geonet has been able to install over 9,000 of these things through contributors to the network who are setting up these satellite mining stations we estimate that we've you know built out infrastructure that's sort of a billion doll level of capex that would have been done and that's happened in two years incredibly fast and with you know the foundation taking in about 7 million inventure Capital to do that so the capital efficiency is just phenomenal both in terms of the amount of money it saves to deploy infrastructure through a deepen model the fact that the participants can get upside through rewards um by deploying this infrastructure and then all the downstream Economic Opportunity for things new classes of applications that can be built because the infrastructure is there that wasn't there before so it's just in a phenomenally Capital efficient way to get get stuff built yeah Capital efficiency I think is the core of the thesis for deep in and Jules I have a followup to uh what you asked because everyone in the world is talking about Ai and everyone is talking about the GPU shortage y um can you talk us through how render helps address that and you know maybe some numbers about what's available on the network yeah and you know how the pricing might compare so you know the the generally render is tapping into consumer gpus of which there's hundreds of millions and we have a weight list of about a million and these gpus are highend Nvidia gaming cards like the 4090 the 390s those have 24 gigs of RAM um what's in a Data Center and what what everyone's sort of hoarding these h100s that are you have this expensive interconnect those are those are effectively um super expensive I mean even if you wanted to just do rendering the data center version of that consumer GPU is 10 times more expensive that's why we first ran into this problem when it comes to AI though the trends are changing pretty drastically so for example open AI put out 01 their new version their new model for chat GPT it's much smarter uh it takes a little longer but open I put out a graph which is we actually can take less time to train the cost of training um can be replaced with the cost of inference in other words running it you know in real time on a lower end GPU and it gets the answer better and so that Trend where you don't need to do as much training in an expensive data center and the training sort of happens in real time as you're running a query is where I think things are going and I think that's where having this massive sea of consumer gpus with enough memory to provide that service is going to be huge for the future of you know GPU training and inference and I do think that the cost savings will be enormous and that's where we're seeing first with generative video um and images right that's fairly easy to sort of take out of that that system we have um you know aad who started stable diffusion an adviser to render sort of helping guide a lot of that um that Journey but I do think that the um that the world that we're also heading towards and this is why also I started render is towards these um world models where you can simulate games you can simulate physics you can do of course video tell stories but all those things are moving towards this one multimodal model and I think those will be running largely on on highend consumer cards with the training step happening really from users data from a lot less of a pre-computed training step than you have now that is what's costing those hundreds of millions of dollars thanks that that that's helpful uh Ariel I have a question for you on how deepins tap into real world demand yeah dollar denominated demand you know actual like customers coming into the network and not circular within the crypto economy yeah can you talk about some of uh how Hive mapper does that and how you might see other sectors of crypto yeah crazy we have like real contracts with real [Laughter] dollars um yeah so so I think the the best way to think about Hive mapper there's really two parts of it right there's the protocol layer that does all the incentives all the coordination that's one part and then there's Hive mapper Inc the business that looks like basically every Silicon Valley startup that you could ever imagine that is building data products on top of all the data that is being collected and processed by the network right and so you know you got to ultimately show up to a customer's door right you have to build products that enable them to integrate our Pro our data products into their systems and then you know you need to support them right they're not going to be like hey I'm going to if something's going wrong they're not going to yell at the network they want somebody to yell at which is us right and so you know from that perspective and a lot of customers are interested in the crypto side I think that one of the things that when we got into this people are like ah like normal businesses you know like car manufacturers they're not going to deal with crypto stuff and I've actually found the reverse right I think they're curious about this right they're like hm that's interesting so let me learn more about it but ultimately when it comes time to doing business they want somebody's neck that they can actually choke if something's going wrong right and so fundamentally you have to be able to do both of those things and you have to do both of those things really really well um Abbi I have a question for you related to touching the real world uh helium is unique amongst a lot of uh deepins in that you've actually tried to own the demand side with helium mobile can you talk a little bit about you know how the owning the demand side versus carrier offload and selling to other people who own the demand side and how the economics of that work for uh people who are participating in the network yeah I'll I'll even take a step back and kind of explain those two terms very jargony and maybe some of the audience would appreciate it so you know helium mobile is uh is a consumer product right it's uh if you're in the US um you uh sign up and you download an app you sign up you get an eim and it is now your carrier right and uh if you look at your your phone it says helium in the top left corner if you have an iPhone for example um and so you know to Ariel's Point like the you know the the throat to choke there is there is a customer service team that will will you know answer your emails when your phone service is down and that's obviously very important for us on our phones um and that is for me I I see that as as a way for us to dog foot our own network um it's a w with you know over 100,000 subscribers uh over you know the past just few months um we've been able to see how to build a a new kind of network one that both leverages uh this this sort of new supply of coverage um people deploying helium hotpots and also still fall back to a traditional carrier like T-Mobile um which you know we're partnered with and and and we're able to provide that global Service right so even though you come to Singapore you are using that sort of T-Mobile relationship with SEL in order to to have coverage here so that's that's one side of the business and that's a consumer side of business and and typically these kinds of uh businesses we're talking you know you're paying sort of wholesale rates for data um and you're you're spending anywhere between one and maybe $3 a gigabyte in the US sometimes up to five uh in order to to to be able to piggy back off the service this is what's known as an mvno um in the industry so that's that's one kind of business that that can be built on helium the other business that is the sort of underlying business is this what we call Carrier offload and and the idea here is that any carrier out there wants to be able to provide a better high bandwidth service to their customers wherever they are typically we think about high bandwidth applications as places with a lot of slow moving people um so think of sports stadiums malls uh promenad in cities uh and coffee shops and and places like that places where you might be sitting around and you know streaming Tik Tok or whatever um and these are these are the places where there's a a tremendous am amount of demand and the traditional carriers don't really have the flexibility to be able to rapidly deploy demand uh deploy the supply into those places and you sometimes see this with music festivals right um if you've ever seen um Coachella for example they might stand up a tower just temporarily for it and it's still horrible like you know there's still like 10,000 people trying to connect to one radio that's up there and and barely holding on for deer life um and so you know that is a very very different business what that that is a business that allow that allows other networks um you know major carriers around the world to be able to leverage this uh Greenfield Network this this network that's being built by this community um in order to supplement their existing user experience so for the AT&T user for example or the T-Mobile user or the singtel user here uh those customers don't know that they're using the helium network but helium hotspot deployers are getting paid for providing a service for this network that they also don't have a relationship with and this is this is sort of the power of deepin right it kind of goes back to that thing which is that that that point which is that you know these are all participants that are all incentive align to provide the best possible service at the place where they are and they don't really want to think about like who's authenticating to whom like all that should be taken care of that's a user experience problem that that should be solved by the by the network itself so Mike I have a question for you I think at this point it's pretty uncontroversial and empirically proven that deepin can scale a supply side of the network with token incentives and you know we have four examples of that on stage today and you know many more out there in the industry can you tell me more about how you think about scaling the demand side at geonet and um how that demand engagement function works for you yeah so in the in the field of high Precision geolocation there's sort of two core markets you can think about their traditional markets that have been using centimeter accurate positioning services that sort of invented the technology and those are things like civil survey uh High defition mapping construction farming and one of the nice things in our space is there standards that are strong and so we adhere to those standards and we're able to offer just a much better value proposition to the customers lower cost better coverage um easy to integrate and so that just lets demand naturally flow in we don't have to work that hard to create that demand and we've been able to add 30 B2B partners that are using uh the service and bringing it to their customers then because the infrastructure is so ubiquitous and has grown so quickly we can enable consumer facing applications and that's really super exciting for us cuz those are the first time that people have had the experience of having hyper precise location which I could have used a number of times here in Singapore where I found the the phone position to be quite poor and so we're looking into applications we have a new partnership with Anoka that will bring it into gaming uh bringing it in again to Consumer robotics I think we all have many iPhones I devices now I think few years from now you'll look back and see it's a turning point in the in the movement of Robotics we're going to have Smart robots mowing our Lawns for us we're going to have drones helping us with security it's really that kind of infrastructure that we have enables the service to be provided to new applications that are really popping up and going to grow revenue for the network tremendously U Ariel I have a simple question for you let let's just make things really simple for the audience why are deepin tokens valuable why is honey valuable yeah um very controversial question uh so I mean fundamentally it comes back to the demand side right like if if in our case we're building well let me take a step back like there are really two customers or two main customers for mapping right there's like all of you guys who use maps every single day in your life there's about 2.5 billion people who use Google Maps every single day and then there's probably like anywhere between half a million to 750,000 businesses that pay for maps to integrate into their products and services right and so both of those are being monetized in different ways and so if we can show up to the customer whether it be a businesses or consumers and offer a product that people want that they're willing to pay for or that we can monetize them in some way then fundamentally that flows value back into the token um and there's a couple steps for that to occur but fundamentally that is what creates value and I think that you know in the in the crypto the thing that's interesting to me about the whole deepen category as a whole is that it is I believe really the first thing in the crypto world that's extending Way Beyond on just people doing crypto trades of fun tokens here and there or store value type of stuff into the real world right in our case there's billions of consumers and many millions of businesses that rely on maps in the case of helium it's you know basically everybody with a phone is paying for wireless service and so if crypto is going to be relevant and much bigger than it is today 5 years from now 10 years from now like all of these products and services that you see here on the stage have to be successful cuz otherwise is just like a bunch of people like trading with one another which I think is cute and interesting but not very big by why is h&t valuable so um h&t we launched it back in 2019 it was given directly to hotspot uh owners who are providing coverage um and that same h&t is burned and used uh for uh by people that were using the network so initially it was an iot Network so if you had a sensor you had to burn h&t in order to pay for the network and it was a you know it's a a fixed cost that a sensor deployer would always know about um you know you send a million you know uh a million messages and it and and it costs a very small amount of money um a year uh on the mobile side it's it's the same right it is agent's issued um it's you know maybe purchased by a mobile carrier uh someone like hel mobile for example um and then it's burned to use the network so you know the the individual uh sub subcriber doesn't need to think about h&t and shouldn't ever think about h&t right they should be thinking about having great service and their service provider in this case um hey mobile uh is the one responsible for buying the h&t burning it and paying hotspot owners for their work got it so it's valuable because real world demand is used to then burn the token MH basically yep I would also add we we we copied that model I mean you guys did such a great job that we apply that render you know when render started it was you know the token was really just about mapping to compute costs we also implemented a burn mint equilibrium model which has followed what you guys were doing on and also went on salana and I will say that also I mean being in the in the web 3 space the crypto space is super interesting for media generation right I mean you're talking about you know trying to prove whether something is real or who made it more importantly though think about nfts like $800 million worth of nfts including BB poox all that stuff was created in our software and on chain and those things are provably every single texture every piece that went into that is uh is part of the uh receipt that is on chain so if you're thinking about the future of media where you know you have to sort of Safeguard your IP your faces as an actor but you also want to pay out royalties figure out the model Beyond streaming crypto is great for that I mean you can split the uh you know you the proceeds of somebody viewing something 64,000 different ways and pay all those different royalties out that's really hard to do outside of this system so in our case I think you know not just the sort of value of the token beond a BME model but also having this capability for provenance authentication royalties that's what crypto is really about and I think that's one of the things that makes um The dein Narrative on top of render also you know really valuable Mike would you like to add on geonet yeah same thing as helium bernon meant equilibrium is a very powerful way to translate direct usage and utility of the network back to um token holder value and I think as a network grows in utility the token value grows in and and value as well so it's a very direct I think burnman equilibrium is a very powerful thing our network is quite deflationary as well in the sense when we burn it we actually take it out of circulation by sending it to the burn address um we have a annual having schedule so these things help time sort of the buildup of the revenue with the buildup of the supply so that you can have a stable but growing ecosystem and I think that's um it works well all right I want to shift gears a little a little bit and ask you about composability uh and how you think about composability either with other deepins or with other protocols on the chain that you're on uh you know D5 protocols Etc and maybe RL you can kick it off sure um okay so remember I was talking about that device uh so there's a device that looks and smells like every other dash cam in the world and you install in your vehicle now we need to ultimately take the data we're producing on that device which is all the map data and we need to offload it so the question is how do you offload it there's a lot of different ways um but this new device that we're introducing uh this year uh that there's a lot of pre-orders for called the B uh that device will actually use the helium Network in the US to be able to offload all the data right so if you're a customer of Hive mapper that purchased the device the B device and you're just mapping then you could also go and utilize the helium data plan which is what 20 bucks a month uh very cost effective uh for that device and then boom you're offloading right and so I think it that's a good example of where two teams that you know kind of share a lot of the same values um they have great product great technology as well um kind of coming together and integrating so I think you got to be choosy about what you're going to spend your time on uh CU I think there's a lot of projects that just kind of go out there and be like I'm going to partner with this I'm going to partner with this and it's like it makes a good press release in the moment but there's not a lot of value for the customer on the other side um and so in this case I think it works great like we we want that like real- time data and helium is super cost-effective for that device Jules how do you think about composability with other things like nfts and uh you mentioned you know provenance and proving that an artist created something uh how how do you think about tying that all together using composability within a single chain well I think you know just just in the deepin space Right storage is really important um so you know connecting render to you know decentralized storage makes a lot of sense just on that first layer also having other chains that handle compute that can work within our system on on their own you know on similar networks or things that are running out salana makes a lot of sense as far as the the end point when you're selling something I mean you know I was bringing up the example of nfts because what you do with with provenance can be you know can create a million different types of businesses you can have of course storefronts where you buy digital art but you can also then create user generated content where you can borrow certain pieces and elements remix them you know do Advertising based um models I mean all those things can be built on chain and I think there's just a ton of of D apps that can be leveraging this so for us you know we want to make sure that there's um you know the protocol support that and then any other deep in projects that are you know really fundamentally interesting again we don't want to focus neily on everything but again and things like storage and also any sort of um you know even games right that that have been building Virtual Worlds I mean the truth is you want to have this kind of world model that we're sort of going after with render as the underpinning for that and I just see foresee just a lot of different applications built on chain around this kind of infrastructure so composability for royalties for authorship and for just jump starting whole new types of business models um and democratizing that makes a lot of sense for us abai earlier this year you um had a tweet around composability no it was a good one uh it was around composability for dpin and Defi and it was specifically around Jupiter and how many of uh Jupiter's users were actually from the helium ecosystem can you share with the audience kind of the content of that tweet and then also the conclusions that you were able to draw that data man that's a long time go I forgot what I tweeted about last night um okay I I mean I I I definitely remember the context uh you know we were moving to salana last year um and one of the many reasons why we moved to salana is because of composability because there are protocols that are building the important things that I don't want to build myself right as a as a protocol Dev right like I don't want to necessarily build the the the routing engine for all the different D5 protocols that are out there I don't want to build um you know you know some of the sort of staking things that like someone like jto might be building um you know I like these are the kinds of things that we saw in salana that were available to us that we were really really happy about being able to to use with Jupiter specifically um you know we have a wallet and one of the things that we try to do with our wallet is we um uh you know we want to it's a it's a fully non-custodial wallet um it allows you to uh claim you know manage your hotpots claim the the rewards for all the work that your hotspots are doing um and also do things like swaps and we were able to just very simply uh natively integrate uh Jupiter in our swap interface um we didn't charge any fees on it and we just basically used the Jupiter API directly and the Jupiter smart contracts directly and you know this was really really powerful for us when we moved to salana um because it was just there and it just worked and that team is awesome and you know some of them are probably around here um uh that was one thing that was really great for us the other thing that was great for us with uh with Jupiter specifically is a lot of folks are coming to crypto and sort of entering into these deep ends and that is the first thing that they've ever done they don't really think about like what is the you know gas token or the rent token for the chain that they're on um oh I have some hentt how do I use it how do I send it to someone I need Soul I'm not really sure what you know how do I go get it do I have to go to an exchange or whatever and one of the things that we were able to build into um into our protocols really we sort of hid the ux of of Jupiter here is to be able to do a gasless swap from h&t to soul and that's because of the infrastructure that they built and so that's the way that we were able to effectively onboard 750,000 wallets onto Jupiter when we migrated and and so what we saw in in practice was as folks were lazily migrating and moving their assets from our old chain to our new chain they were starting and Ed Jupiter from day one and we saw that we were um you know a a massive driver of their traffic and and yeah like awesome team and they were like hugely supportive of them and they've gotten much bigger than that since then we are now a small fish in their in their pipeline but you know it was great to have another team like that building in the salon ecosystem that was that was just so uh that had built a great product all right so for the last question I want to talk about ecosystems and how you each of you as Builders think about what ecosystem you want to be in and how you chose the L1 and the ecosystem that you're building in and since you kind of answered that question just now maybe start with yours go that way well you know we started as an erc20 token in 2017 and then it was I mean the gas fees were crazy and it was impossible to get node operators paid quickly and so we moved to tomatic you know not not polygon to didn't really work out great and you know our sort of um interest in salana was started because it was efficient I mean I love Raj and toley I think that they've been absolutely incredible as supporters of what we're doing and giving us all the resources we need to make that simple there's also other amazing projects right in salana ecosystem like what you guys were doing which helped us move to BME so I do think that the um choice for salana now it's obvious I think that three or four years ago when we were first considering it like in my first breakpoint talk was where I was like where're we're here because we believe this is the future for you know just efficient uh services and it's proven to be true but there's so much more in the salana ecosystem as you're pointing out in terms of all the other interconnects and uh yeah I think there's nothing else quite like it so for us salana has been been fantastic and it's been the right choice yeah for us I mean we started off in salana um I think we were Technic although helium is definitely the the George Washington of deep ends uh I think technically the first deep in on salana um and so we didn't you know these guys went through a much harder process of kind of migrating but for us it was all about um you know cost of actually building on top of that right like most people don't realize like we want all of our software developers actually building the map not you know building all of the crypto infrastructure blockchain infrastructure type of stuff and so we really just have one engineer who works on that right and and I view that as great right I wish it was only half an engineer's time um and so that has been just incredibly cost efficient both in terms of the transactions itself but also the total cost of ownership of the development cost that we spend in terms of building on top of it so so we love it yeah I think the ecosystem question goes and the chain goes a lot back to the composability uh question in the sense that if you put these Services together it's very important for them to share a common chain so we started out on polygon and we have projects that are using geonet for either proof location or high accuracy location um that are interested in building in the polygon ecosystem but a lot of deepen projects have chosen salana and so when you think about putting a composable service if we want the geonet token to be the the token of choice and the service of choice for mapping for example we need to be in the salon ecosystem so that's really the value that we see the cross chain as being able to support to use the geonet service you have to burn geonet tokens so that is much more effective and easy to implement for another web 3 project if you're in the same ecosystem so that composability drives a lot of the choice of chain and clearly in salana um ecosystem deepen is emerged as a huge Trend and and a really powerful Network for building those so that's why we're deeply um working to move into the salon ecosystem wow all right so thank you so much for sharing all your insights with the audience today appreciate it and yeah thank you everyone for uh listening [Applause]