What is DePIN? The Case for Decentralizing the Real World — with Guy Wuollet

Transcript

There are lots of things that it's a bad idea to decentralize. So, we're going to talk about why would you want to take a physical infrastructure network and decentralize it? What benefits do you get from that? One simple way to create a network that has good adoption is to get your potential users to not only pay for it in advance, but also to own and operate it. The potential to move from incumbent atrophying monopolies to web 3 products is a huge one and I think in many cases is a bigger change and shift than moving from pretty good web 2 products to even better web [music] 3 products.

Hey [music] everyone, my name is Guy Wlette. I'm an investment partner here at A16Z crypto and one of the things that I'm focused on and very interested in personally is deepen. Deepen means decentralized physical infrastructure network and it's a term used to refer to a category of projects that we think are pretty compelling. So to start with I'm going to tell you what DPIN is and why you should care about it. We would define deepin as any sufficiently decentralized network that uses cryptography and mechanism design to ensure that a client can request a physical service from a set of providers. And I can talk a little bit about more what that means in a minute.

To give you kind of my historical background or context here, I think it's helpful to start with how I became interested in this set of ideas. Back when I was in college, I was particularly interested in the idea of building a decentralized internet service provider with the insight that a lot of blockchain projects wanted to provide censorship resistance, but they only really did so at the application layer. and they didn't particularly consider what would happen if your internet service provider or government decided that it was now illegal or changed the rules around how you were able to access and participate in a blockchain network. And so if you care about one layer of the stack as being decentralized, if the layer that it operates on top of is fully centralized, you lose a lot of those benefits. Uh so I think it's helpful to first start with

what I think is one of the important distinctions between deep pin projects and call them blockchain projects or other decentralized networks which is how you verify the service that the network is providing. And so on something like a blockchain usually uh you do verification through consensus. Consensus mechanisms are you know a very wellstudied academic field and people will often confuse them with civil resistance mechanisms like proof of work or proof of stake. Proof of work and proof of stake ensure that you know a single computer or a single person or a single dollar can vote in consensus and you form consensus over some set of state changes uh to ensure that everyone participating in the blockchain agrees.

Some other approaches to verification for a traditional blockchain are uh succinct proofs often also called zk proofs or you can use trusted hardware like enclaves although some people would tell you that's a very bad idea. I won't you know give you too strong an opinion. The distinguishing factor is that those all operate over a virtual machine or an inherently digital uh service or agreement. So if Alice wants to send Bob money on a blockchain, you are agreeing on a purely digital exchange of value, something that can be represented in in binary or or digitally. Now physical infrastructure networks deal with things that are not able to be represented in a digital or a binary format. For example, if Alice wants to send Bob electricity

and receive a payment for that, or if Alice wants to send Bob uh a packet over the internet, which is a an information theoretic transfer, not purely uh a digital transfer of value. Uh another simple example of this is any ride sharing or food delivery protocol where you literally have to move from one place to another and hand something to someone else. You couldn't ensure that that was done correctly. couldn't verify that a ride sharing trip was completed in a purely digital context. And so I think that's one good way of delineating uh how we think about the difference between traditional decentralized networks and what we'll call decentralized physical infrastructure networks here. Okay. So let's walk through by example what at least I

believe should be called deepen and what should not be called deepin. So start with no because that's everyone's favorite thing to argue about. What is not deepen? Bitcoin mining. Bitcoin mining is not deepen. Decentralized AI projects are not deepen. They are purely digital. Cloud compute projects or cloud storage or cloud networking. People often call these things deepen. Why do they call these things deepen? I think they call that deepen because it has to do with literal energy where you're uh using excess energy from something like a a power plant or uh an oil rig to mine bitcoin which is a cool idea but is not the same as a decentralized energy project. People often call decentralized

AI or decentralized cloud projects deepen because they also do not use a consensus or a succinct proof uh approach to verification and they don't because it's way too computationally intensive to try to form consensus on the inference of a LLM or to try to form consensus on the sort of compute workload you would want to traditionally run in the cloud. So these sorts of projects use very similar verification approaches to deepen projects. So that's the the first reason I think people confuse these. The second reason is both of these projects often like to create a token and provide it as an incentive and a subsidy to get people to use the network. And this is sometimes what people colloquially mean by deepen is I

start a deepen project. I make a token. I give it to Alice. I give it to Bob. I give it to Charlie. Everyone gets a token and hopefully they do what I want and we build a network effect. Then I stop giving people tokens and I start charging a fee and we've now built a decentralized network. That's a cool idea. It's a good idea. Often helps you solve the cold start problem. It has nothing to do with physical infrastructure. Notice like no part of what I've just said relates even remotely to physical infrastructure. So these projects are very cool. I'm a fan of them. We have invested in all of them, but they're not deepen or at least they're not the deepen I'm going to talk about. So what is deepen? Energy projects, telecom and

transportation. So what is deepen? I think decentralized energy projects should be categorized as deepen. Energy is inherently physical. You could have some argument about digital signals or represented through electricity. That's fine. I think that's sort of a a useless abstraction. Decentralized energy is particularly interesting because the energy grid itself at least in the US is you could call it decentralizing or becoming more distributed from centralized power plants that generate most electricity and uh centralized utility companies that both generate and transmit electricity to all consumers into a world where people have solar panels on their house or batteries in their garage and are able to generate and store their own electricity. So we'll come back to that telecom. Sometimes people say uh

well if helium the most well-known defend project that is focused on Laurowan networking and 5G networking well I think of telecom as being a digital service. So why is this a potentially digital service in the physical infrastructure category? Well, the simple answer is a lot of what Helium innovated in was convincing people to buy a box and to physically install it in their home, small business or apartment building, which is inherently a physical activity uh and not a digital one. And the second answer to this is the service you're trying to verify in a telecom context is not a state transition function in a state transition machine. It is instead that one person transmitted information to

another person in a different physical location which cannot be verified in a purely computational or digital sense. And then the third which is hopefully the most obvious is transportation. And if the whole goal of the project is literally to move something in a physical space A to physical space B, hopefully you can understand how that would be a physical infrastructure network. Cool. So now that we've talked a little bit about what decentralized physical infrastructure is not and what it is, let's talk about when you should build a deepen network and why it matters. So let's talk about what sorts of networks uh physical infrastructure networks are naturally a fit to be decentralized because inherent in

decentralized physical infrastructure networks is that you have decentralized. There are lots of things that it's a bad idea to decentralize. So we're going to talk about why would you want to take a physical infrastructure network and decentralize it? What benefits do you get from that? The first and probably the most used or well-known is capex. So, one thing you can do if you're building a physical infrastructure network is instead of one company having or raising a large amount of money and then spending that money to build physical infrastructure, you could instead sell peace meal uh part of the physical infrastructure network to potential users and have them own and operate the network themselves. And this is precisely what helium which uh I would argue is the first modern deepin

project and likely the most wellknown uh did. So they built a Laurowan network. Laurowan is a long range very low power uh and very low throughput networking standard and Helium built the largest Laurowan network globally. Now there's a cold start problem in building a new telecom network which is it's very expensive to deploy hundreds thousands millions of boxes to uh comprise that network and you're uncertain if people are interested in using the network before you have spent all the money to deploy it. So no one had created a broad deployment of Lauraan at the time because it was uncertain that there would be good use cases for it. And Helium believed very deeply that there

were a lot of potentially quite good uh IoT use cases for a Lauraway network. And so instead of trying to find one company or person uh with a a very large amount of money who would buy all of the nodes in the network and deploy them individually, Helium sold Laurowan routers to anyone who was interested in buying them. Now it turns out this is actually uh like a fantastically useful approach. They sold uh more than a million of these routers and people were willing to pay in many cases hundreds or on sometimes in the secondary market thousands of dollars for a helium node.

Uh which is surprising. You would not think that people are really excited and interested in operating new telecom network. So why did this happen? Well, the first answer is there are a lot of people who are nerds or hobbyists out there uh that are actually particularly interested in participating in a network like this. Uh the second answer is that by buying a router and participating in the Helium network, you were owning some of the network. And so when you bought a router and operated it, you would receive tokens uh that represented an ownership in the network for operating your router. And it turns out people were really excited about this idea. So the market price of Helium tokens went up, which made the opportunity of buying a router and running it much more financially appealing. This idea applies

pretty broadly and is something that people often uh use to describe deep pin more broadly, which is as a crowdfunding mechanism. Another way to view what I've just described is to say uh what if you get all of your users to pay in advance for the product you're trying to sell them? Well, that will make it much easier to build a high fixedcost product. And it turns out that physical infrastructure networks often have very high upfront fixed costs. And so one simple way to create a network that has good adoption is to get your potential users to not only pay for it in advance but also to own and operate it. And that's one of the reasons uh we're excited about deepen is this idea of userowned and operated networks. The

second reason to decentralize a physical infrastructure network is a reason that you see all the time for things like uh DeFi protocols or traditional blockchains, which is this software idea of composability. Much in the same way people often say that uh compounding interest is an essential component of finance, composability is, I would argue, the essential component of software that has allowed it to innovate much faster uh than you see in physical services. uh there's the the often quip that there's been a lot of innovation in bits but not very much innovation in atoms and I think one reason you would say that is because software is composable whereas physical services or infrastructure is not and so if you build a decentralized physical

infrastructure network the hope is that you now have composability because anyone can contribute to these networks uh help build them and anyone can use them you're able to combine physical infrastructure networks to create a sum that is greater than its parts. In the same way you have money Legos for DeFi projects, you should have infrastructure Lego for DPIN projects. The third compelling reason I think uh more people should build decentralized physical infrastructure networks or when is a physical infrastructure network a good fit for decentralization is when there's a lot of latent capacity. I think this is another reason sometimes people refer to Filecoin as a deepen project or the the first deepen project and it's

because Filecoin had the idea of taking the excess capacity on uh anyone's hard drive and contributing that to a file storage network. One instance of this that I'm particularly excited about is the idea of building virtual power plants. uh virtual power plant meaning the aggregation of a bunch of small electricity generating devices like solar panels that from the perspective of the energy grid can look uh very much the same as a traditional big power plant. you as an individual, uh, a normal homeowner, a small business owner, you have very bad bargaining power when, uh, talking to someone who's participating in the energy grid, if you have one solar panel that produces a little bit of energy and would have no individual marginal benefit. But if you

can get together thousands, tens of thousands, hundreds of thousands of those solar panels, now you actually have a massive uh generation capacity and you have much better, you know, bargaining or negotiating power in terms of getting good economics and being a large participant in the grid. So you aggregate this kind of latent capacity that has gone unused. Often times it's much more useful from the perspective of someone who's interested in buying that capacity, but you also have much better economics in many cases. So, one of the ideas uh that we're excited about is the idea of uh transportation network as a a potentially impactful deepen project.

And uh you have existing ride sharing networks that look a lot like uh a decentralized ride sharing network. I I think the best example of this is is Uber uh where anyone can sign up as a a driver, anyone can sign up as a rider, and you're matched with someone. It's a a fantastic product experience. I think the difference here is that I would call Uber a distributed network as opposed to a decentralized network. And what I mean by that is the network topology is distributed very broadly. There's not a central service that provides all of the rides or a central customer that consumes all of the rides. So the network is distributed, but the economics and the control of the network are very much centralized. If you're a driver or a rider, you show up and it

quotes you a price and you have no influence on that price. You either take it or you leave it. If you are a driver, you do not get all of the payment or even in some cases a majority of the payment that uh a rider is contributing to the network. So all of the routing in the network is done by a centralized party. Uh and the drivers and riders in the network have no control over how the network evolves. Now a ride sharing network is very naturally a userowned and operated network. In the current case, it is a company-owned and user operated network. And making that shift would push a lot of very valuable economics and control out to the people who use and care about the network. And so the difference between something like

uh the current version of a ride sharing network in Uber and a potentially decentralized equivalent is that instead of paying people in dollars as a driver early on in the network, it'd receive tokens that represent ownership of the network. And as the network grows over time, those tokens become more valuable. So some of the early uh drivers or riders in uh a decentralized ride sharing network would perhaps make significantly more money and have significantly more control of the network than were they to be paid in in dollars. So let's take by example a decentralized energy uh network or investors in a project called Daylight that is trying to build a decentralized energy network. So this has definitely been influenced by them, but to try to

make it a bit more abstract and clear, uh I won't talk about them specifically. So one thing that's currently happening in the American energy grid is there's a lot more generation capacity at the edge of the network. This is true because people are putting solar panels on their roofs or batteries in their garages for an electric car. It's also true that it's very hard and expensive to build a single large centralized power plant and to connect those power plants to the energy grid takes several years in many cases. So if you wanted to significantly increase the amount of energy capacity available, it would be very hard to do that in a short period of time. This is particularly salient now because lots of people want to run AI workloads uh which are very energy intensive and there's no good way to add a lot of new capacity to

the American energy grid in a short period of time. Uh so you know the grid today maybe looks something akin to this centralized power plant. just makes all of the electricity and then it sends all the electricity to homes. Now, you can see how this is naturally a relatively centralized topology. Uh the alternative to this would be something like this where this is still your centralized power plant and this still makes a bunch of electricity, but now perhaps you have a solar panel on your roof and you don't need electricity flowing to you the same way. you actually perhaps sometimes sell electricity back to the power plant and to the grid more broadly. There are

other cases where you might have, you know, one home in the neighborhood that has a solar panel or many solar panels and generates so much electricity that other homes in the neighborhood don't need uh to get power from the power plant. This taken to its logical conclusion is just the idea of building a data center that also happens to be a power plant that happens to have a bunch of solar panels on the roof of the building or nearby. This is a much more decentralized or distributed network topology. And so moving from the old world to the new world, there's naturally an opportunity to change how the economics and the control of the network work as well. Something that's interesting about energy grids as opposed to other sorts of networks is the business model. So for a very long time, the way the energy grid has worked

in America is that there's an an energy company and they both make all of the energy and transmit all of the energy to you. And by analogy, this would be like if you know YouTube and Google and Netflix owned both the internet cables themselves and the movie studios that made all of the content we watch. Uh that's a very different world than the world we live in where you know content is built is created by many different people all over the world on YouTube and a different company owns the internet cables. So naturally in this kind of more decentralized world you end up having an opportunity to change how the economics work. And so if you contribute your solar panel or your battery or even information about uh the state of the

energy grid at your home, you are able to receive a token reward that represents an owner uh an ownership stake in this call it virtual power plant. And now all of a sudden if you get uh tens, hundreds of thousands of homes, millions perhaps participating in this, not just homes, small businesses potentially, uh apartment buildings, so on and so forth, you're able to sell energy back to the grid with much more appealing collective bargaining potential. I guess the other version of this would be no one is interested in buying just your electricity because it's not enough to make a difference.

But if you get a bunch of people together and say, "We're willing to sell our electricity back to the grid, uh, then [snorts] that's a much more compelling value proposition." One important note both for telecom networks and for energy networks is this touches up against traditional regulation in a bunch of ways and you could obviously see why that's true. So for a project like Helium, they're now mainly focused on uh 5G connectivity and they use CBRS or citizens broadcast radio service which is unlicensed spectrum. Normally if you want to use frequency spectrum for telecommunications you have to go to the government. They have an auction for that spectrum and you pay them a lot of money and they say only you can use that spectrum. Uh and one of the reasons

helium has become popular and possible is because there is spectrum that anyone can use and is freely available which is CBRS. Uh there are also a lot of important regulations for energy grids about when you are allowed required or forced to sell energy back to the grid and what price the utility company or other takers in the grid are required to pay you. So in some cases the utility company which is a highly regulated entity is always required to sell energy to its customers which is a very strange thing because perhaps the demand for that electricity fluctuates wildly or perhaps the ability to generate electricity fluctuates wildly. Uh and so oftent times energy companies or utilities have very sophisticated people

who do hedging basically on their ability to supply energy to the grid and the price that people are willing to pay for it. Now, all of this is a very highly regulated space. And so, the broader point here is that if you're interested in building a physical infrastructure service, it has probably existed for a long time. And the government has probably made a bunch of rules and determinations about the way it should work. And so, there are some physical infrastructure networks that are very are going to be very hard to decentralize because of the way they currently exist uh and the companies or government regulations that currently exist there. There are others where there are, you know, broader tailwinds and existing companies have outdated business models or the government has, you know, recently passed or changed

legislation to enable a more decentralized and distributed network topology. So, there's no value judgment comment there, but it's definitely something to think about before you decide to start a project or a company focused on this idea. And hopefully, we get many more decentralized physical infrastructure networks. We've been talking a little bit about decentralized physical infrastructure networks, which I think are one of the most compelling opportunities to build decentralized networks today and one of the best use cases for crypto, but very briefly, the potential to move from incumbent atrophying monopolies to web 3 products is a huge one and I think in many cases is a bigger change and shift than moving from pretty good web 2 products to even better [music] web 3 products.