211 - Is DA a Good Business Model? - Neel Somani & Jon Charbonneau

I just think it's hard to become money we only know two digital assets have that have ever done it Bitcoin eth and I suspect that part of that is because of the fact that eth and Bitcoin are constrained at the base layer as a result it forces this asymmetry of demand or in congruity of Demand with respect to supply and it forces it to be deflationary so I don't know if it's possible for that to ever happen again it seems like more was like a freak accident of History where if you tried to deploy a constrained base layer today people would be like why would I use that over Celestia salana all the other guys but that basically means that only

Ethan Bitcoin will ever be money welcome to bank list where we explore the frontier of Internet money and internet Finance this is Ryan Sean Adams I'm here with David Hoffman and we're here to help you become more bankless Celestia just launched and is already worth over $20 billion at the time of recording that's the 11th most valuable crypto asset we also have aan da that's coming out soon and today we're taking an in-depth look an exploration of data availability it's

also known as da also known as data publishing if all of that sounds like a lot don't worry we spend the first 10 minutes of the episode actually explaining it and why it is so important to understanding block space and your broader crypto Journey so the topics today number one data availability what is it why does it matter number two what does cheap da unlock for web 3 number three is it even a good business model how valuable is it to be a DA provider number four we get right into the heart Tia is worth $20 billion why is this

actually a race to the bottom is this a race for a commodity what does the future have in store for the entire Market there has been this growing hype and excitement inside of the Celestia ecosystem now being um packaged up in the Tia is modular money a meme uh this is partly an exploration into that subject matter uh maybe a prep for a conversation I'm going to have with Nick white from Celestia exploring that meme unpacking that Meme specifically this this I think is more of a a third-party

Bird's eyee view exploration of the the nature of da um what is this business model what are the margins like what are the cash flows like why is Da valuable what is the long-term future of Da why will da consumers like rollups and other chains and other ecosystems why what choices will they make and why will they make them uh I think this the content in this episode is going to illustrate some of the decision trees that many rollup providers da purchasers are going to have to make in the coming years uh as

this whole da um da universe unfolds we also of course get into conversations around Proto D sharding and full DK sharding and what is this difference between ethereum native da for ethereum Native rollups versus Alt da there is a big choice between choosing electing to consume ethereum da versus non- ethereum DA what is the nature of that choice the two guests that we have on the episode today today John sharbono from DBA people will know John he writes insanely depth research articles all about different subjects inside of the

ethereum and broader crypto world and also Neil somani from Eclipse we've had on Neil once before he is the guy building the salana of virtual machine as a layer to on ethereum consuming Celestia for da so both of these guys know a thing or two about the world of Da and the choices around that ecosystem as a disclaimer before we get into the episode we are advisers for Igan layer the producers of Igan da which has brought up a handful of times throughout this episode and also John's investment firm DBA is an investor in Eclipse who the other guest Neil is the CEO of so

let's go ahead and get right into the episode is Da a good business model but first I want to talk about some of these fantastic sponsors that make the show possible especially Kraken our preferred exchange for crypto in 2024 if you do not have an account with Kraken consider clicking the links and the show notes getting started with Kraken today INX Nation I'm so excited to introduce you to John charbono a researcher investor at DB a John is known for writing some of the longest and deepest research pieces out there exploring the frontiers of all of the leading ecosystems that we have in crypto ethereum Celestia salana

wherever there are interesting research topics John is out there sifting for gold John welcome back to bankless how's it going good to be on again Neil somani is the CEO of eclipse a project bringing the svm to an ethereum layer 2 Eclipse uses ethereum for settlement salana VM for execution Celestia for data availability and risk zero for proof verification and Neil like John has been on the Forefront of the modular design space and if you couldn't tell from how eclipse is constructed he has a thing or

two to teach us about the role that data availability has to play in the world of network design Neil also welcome back to bankless thanks for having me guys so there is a broader conversation I think going on right now about more or less what the hell is Da what the hell is data availability and not just from a technical perspective which we will also just answer that 101 question but also qualitatively what does it mean to have good da what does it mean to have bad da uh is Da a commodity how what is the future of Da is it the bandwidth of web

3 what is its future role in trajectory as the future of da um uh expanse and but I in order to really just nail down the one ones I want to just Define data availability it's a weird term it's one of these weird terms that the crypto industry comes up with normes will hear it and they're like this is whatever podcast I'm listening to is not for me uh maybe John you can start by kind of laying down some Foundation what is Da how is it a part of just blockchains uh and why is it important yeah one of the

many unfortunately named things in blockchain which we've kind of realized after a couple years still isn't super clear um so I mean kind of it's kind of helpful to start with like the basic idea of like what is a blockchain in the first place it's like some way for us to agree on some form of shared State and like we can update that thing and then we agree on the new shared state of the world um so like fundamentally what you want to have there is everyone knows what the state of the world is and then like knows what it's correct um and kind of fundamentally to do that you both need to be able to like verify that this like new state that you've been told um

is correct and also just to like have all the data there in the first place just to know that and understand what it is um so like the the most concrete way so while so what you should also realize from that is like we always talk about it in the rollup context for the most part um but fundamentally every blockchain ensures data availability like that is just a fundamental guarantee of them um rollups just kind of highlight it and that we kind of like strip it out and think about it separately which is why we've just always kind of like taken it for granted that every blockchain ensures data availability by default of full nodes just download all the data and if all

the data is there then you know that the data is available if the data is missing then you shouldn't sign off on that block um that's like generally how any traditional blockchain will work whether it's salana Bitcoin ethereum Etc um for rollups the like the area where it gets highlighted and we see it differently um is they are posting their data to another chain to a chain like ethereum um and you will see data availability you need it for for example like the clearest example is for optimistic rollups um you know you can't make a fraud proof um to show that like some State transition was invalid if you don't have the data available um so

while it's something that we kind of like highlight and see differently in the rollup world it is just fundamentally a guarantee that like any blockchain needs to have is hey if we're all going to agree on what like the state of the world is and everything about it you need to know what the data is um so it's just like a very fundamental guarantee for any John if we're this is a core fundamental guarantee of blockchains what do they do they they tell you whether the state of the world is true or not right okay so and then we have in in kind of these modules sort of the what we've talked about is the consensus layer the data availability layer and then the execution layer one framing of this that

I've had is like the consensus layer tells us like what's just happened or like what's true I think of it like that and the da layer is like what's happened in the recent past it's kind of like there's some history to of sort of archive associated with it um tell me if that framing works for you or or how you'd modify that in terms of what we're actually concretely talking about here so it's not so much the history um and this is part of the reason why I would say data availability wasn't kind of unfortunate name in hindsight because it does sound like that um but we're really

referring to just a very like short-term guarantee of I am able to download this data and I know that it's available it's not a guarantee that like you can go get the data from 5 years ago that ethereum had or anything like that um that like kind of historical storage requirement is a little different John danrad has called it data publishing do you like do you like yes it so it I definitely do think it is clearer um a lot of the other data availability people are also in favor like Mustafa and Celestia like I think they're generally in favor of it I think it is just so hard to like change the

name of it at this point um to something that is like probably a little bit better but has just been MD into this gigantic thing over time um but that but that is a clearer way to think of it um and that is like actually how the idea was kind of described in the first place like back in Bitcoin world is basically is this and and when we when we say data publishing is is maybe clear because like I I guess that's what in fact we are doing when I think most normal people They hear the word publishing they they think about it in a non-blockchain context so they think about like publishing to the web or I'm

publishing my blog or I'm publishing this uh podcast to my RSS feed right but when we're talking about data publishing or AKA data availability we'll use those terms interchangeably through the rest of this episode we're talking about publishing that data to the blockchain right so and again what do blockchains do sort of a you know truth computer for us it's a truth engine it's a it's a trusted compute layer so that's what we mean in this context of of data publishing yes yeah you could also publish to the internet and that would be some form of data publication it's

just very bad da uh as opposed to a blockchain which is a gold standard okay bad da bad da can we like unpack that a little bit why is that bad da and can you like frame it up as like okay did in crypto did we invent good da so Neil what what what is why did what did you mean by bad da and then maybe you can also go into what is good da so to me good da is something that's one verifiable so anyone could determine whether da was performed properly and whether the data was actually published and second it has Better liveness

Properties meaning that on the internet theoretically whatever web server published that too they could refuse to serve you that block and that's okay for archival data just because you don't actually need that to determine the fork Choice Rule and to determine the current state of the chain as John was pointing out once you've already finalized a particular tip or a particular history but um but for good da it's probably better to to optimize on good da for things like um like when you're in the active process of final of finalization so Neil when you say good versus bad

right we're we're starting to um add some subjectivity to this I think and like maybe the question is like good for what or bad for what because I think the way I publish you know my uh newsletter is probably fine for the purposes of a newsletter it's just content that I'm trying to I don't I don't need sort of the trust guarantees of a blockchain but maybe you might argue is bad for use cases like um let's say transferring millions of dollars large amounts of of value so with with

what kind of context what types of apps are you are you judging uh da through this lens of good versus bad is it basically internet of value types of apps you need stronger da guarantees whereas like just Communications regular old internet copy paste jpeg type apps it's fine to just use the protocol layer of the internet yeah tell me about that I think it's more about what's the threat if data were to be withheld and sometimes that's a more meaningful way to think about da in the context of a data withholding a tag meaning if

someone didn't give you this data what's the worst damage they could cause and for example for a blockchain once something is already finalized if they withheld historical data that might mean it's harder to bootstrap a new Full node and you can't replay history from the beginning of time but at the end of the day you know that there was some safety guarantee given by the fact that the whole Quorum voted off on something and you had consensus at some point whereas if something has not even been finalized in someone without data then you could actually have Forks you could have safety violations uh then it's much more meaningful to have good da in in those

circumstances so what can happen if the data is withheld I guess the stakes are higher with high value use cases like actual you know peer-to-peer uh money transfers that sort of thing whereas so it depends on the depends on the blockchain so for an L1 if you withheld data then the blockchain would literally fork and then you'd have two sets of participants some which have seen the blocks and some which have not and then it depends on who you're connected to uh in terms of what state you're going to be receiving but if you're an L2 then that means that ether the ethereum

bridge could literally be compromised in the case of an optimistic roll up where the executor could publish an invalid State Route and then theoretically if the block were withheld then no one would be able to wipe that invalid State commitment and would eventually be finalized so those are the it it somewhat depends on uh what the blockchain construction is to determine what's the worst case outcome of a data withholding attack okay John I'm just you know mining for data here do you do you have anything else to add to like to just explaining data publishing and data availability to to normies what else

would you add to kind of tie us off and then we'll get into the deeper subjects here I think we covered it well there I I would just finish it with like that simple mental model and description of publication probably is the right one for most people to think of it in the same way that you know you can publish something to any website and based off of that we can have some reasonable assumption that like everyone saw that data and if it was there for long enough everyone should have been able to download it um but it's not a guarantee that like if you publish something that you know a year later the sign it you know has taken it down or something else um but your data was published um and

it's just a matter of like that strength of guarantee and that everyone saw it and then we could all verify that together I actually do really like bringing in the web 2 database conversation here just as a frame of reference where we have like our censorable substack or like our bankless YouTube that was actually taken down one time and deplatformed our data was withheld from us and this is this has caused issues in like the web TW space like we call it deplatforming um and then when you apply a new technology called a blockchain that introduces

property rights and settlement guarantees withholding data is a same it's a similar problem but now we're talking about that data was actually your money like not just your YouTube videos uh and so maybe maybe that can help uh listeners come in and reframe this conversation for data availability John talking about it from an economics standpoint data availability is a commodity it all block all blockchains have them Bitcoin so like just like you said it's a core part of every blockchain if you are a blockchain you

have data availability your blockchain has qualitative properties therefore the data availability of your blockchain also has qualitative properties Bitcoin is data availability with a proof of work mechanism ethereum is proof of stake with a TR how but like data availability it's it's a commodity like it's it's a resource how do we categorize this thing I I think that's a good way to categorize it is that that it is effective just kind of a base commodity that um is kind of an input to most of these systems the way that most people will use that term then is to

kind of try to make the point of oh it's a commodity like these different things are all the same whether I use Celestia or ail or Ian da or ethereum whatever that it's just a commodity you know they're all kind of the same thing um that is the part that I would in practice at least kind of disagree with and push back on um is that like I would agree that in the limit of like if we have all these perfect systems and everyone is like yeah there's a bunch of perfect da layers that have all the guarantees you want then yeah it should be like relatively fungible between them um practical reality is there is a very large difference I would say in the

quality of block space that is being provided by one da layer versus another one today um because otherwise you can take this to the extreme of we're actually back to the web to of like posting your data online you know is data availability in some sense like why why don't we just put all of this on ads like why do we need a blockchain in the first place um and so there very much is like this kind of spectrum of like there is different I would say qualities of block space um and there's different features that come along with them of whether it's you know certain Network effects of being in the same ecosystem together um others lend themselves uh

much better to data availability s data availability sampling which means people can more easily verify it um so are these kind of like different features that definitely do differentiate in practice like one da layer versus another in the short to medium term I think that's true I think in the long term pretty much all these folks like avale Celestia I da the features like data availity sampling and having really good decentralized Rel layers will essentially become table Stakes because if they don't have those Baseline features even things like soft confirmations to better support based Roll-Ups these constructions will become

so popular that if you don't offer the features that are conducive to those constructions you'll just be non-competitive um so that's kind of how I'm thinking about it in the long term and then ethereum da obviously is in a category of its own assuming you're using ethereum for it's for the fork Choice rule just given that now your trust minimized with respect to ethereum Okay so there's so many things uh that like we we just brought up there but so David's basic question is is Da a commodity and your answer to that both of you I I think you guys would say yes it's a commodity and bankless listeners

have heard us talk about um block space as being sort of the core commodity that this entire revolution has uh has created like the thing that we made was block space and is it accurate to say uh John that da is just like one component of block space in fact inside ethereum on kind of like main net before it sort of modularized and it sliced out da as a separate part it basically included da as part of that block space and just da is just one component of the larger suet commodity which is block space is that

correct yeah yeah I would agree with that yeah really data avilability and then consensus over that is kind of like the fundamental Baseline of what any blockchain um is like looking to that you need to have and to provide for them okay and then John you were making the argument that not all da is the same so like some Commodities uh like da Commodities have different properties than other Commodities but but Neil you were just saying well well actually that's true right now but over time they'll sort of fuse fuse together they'll become kind of like a a baseline

this is the set of functionality and features that all DA has and it will I I guess maybe another word is commodify right they'll all become kind of uh similar to one another but then you made an exception there Neil and you said but uh ethereum da might be in a category of its own can you say more about that like why is ethereum have any special status and I would love uh John's take on this too so it's mostly because of the fact that rollups typically Define their Fork Choice rule on ethereum and this is actually something that we were chatting about with the Celestia team just recently

they said why don't you just Define your fork Choice rule directly on Celestia in some way and the reason for using ethereum is that now assuming you also offer forced inclusion on ethereum then theoretically your rollup could operate even if Celestia goes down and that's a pretty big Advantage so by using ethereum for da as well it means that even in the happy path you're um you're only relying on ethereum and there's no risk that celesia going down could impact the safety or liveness of your rollup so that's the reason why it would be different but for example if you were to have a rollup that's subtle to Solana

then I'd argue you should also use salana for Di because that would be the trust minimized construction for that kind of rollup is what you're saying just there Neil is there's like um a resonance between using the same layer for da and settl M like if you're using both and both are the same there's just a little bit of um sum greater than the parts uh that emerges when when both a DA and the settlement are the same layer it's essentially avoiding the same issues that come with bridging because in order to use Celestia da with

ethereum settlement at some point you have to relay some data from Celestia which is a signed data route so everyone's posting their transactions to Celestia Celestia Quorum signs off on some succinct representation of that and that has to be relay to ethereum and doing that in a decentralized and trustless way runs into the same exact issues that bridging runs into understood understood Neil you've also used this term Fork Choice rule could you just unpack that a little bit for listeners who might not be familiar with that term just explain like I'm five what that means so let's say I have some

set of transactions and then I have an alternate set of transactions and they lead to different state routes how do I decide which one is a fork yeah and then choosing which Fork you should go on is or which Fork is canonical is the fork Choice role okay understood getting into just more of this like the qualitative nature of this da conversation both of you have said like yeah da is a commodity and then also there's this set of features that all da providers will need to be able to provide in order to be competitive at all which means uh to

me that there's like this natural convergence towards some conclusion about all competitive da layers which also kind of sounds like there it's a it's a race it's a race to this conclusion and it's also like data is cheap in the grand scheme of things like I like Hardware is like you know one terabyte is like not that much money these days it's like $25 is that a fair comparison and is that if like if data uh data on a blockchain is is the cost of this approaching that level of

commodity like nature where just da costs are approaching you know zero is that is that a fair conclusion to arrive at John I I'll start with you here definitely is a race towards this kind of like theoretical equilibrium of exactly what you described of it should basically approach like what is the cost operationally for the most part of like providing this service in the long run um and I do think it will continue to Trend closer to that over time um the question is uh like there are many things that in crypto are in practice

this is the equilibrium and you end up here and then like you know everything is in practice though you don't actually get there at least for a very very long time um there's a lot of markets I mean like the same thing as like proof a proof of stake mechanism in theory the equilibrium is like 1% as all the stake and that's it but generally in practice there's going to be a very long road to like actually get to those systems and it might not actually get there for this I would I would expect that in the long run the cost of like using one of these da layers does get pretty close to like some reasonable multiple on like what is the operational cost of like literally

providing this like x amount of bandwidth that is replic that is like replicated across end nodes and however many times um I I do think that you should approach that over time but in practice today I mean like we're so far away from that of like in reality there is one specialized da layer that is live today like it's Celestia and everyone else is like on the time span of like years when does it actually get to something similar than that and even Celestia is in its like very very initial state where like their their current state now is nowhere near oh yeah we can just theoretically provide all the bandwidth that everyone needs

and you know it's going to cost 3x of like whatever the bandwidth cost is like it needs to scale like scale like a thousand times more um to start even getting close to those kinds of numbers okay John so your long-term prediction is that the cost of Da is going to decrease right uh similar to other collapsing I guess uh you know Tech CA like something something equivalent to maybe like mors law and that sort of thing but that will take some time to play out uh is I I guess your prediction and let let's connect some dots for for listeners right now about the implications of that right so there's uh

kind of I guess two stakeholders we could talk about one is the implications for Value AC Rule and sort of investors who are looking at uh da layers themselves like if you're looking at something like Celestia or you're trying to invest in I da the question is what will this ACR value over time where where in the value chain is it so there's a set of questions for investors but there's also a set of uh you know questions and answers for users and I think one implication would just connect some dots very quickly is the cost of transacting on blockchains is going to

decrease a whole lot in the near future and then whole lot more by orders of magnitude in the long-term future maybe let's just talk about the the users because there's always going to be more users uh than than there is um you know investors here and so is this effectively crypto's Broadband moment when we see da cost collapsed almost nothing that means blockchain as a commodity the usage of blockchain sort of collapse uh towards zero um what are the downstream effects on on that of of

that for users so it should mean at least the base input cost of spinning up one of these systems should approach zero um or approach some reasonable multiple and like the literal operational cost of providing it um it doesn't necessarily mean that um the specific application the cost of using something has to go to zero um there is still stickiness and value in like a specific piece of shared State on on top of what may be a very cheap um piece of Da so like if you have an application that is super in demand um and that is

on top of a DA layer that is like super super cheap great your input costs are effectively zero um but there still can be obviously a cost to interact with that application um if it is valuable enough that they decide to like charge extra fees and now there is some premium am execution yeah I get that so it's definitely understood I'm just wondering if you kind of like um hold with Chris Dixon's take on this and he calls himself sort of an infrastructure determinist right so when he sees the cost of infrastructure approaching zero then he's like oh there's going to be incredible app ecosystem and like I don't even need to know what apps are

built on top of this I'm just going to invest in this thing that is you know uh the infrastructure whose cost is approaching zero right and so uh his take on this would be like look at um bandwidth scaling on the internet right and for some listeners getting involved in the internet I it's the era of dialup modems right and uh the 56k modem and then we had our bandwidth era on on the internet and that didn't mean that the apps we used on top of the internet approached zero but the bandwidth of those apps uh was effectively multiplied by orders of magnitude do do you like

that analogy does that kind of work here are we about to enter a high bandwidth ERA with da costs going so low I think to some degree it's about like rewriting the software so it can take advantage of bandwidth increases that's what Don sharding is essentially doing or the upgrade known as Don sharding is doing for ethereum and I think the phenomenon you're pointing out is essentially neon's law every year there's some percent increase in bandwidth year-over-year but I guess my point is that I don't know that before then there's anyone who's really properly

positioned to take advantage of those bandwidth increases though there's Celestia Solano who claim that they're well positioned to take advantage of it I think there's this informal like upper bound on how much bandwidth a blockchain can reasonably support because you need light nodes to actively verify the bandwidth um so like to to put it in another way if you look at Bitcoin CPUs like kept increasing Over Law over time like we we have Mor's law that's a true thing yet Bitcoin doesn't use more CPUs in the execution of programs and that's intentional it's because there's a security constraint

that like those CPUs are used for Cil resistance for Bitcoin and they're not trying to use as many cores as possible so similarly for blockchains like some blockchains might have an informal design goal that they don't want to expand bandwidth too much because that can potentially compromise the security of the bandwidth or the blocks that are being propagated I think if uh we do this episode well the original motivation for this is really to understand the economics of selling data availability as your business uh and like if if we have a bunch bunch of Da

providers like where where do they end up on coin market cap is really like the goal I'm trying to get here and and so and we already kind of talked about the race towards um what John I think alluded to is like limited margins and and we I think we'll return there later but first in order to really answer one half of this question I want to talk about the buy side uh which is the question of like what does Da unlock with cheap da what are we able to do in crypto that we weren't able to do before like what new fertile grounds have does cheap da provide us and Neil your

project. Eclipse like I said in the intro chose elected to use like Celestia da for da because celestius got the cheapest da so what did that choice unlock for you like what were you able to not do before and now you are able to do and I think this conversation with eclipse is actually will be the first of like a blossoming conversation for the future of rollups and even cheaper da but let's start with E clipse like what could you do because you elected to choose Celestia da I think that's the right framing because if we were able to we would have just used ethereum da if

it were able to support the use cases that we wanted to and I think that in the long run we probably will migrate to ethereum Da but those use cases are things like Central limit order books for a market maker to effectively use that and to subm make cancel orders transactions have to be on the order of 1/1 100th of a penny and even one one even one/ tenth of a penny is possibly doable but beyond that there's no way it's economical so that's one use case things like less economical use cases such as full on chain games deepin networks even certain types of social

social networks like I'd argue the experience of using friendch could have been potentially smoother if there was just more abundant block space and therefore transactions were able to flow through more easily so those are categories of apps that I think are better enabled uh when you have really cheap da I that kind of just um invokes memories of 2017 to me when we could Daydream up anything uh about some sort of crypto project and it was all very high in the Sky it was all turned out to be like Ico Mania at the time but in hindsight there were still like lots of nuggets of very broad very grandiose

ideas of what we could do with blockchains and maybe da was actually a fundamental constraint to every single one of them there's a growing um motivation for further experimentation in layer 2os uh we really like the last like two three four years of Layer Two experimentation has been relatively constrained maybe one of those constraints has been what we don't have the da to really experiment with like John Maybe I'll ask you to kind of carry this conversation forward like cheap da like what can what can the rollup landscape really do with cheaper da it basically enables all of these just

types of applications of it it it is similar to any layer one being like cheap and scalable it is fundamentally very similar to that um it's mostly that l2s in particular have been the area of crypto that has had a very challenging time to do that um like technically anything you could do on a roll up or an L2 like you could just do on a centralized server um so it's a matter of being able to to do it in the context of that type of chain um and particularly like l2s for ethereum um where in practice that is still where the majority of users and assets and