Prompted by @dankrad, it's worth looking at the type of blockchain applications that make sense, and what sort of demand we're looking at. I largely agree with Vitalik's blueprint (minor disagreements noted, and also my post on the topic), so this is the template we'll follow. As always, my blog posts are pure rambling, and I have thought and searched about this topic for less than an hour since that tweet.
Money: considering population in countries with unstable economies and/or poor payment solutions, I estimate this at ~1.5 billion. It's very important to note that this number is down-only historically - we have pretty good solutions like mPesa, PayTM, AliPay around the world. In the US, the average person makes 1.2 transactions per day. While this is almost certainly going to be lower for the type of economies that actually need this usecase, we can roughly estimate this at ~2 billion transactions per day. However, again, this will decline over time.
DeFi: I've argued before in multiple ways that DeFi is a niche category in terms of number of entities and transactions, even if the $ values can be very substantial. All considered, I don't expect more than a few million tx/day, though these are much more complex txs.
Identity: this one can vary wildly. I can see a scenario where SIWE lives alongside Google, but it could easily remain a niche for blockchain nerds. Either way, the interactions with blockchains will be quite limited, basically just registration and modifications. Most of identity will be about verification, which are read-only actions. Even if there's widespread adoption, daily transactions will be a few hundred thousand txs per day in the long term, and probably much lower (even if the off-chain verification stuff are hundreds of millions per day!)
DAOs: this is very much a niche. Only a certain type of organization will benefit from being a DAO. It's hard to put a number, but I can't imagine this being more than 1% of all companies. Even if all proposals and all voting was on chain, it's still few million tx/day tops.
Hybrid: I have argued in the past that hybrid makes sense for most cases. As I’ve argued before, a lot of purported usecases for “web3” or blockchains don’t actually require both peer-to-peer operation and strict global consensus. Even if public blockchains were free, they would still not make sense, and would be more efficient as hybrids. They are much more efficient with hybrid solutions where only the specific aspects are decentralized. Now, here I can certainly see a spectrum with billions or even trillions of actions/day. But most of it will be centralized, some of it will be validiums/optimistic chains, so it doesn't quite count towards “decentralized blockspace”.
Some miscellaneous things: Ponzis & gambling is of course a huge product-market fit for blockchains. I do expect ponzis to be relatively price elastic at the bottom end (even if the premium ones may even be Veblen goods!). It’s hard to put a number, but I’d say there are hundreds of millions of people into gambling! At the same time, I expect a lot of these ponzis to happen on validium/optimistic chain-like solutions. These usecases simply do not benefit much from maximal decentralization or security. As mentioned, the premium ponzis that’ll compete for decentralized blockspace work quite differently - they are built around the idea of scarcity, and that necessarily limits the number of transactions per day, particularly with the most popular ponzis thus far - NFTs, where you have very very few transactions. Exchanges are a bit of an anamoly because they are both high-value and can aggregate to a ton of transactions. However, it’s very likely most transactions are fine on centralized or validium-type solutions. Even still, I will definitely count dozens of million transactions per day on L1s and fully secured L2s. It’s important to note that exchanges are very compressible in rollups. Indeed, we’ve seen dYdX only consume 5.35 bytes per tx. So, post-EIP-4844 a fully decentralized Ethereum rollup can already service up to 1.3 billion transactions per day without requiring full danksharding! Finally, there’s the elephant in the room - MEV bots. Theoretically, this type of transaction is heavily price elastic, and more and more will pop up to fill in what I’ve previously called “micro MEV” opportunities. However, we have seen every single execution layer set a minimum fee to cut off rampant MEV & other worthless spam, and I expect that will be the case for all generalized L1s and L2s. (I have some old posts about how app-specific L2s can circumvent this)
Addendum: Just wanted to address, “but what about all the new applications that we can’t imagine!”. Yeah, maybe, but the “we’re so early” phase is long over. We have had application-specific chains for nearly a decade now, and Ethereum has been live for 7+ years. In the last couple of years at least pretty much every software developer has been exposed to blockchains. The longer we go, the less likely it becomes we’ll get some massively revolutionary application. I’m not saying it’s not possible, but rather, there’s no reason to consider it until we at least see some hints of. Nevertheless, I’ve certainly left a lot of leeway in my estimated number below to account for unpredictable outcomes.
As I’ve argued in many places before, the TAM for core decentralized blockchain applications in terms of blockspace requirements isn’t that high. Ethereum L1 + rollups + EIP-4844 are already sufficient for most valuable usecases for the foreseeable future. OK, I just realized I kind of didn’t address Dankrad’s prompt. But looking at most of these usecases - they are actually quite inelastic by nature. They tend to being a) hard to substitute and b) essential and well-specified financial applications. (E.g. you aren’t going to register 5 ENS, and you aren’t going to split up your grocery payments into 5 parts just because it’s free) Also, there’s pretty well-defined ceiling. If we assume blockchain apps fill all the niches that Vitalik describes, we’re looking at something in the ballpark of ~10 billion transactions per day for “decentralized blockspace”, following which there’ll be a rapid cliff in demand. Of course, there could be billions or even trillions of transactions per day that opt for Hybrid solutions, but I’m not counting these. Semi-decentralized execution layers definitely have their work cut out though!