Scaling Up and Not Sideways — the Future of Blockchain Scaling

Scaling blockchains while keeping transaction costs low is one of the toughest challenges for blockchain computers. There are two main attempts to solve this:

1. An increased amount of ‘fast’ blockchains — Solana, Binance Chain etc. All attempts to create blockchains that can process more transactions/second (‘TPS’) by utilizing better hardware and/or giving up on some aspect of decentralization and being a more concentrated network of computers.
2. Rollups (e.g. Starkware, Arbitrum) that are compressing transactions in different techniques and record them on existing blockchains (currently only Ethereum).

Each of these two presents a different approach to scaling.

I will explain here the logic of why creating faster blockchains doesn’t make any sense. I will argue that *most* value will eventually be captured on one or two blockchains (similar to operating systems oligopoly with Android, iOS, Windows, OS X) and that’s it.

This is opposed to the ‘multi-chain’ theory that many investors are suggesting.

(p.s I’m not saying there will not be room for niche players, just that most of the value will accrue to 2 winners at the maximum).

Why?

Let’s start with how blockchains work very briefly. Blockchain is a set of computers that share a state by consensus. Ethereum also introduced general compute to the picture, so these computers also can share code, execute it, and then record the new state through consensus (‘smart contracts’). A shared global computer.

This computer is very good at being robust and keeps on going without anyone being able to shut it down (this is the censorship resistance property). It is not so good at computing things. That is because for it to become robust at censorship resistance it utilizes compute redundancy (all computers must do all of the computations).

That is why compute on blockchains is inefficient and will never, ever, scale in its current form. It doesn’t matter how good of hardware you’ll get, or what tradeoffs you will make. It just doesn’t make any sense that *everyone* has to compute *everything* (On Ethereum that means that every piece of code is executed in parallel ~10k times right now).

The redundancy that creates the base property of blockchains also limits their ability to scale as a global computer.

Any tradeoff that you will make (except for completely giving up censorship resistance and basically becoming a server) will not overcome the fact that you accept some redundancy in compute. Redundancy doesn’t scale well.

This is where rollups come in. With rollups, you keep very little compute on the base layer and perform the logic of smart contracts on a higher level. Rollups are executed on a centralized computer that is not part of the blockchain (and therefore can scale compute) but still get the censorship resistance of blockchains by utilizing different techniques (won’t get into this in this current piece but if you want to learn more learn about ZK rollups / Optimistic rollups).

Once we externalize the compute to a higher layer, we can 1) make any kind of computation we would like, even very intensive AI computations 2) make transactions really cheap (probably zero with Starkware L3 and fractal scaling) 3) scale to billions of users.

So summarizing up my argument:

1. We will never scale to one billion users with redundancy in compute
2. We will never give up on redundancy in compute with only using parallel computing machines (blockchains)
3. We need to outsource compute and keep the decentralization
4. Rollups solve that
5. Now there are zero reasons to use ‘fast’ blockchains (from the technical perspective)

And yet, why should there be more than one winner?

This is the same reason we have a few operating systems. Human and Financial capital flowing to a few base layers right now will create culture and innovation in a few places in parallel. My guess is that not only Ethereum will adopt rollups, but other chains will also adopt rollups in 2022/3 and we will have a few candidates as base layers with rollups on top. There will be some loyal users using some other chains than Ethereum, and then the grassroots will grow into trees in some places (it might also be the case that the loyalty will be to a specific app on top and not the base layer itself. Such as a game on Solana).

In the same way that most users don’t really know the technical difference between iOS and Android (except for application layer differences), I’m guessing it will be the same for different blockchains. The differences will be around the culture and nature of applications building on top (like gaming on Solana). But the bottom line is that there is no technical reason to have a few base layers to answer the demand boom we are seeing today (If I were some kind of a central designer that architects the system from scratch I would probably just create one or two blockchains).

So I think the future will be up and not sideways. Don’t agree with me? I’ll be happy to hear your thoughts and counter arguments.