Luca Donno - Ethereum’s Roadmap to 10M TPS - TOKEN2049 Singapore 2025

Hey everyone. I'm Luca. I'm a researcher and that's why my slides might look a little bit different uh from the usual slides. Uh I'm a technical person and I'm here to talk about Ethereum's road map to 10 million TPS. And you know in the DeFi world we usually got used to ask ourselves when we see like those super high yield values uh we used to ask ourselves where does the come from right and in the world of high TPS chains um you should ask a similar question to yourself which is where does the scaling come from because in the same way that for very high yields uh there's probably you know something fishy going on you're probably the yield right um for these very high TPS chains uh it is likely that they are compromising something so it's very important to know uh when you use these chains uh where does the scaling come from and one of Ethereum's core value that we want to preserve uh compared to uh the other projects other uh blockchains is that we want to maintain the property that users should be able to independ independently verify their property rights with no trusted intermediaries.

So what we mean by the by this is uh we want users to be able to run blockchain nodes. We want users to be able to run their transactions on their own hardware and this hardware has to be cheap has to be um we don't want this to be very expensive. We don't want to be uh having you know terabytes of RAM machines to be able to verify the tokens that you're holding and whether the state that you're touching is correct. So it's really important for us to uh for users to be able not to depend on any intermediary to know that your tokens are there. Uh, and of course, if you want to have this property, the bottleneck is you need to have a blockchain that goes at the speed of this lowest node that you want to support, right?

Ideally, we would want to have um users running their full node on their smartphone or maybe on on a smartwatch, right? And these computers are very slow. Hence, um you might want to rate limit the speed of your blockchains to be that slow. And this is what Ethereum has been doing. If you look at past performance, Ethereum has been doing like 12 TPS.

Right now, it's doing around 20 TPS, which is pretty bad, right? Uh but this is because we want to preserve this value. We want users to be able to run nodes. Uh the thing is that there is some new tech uh that allows to um allow nodes to be able to verify things that are much more powerful than them. We call this an execution verification as symmetry and there is new cryptography that allows this.

Uh there are fraud proofs and there are ziki proofs. Zik proof seems to be more promising the f proofs. But the idea here is a node instead of having to reexecute all the transactions to know that the state is correct. They can just verify a ziki proof which is this cryptographic proof right. Uh but the problem is the tech is nowhere close to be mature enough to be used for ethereum.

This is new tech. This is experimental. Uh it hasn't been on production for a very long time. So right now we cannot use it uh for Ethereum to scale Ethereum. So the short-term solution that we found is okay let projects that are willing to take such risk experiment and iterate uh and that's why uh we got to rollups or L2s.

So what L2s do is they use zika proofs to scale their computation. They have much higher throughputs than Ethereum, but they produce the Zika proof, which is this uh very small cryptographic uh proof that they send to Ethereum. And these proofs are much cheaper to verify than the execution that they cover. Um and right now Ethereum is doing uh if counting the rollups is doing around 300 TPS and which is again m much better than what Ethereum is doing uh today at 20 but it's still not you know 10 million as I promised. The thing is uh the ZK technology is getting more mature with time.

It's getting faster and the Ethereum foundation and Ethereum in general has committed to bring the Zika technology on Ethereum itself. Uh there is new initiative um for for realtime proving. There is a website showing the stats uh on the progress towards being able to zero L1 directly. uh we need some more performance improvements to make it viable to make it fast enough. Uh but this is happening like very soon maybe in one year maybe in two years uh we'll be able to use this technology on Ethereum itself and make use of this asymmetry for scaling.

Uh the thing is why is it the case that with this technology rollups are only doing 300 TPS and the limit here is uh in data availability. So the way that these uh rollups work, I'm not going into the details, but they need to post their transactions on Ethereum uh to be able to inherit the security properties like ordering, consensus, and censorship resistance, right? But the thing is that the data throughput that Ethereum supports only allows for 300 TPS max. The good news is uh on the 3rd of December of this year there will be a new upgrade for Ethereum which introduces blob sharding. This is again a new cryptographic primitive uh that allows to scale how much data throughput the blockchain can support while still keeping uh verification low.

So potentially uh we'll be able to support uh very very high throughputs and nodes will be able to verify that all this data is available uh just by downloading a very small part and but with the same guarantees as if they were downloading everything and this is again new tech I don't have time to go into the details um but this is called blob sharding and one week later this upgrade Ethereum already bumps the data throughput limit to be able to support 500 TPS and one week later again there will be another bump to be able to support 800 TPS. So from 350 that we have today to around 800. Um but the target for this upgrade so if everything goes well if we see that the networks reacts well. Uh the idea here is to bump the DA throughput limit to be able to support almost 4,000 TPS. So we're kind of getting there, right?

Um the thing is if you look at other blockchains that are also working on DA and one of course wonderful example is Celestia they're doing amazing job in scaling data availability uh in the next upgrade which will happen soon. Um they will already be able to scale uh data availability to support 120,000 TPS. Uh and this is you know just to give you an idea of what is possible to support with uh data sharding. uh and their long-term goal is to reach 1 gigabyte per second of uh data throughput that roll-ups can use right and if you do the math this is a simple linear function uh of how many transactions per second uh this can support you get to around five to six million TPS uh which it's a very good number right the thing is with all these new cryptograph uh cryptographic uh innovations we kind of solved the verification bottleneck. So as I said before right now today Ethereum is slow because we are bottlenecked by the speed of re-executing all the transactions by the speed of verification.

With this new tech that unlocks the execution verification of symmetry. We solved the verification bottleneck and now the bottleneck the bottleneck is on the execution side. how fast we can produce blocks. And if you look, you know, at modern blockchains, it is believed that the serial execution performance limit for a single blockchain is around 10,000 TPS, right? Maybe there is one order of uncertainty for this value.

Uh maybe it's 100 uh maybe it's 10,000 TPS or maybe it's 100 TPS, but still this is not uh 10 million as I promised initially, right? So if one single blockchain can only do maximum let's say 10,000 TPS, how we can get to 10 million TPS and the the simple idea here is you can have multiple blockchains that that run uh in parallel uh that run in parallel, right? So instead of having just one blockchain that does 10,000 TPS, you can have multiple ones in the same way that Ethereum today has L2s and rollups like arbitrum, base, linear, zikync. And all of those will be building blocks in parallel. They will not be limited by each other's speed and all of them will be publishing uh data transaction data to Ethereum.

Um so uh there have been many criticisms uh around L2s and rollups because um there is this stages column on this website. Uh this is a website from L2. I work at L2. I spent a lot of times you know uh a lot of time uh looking at these projects and what stages are uh is basically an indication of how decentralized those systems are. If you're stage zero, you're fully centralized, which is bad.

If you're stage two, you're fully decentralized, which is good. If you're stage one, you're kind of in the middle. And if you look at all the top L2s, all the top rollups, none of them is stage two, which is kind of bad, right? Because we want to be super decentralized, but none of those systems are. And we need to understand why.

And one thing that I'm also working on these days is native rollups. And what what are native roll-ups? So the idea here is all these projects need to be centralized because Zika proofs as I said before is experimental and they want to have multisigs. They want to have security councils to protect from bugs. This is one reason and the other reason is they want to be able to upgrade to maintain feature parity with Ethereum.

If Ethereum introduces a new feature, they also want to upgrade to have this feature. So they need to have some again multi-IG security council that can upgrade which is again a risk of centralization risk. But here the idea is if we implement native rollups on Ethereum they will be able to reuse the same ZK verification infrastructure that Ethereum already uses. Uh they don't need to deploy their own stuff and they don't need to maintain their own stuff and if they reuse what Ethereum already uses if Ethereum upgrades they would also upgrade automatically. So the idea here is no dedicated proof system.

You can use the same zika stack already used by L1. If there is no dedicated proof system, there's no need for security consoles. You automatically upgrade with one. But still you you can support custom sequencing. For example, you can have centralized sequencing with super fast preconfirmations like 200 milliseconds preconfirmations that uh projects really like.

You can have custom gas tokens. For example, you can support uh a chain. You can be you're able to deploy a chain that has a yield varying stable coin as a gas token. You can capture all the fees uh yourself and you can easily opt in or opt out of this feature of being native. So the TLDDR of native roll-ups is you can see native roll-ups as customizable Ethereum shards or Ethereum copies uh with full potential Ethereum security.

And here I say potential because of course you can customize these chains and you can customize them in ways that can lower their security. But potentially they can be as secure as Ethereum. So the TLDDR how to get to 10 million TPS in three simple steps is first you get data availability or like data throughputs which we also call blobs. Uh limit to 10 million TPS and again this is feasible. Uh data sharding scales amazingly well.

Then you push the single chain limit, the single um execution environment limit to 10,000 TPS. And then you get uh 1,000 of these chains doing 10,000 TPS each for a total of 10 million TPS total. And the coolest part of all of this is that all these chains, all these 10 million TPS are fully verifiable on a very light machine. you can potentially verify all of these throughputs, all of these state changes on your smartwatch or on your um on your phone, right? So, this is uh kind of the alloc that the new technology, this new cryptographic technology allows and what differentiates Ethereum from from the other blockchains.

That's the end of my talk. Thank you.