Scaling Etherlink Without Compromise

Good morning everyone! Today I’ll be talking about Etherlink—and more specifically about how we are scaling Etherlink without making any compromise. And before I dive too deep into the details of this talk, I think it is always a good idea to start from the beginning. That is, what is Etherlink, really?

What is Etherlink?

First and foremost, Etherlink is the EVM-compatible layer of Tezos. We have implemented this layer as a Layer 2 blockchain powered by a Tezos smart rollup, but that’s an implementation detail. What matters is what we wanted to achieve with Etherlink: bringing EVM-compatibility to the Tezos ecosystem, at a time where it could make a difference.

Now, we still ended up implementing a Layer 2 blockchain no matter our main motivation—and one we want to be attractive. But as an EthCC speaker put it earlier this week, there is a new L2 chain being launched every day, which means we need to bring new things to the table that our competitors don’t, if we want to make Etherlink a success.

This is why we have built Etherlink according to three core values, summed up in a very simple motto: fast, fair, and (nearly) free.

• Etherlink is fast. Its sequencer is creating new blocks at a competitive pace (every 500ms), providing low-latency to the chain’s users. You only have to wait 250ms on average for your transaction to be included, which enables exciting opportunities (typically in DeFi).

• Etherlink is (nearly) free. As of today, it only costs around a tenth of a cent to make a tez transfer on Etherlink. This is necessary for Etherlink to be a welcoming platform for a wide range of use cases including some requiring the injection of hundreds of thousands of transactions every day. Paying 20 cents once is okay, but with high volumes, every cent counts.
• Etherlink is fair. In this area, we probably are pioneers actually. What we mean by fair is that Etherlink treats all its users equally, by being a permissionless platform where no arbitrary set of third-party entities holds dedicated rights. There is no committee, no chain owner capable of doing executive decision. On the contrary, we have entrusted the governance of Etherlink to Tezos bakers, making it the first non-custodial rollup. Anybody can become a baker, which means anybody can engage with Etherlink governance .

Etherlink in July 2025

We have been building Etherlink for a while now, and it turns out that July is a pretty busy month for us. In 2023, we announced Etherlink publicly at EthCC and TezDev and launched the Testnet . One year later, we had launched our Mainnet  and it had been non-custodial from day 1. It was of significant importance for us that as an Etherlink user, you didn’t need to trust us in gradually decentralizing a custodial solution. If you are familiar with the Ethereum rollup space, you already know that this approach is actually quite singular.

Now it’s July 2025, and we have once again exciting news to share. This is especially true if you are a Tezos baker, since July will be a voting month for you with two key votes for Etherlink’s futureOh, and we have also announced Seoul , Tezos’ 19th protocol upgrade proposal since then. !. One in particular—and I mention it because it is actually a good transition for the topic of this talk—is about a new proposal upgrade notably changing how the gas price is computed to align it better with Etherlink’s real capacityTurns out that we were a bit conservative in Dionysus. When testing more thoroughly, we noticed the chain was behaving better under load than what we previously thought. .

That being said, I would not say that we are happy with the capacity of Etherlink today. It’s enough for the current activity of the chain, as hinted by how stable the gas price is nowadays. But it’s not enough for the kind of ambitions we are having for Etherlink in particular and Tezos as a whole.

This is why we are committed to increase the capacity of Etherlink in the short-term (the next six months) at least tenfold.

Making Etherlink Blazing Fast

Software developers in the room are probably familiar with this expression 🦀.

But when facing such a claim, it’s only natural to ask: how do we achieve that? In my opinion, there are two aspects to consider.

Firstly, we need to execute transactions faster. For each transaction submitted to Etherlink, we want to spend as little time as possible executing it. For this endeavour, we have several plans that will allow us to increase the throughput of Etherlink.

To give a few examples:

• We are not particularly happy with the EVM implementation we are currently using (SputnikVM ). We are making great progress in transitioning to REVM .
• We want to introduce parallel execution where it matters. It is a hot topic in the industry, with already solid results that we can leverage.
• We have identified for a while now the shortcomings of the current runtime our smart rollups are built upon—and we have a team tackling this challenge to deliver a new RISC-V-powered runtime that will come with its own storage backend designed, implemented and optimised for optimistic rollups.

After combining all these efforts together and throwing a few more optimizations to the mix, we are confident Etherlink will be able to process transactions significantly faster.

Which brings me to the second side of making Etherlink fast: volume. Processing the transactions that we already have will make Etherlink more responsive, but we really want is to increase the overall capacity of Etherlink so that when you bring more activity to the chain, the system is not congested. To do that, we need to consider the amount of transactions that you will bring. We already have hundreds of thousands of transactions sent every day, but what happens when they become several millions? And several dozens of millions?

As of July 2025, the sequencer publishes Etherlink transactions inside Tezos Layer 1 blocks, but sadly block space is a scarce resource. To give you a few numbers, Tezos Layer 1 bandwidth (provided by its block space) is around 500 KBytes every 8 seconds. We can increase the block size, we can reduce the time between blocks, but the order of magnitude will not change. This becomes a real concern when we want to scale Etherlink without compromising with its non-custodial nature.

Scaling Without Compromise

We need to keep in mind what it means to increase both the speed of your system and the amount of data that you are processing with it. On the one hand, processing transactions faster is first and foremost an optimization challenge. It is a hard one, but ultimately it’s about using the hardware at your disposal in a more clever way. Processing more transactions on the other hand raises a security challenge, because it requires to publish more transactions. If Tezos Layer 1 blocks are not an option, where should the sequencer publish those?

One popular alternative to Layer 1 block space for a Layer 2 to publish its transactions is to rely on a Data Availability Committee (DAC). In this model, a small group of trusted entities is tasked with ensuring the data is available to participants.

But would Etherlink have stayed a non-custodial rollup if we had done that? The answer is a loud and clear “no.” Why? Because optimistic rollups rely on at least one honest participant that is willing to protect its state. That participant must have access to the full transaction data. If that data is only available via a closed committee, we introduce trust assumptions. And the moment you need to trust a small group for security, you're no longer truly decentralized.

This is why I am saying processing more transactions raises a security challenge. We knew that in 2022, when we started building the Data Availability Layer (DAL) of Tezos..We nicknamed the DAL the Rollup Booster , because the goal of the DAL is to increase the available bandwidth for Tezos’ smart rollups, and it is now live on Tezos Layer 1 Mainnet.

In a nutshell, the DAL is a proof of publication system. Anyone can send arbitrary data to the DAL’s peer-to-peer network and get back an attestation that this data has indeed been published. The same way Smart Rollups have been enshrined to the Tezos Layer 1 protocol, the DAL is enshrined to its consensus. That is, the body of the DAL’s attesters is the same one that is participating in the consensus of Tezos: the bakers.

Now is actually a good opportunity to celebrate that more than 85% of the baking power of Tezos is running a DAL node. DAL is already here today for Etherlink to use in order to increase its capacity. And since Dionysus (the current version of Etherlink), the sequencer can already decide to publish transactions to the DAL. It’s not doing it at the moment, because going through the DAL instead of Tezos Layer 1 blocks has a small impact on the latency of Etherlink’s bridges—but as soon as we need to, we’ll able to enjoy the additional bandwidth provided by the DALx10 compared to Tezos Layer 1 blocks today, x200 by the end of 2025. without any loss in decentralization.

It’s All Coming Together

If you have been following the work of core engineering teams from Nomadic Labs, Trili Tech and Functori working on Tezos and Etherlink, 2025 is a really exciting time. Everything we’ve been hard at work delivering our scaling roadmap for the past four years is now in production. Etherlink is live, bakers have enabled the DAL, and every piece of this technology stack remain true to the decentralization ethos of blockchains.

July 2024 to July 2025 was about bootstrapping this EVM-compatible layer of Tezos we call Etherlink. I want to emphasize how much work that entails. Launching the chain was actually the easy part. Having the chain being used, having an ecosystem in place, having the tooling, the partners, the projects landing: that was the hard part. It’s now done—as much as “being done” in the rapidly evolving landscape of blockchains means anything.

But we’re not done. We’re still preparing for the future to come. This talk was about discussing Etherlink’s capacity and how we intend to increase it tenfold. This will allow to create more room for more users to come, for more projects to be deployed. In the meantime, I do hope you find in Etherlink a welcoming platform for your projects!