The Ethereum Name Service just shipped the biggest upgrade in its history. Notably, it did so on Ethereum mainnet rather than the L2 it spent years planning.

On Tuesday, Oct. 27th, ENSv2 went live on L1, roughly nine months after ENS scrapped Namechain, which was the rollup that was originally supposed to host the revamp.

— (@)

ENS isn't alone, either. Over the past year, a growing list of apps have left L2s, canceled plans for their own, or shut down chains they'd already built, then moved their core contracts back to Ethereum L1.

Return to mainnet, in other words.

The backdrop for this trend is Ethereum's gas limit (i.e. how much computation fits in each block) doubling from 30M to 60M over the course of 2025, such that L1 fees fell right along with it.

Meanwhile, running your own chain has proven expensive. Two L2s, Blast and Abstract, announced shutdowns within four days of each other this month alone.

Let's break down the bigger picture here, starting with the latest domino...

Why ENS stayed on Ethereum

As many of you know, ENS turns wallet addresses into readable names like william.eth.

ENSv2 has arrived as a total rebuild of the protocol. In short, every ENS name can now run its own registry for its subnames, while owners get fine-grained permission controls and fees are paid in stablecoins like USDC.

Most interesting to me is where the v2 lives, though. Back in May 2024, I wrote ENSv2 101 about ENS Labs' plan to bring the protocol to an L2, which later took shape as its own rollup, Namechain. At the time, I said the upgrade looked like "the beginning of ENS's mainstream era."

ENSv2 has actualized and hailed its new era, just not on an L2. On Feb. 6th, ENS founder Nick Johnson explained the U-turn: "Ethereum L1 is scaling, and it's scaling faster than almost anyone predicted two years ago."

Indeed, ENS had seen its registration gas costs drop 99% over the prior year as Ethereum scaled. Plus, Nick argued that an L2 would add trust assumptions that L1 doesn't.

For his part, Vitalik Buterin signed off the same day: "It's a good decision!"

The dominos

ENS is the newest name on this growing "returners" list, but it's far from the first.

For its part, Synthetix was an early mover. Despite being one of the first protocols to launch on an Ethereum rollup back in 2021, it published a post in June 2025 titled "L2s are Dead, Long Live Ethereum," citing fragmented liquidity, among other issues.

Accordingly, Synthetix began winding down its L2 deployments that summer, and its new perps exchange opened on L1 in December. As founder Kain Warwick later put it: "I spent two years of my life memeing the L2 narrative into existence. It didn't work out well for Synthetix in the end hence back to mainnet."

Then, in February, came the post. Ethereum creator Vitalik Buterin argued that L2s' progress toward full decentralization had been "far slower and more difficult than originally expected" while L1 itself was scaling. His conclusion was that "the original vision of L2s and their role in Ethereum no longer makes sense."

I covered the post at the time, and ENS pulled the plug on Namechain three days later.

— (@)

Enjoying this article?

Subscribe to Bankless or sign in

Subscribe for free

This summer, Superseed became another case study. Its whole design used fees from its own L2 to pay down users' loans. Instead, it deprecated the chain on Aug. 15th and is now in the process of "bringing its entire ecosystem home to Ethereum Mainnet," swapping chain fees for app-layer yield strategies.

Then, in September, Trueo, the prediction market protocol behind the DefiLlama markets I wrote about last week, announced it's "coming home to Ethereum" from Base. Vitalik cheered the plan.

This October fielded a flurry of related moves, too:

• ShapeShift moved its rFOX staking program from Arbitrum to L1 on Oct. 1st since most FOX already lives on mainnet.

• Blast announced on Oct. 2nd that it's winding down since its costs exceeded its chain revenue.

• Nillion took its Blacklight network live on L1 on Oct. 5th, and its roadmap is retiring its L2.

• Abstract said on Oct. 6th that it's shutting down on Dec. 15th, calling a consumer-only chain "unsustainable as a standalone model."

Some of Abstract's NFT projects, like Gigaverse's ROMs and BEARISH, have announced moves to L1, too.

State comes home?

There's a subtler pattern running through these migrations, namely that apps are increasingly separating where their state lives from where their computation happens.

Take Synthetix's new perps. Orders get matched offchain, but user funds sit in custody on Ethereum L1, where trades also settle.

Or take Derive V3, which I broke down earlier this month. Its order matching stays offchain for speed, its risk math gets proven with zero-knowledge proofs, and user funds live in Ethereum L1 contracts.

ENSv2 operates in this vein, too. The NFT names themselves live on L1, while cheap and gasless subnames can still resolve from L2s or offchain via CCIP-Read, a standard that lets a name point to data stored elsewhere.

In other words, the valuable stuff (custody, ownership, and settlement) is moving home to L1, while the heavy lifting is going wherever it's cheapest and fastest. That's different than the classic rollup model, which holds your state and your compute, and therefore needs an exit when you want your "stuff" back.

However, note that there are also technical avenues that will increasingly flatten the distinction between L1 and not L1.

For example, the Ethereum Economic Zone (EEZ), which I explained back in March, is an effort to give Ethereum and its rollups synchronous composability, i.e. letting contracts on different chains interact within a single transaction.

Rather than pulling apps off L2s, the EEZ aims to make rollups feel like part of L1. The plans are proceeding, too. GnosisDAO voted in August to turn Gnosis Chain from a standalone L1 into an EEZ rollup, as I covered when it was proposed.

There are newer, more unproven ideas here as well. This week, a researcher going by namnc floated an "ephemeral rollup," or "L1.5," on the Eth Research forum, which takes the state-compute split to its logical end.

In this proposed design, L1 holds commitments to users' state and checks every proof, users' wallets hold their own private state, and computation runs as short-lived jobs that keep nothing afterward. Since an L1.5 holds no state, there's nothing to exit from.

Of course, this migration momentum rests on Ethereum staying roomy. If apps arrive faster than L1 capacity grows, fees could climb again and shift the math.

Glamsterdam, Ethereum's big upcoming scaling upgrade, saw its next testnet activation postponed last week, and per the Ethereum Foundation, no mainnet date has been decided.

Still, Glamsterdam already runs at a 200M gas limit on testnet, over 3x today's 60M. Given what the last doubling did for fees, that bodes well for L1's headroom if mainnet follows.

Whatever happens, it's clear that more and more apps are bringing their most valuable state home to L1, and ENS is just the latest to make this trip.

Now, keep an eye on the next dominos to drop. And in the meantime, if you hold a .eth name you can upgrade it to ENSv2 at app.ens.domains whenever you want. It cost me about $1.40 in gas for three names!