An Ethereum layer 2 is a separate blockchain that executes transactions away from Ethereum's main chain and then settles the results back to it, so people can use ETH and Ethereum tokens for a fraction of mainnet's cost. Almost every Ethereum layer 2 running today is a rollup: it batches thousands of transactions, publishes their data on Ethereum, and either proves each batch correct with a zero-knowledge proof or leaves it open to challenge with a fraud proof.
Layer two networks now carry most of Ethereum's activity. They averaged roughly 2,300 transactions per second in the week to November 4, 2026, against about 21 on mainnet, and ETHBEAT, the analytics site known as L2BEAT until November 4, tracked about $35.9 billion secured on them on November 12, 2026.
The category looks very different from the one most guides still describe. Ethereum is scaling its own base layer again, Vitalik Buterin said in February 2026 that the original layer 2 plan no longer makes sense, at least 11 networks have shut down or announced closures this year, and Base and Arbitrum One hold almost two thirds of the value between them. Security maturity now separates these networks far more than fees do.
Key Takeaways
- An Ethereum layer 2 executes transactions off the main chain and settles them on Ethereum. Rollups dominate the category, and they come in two families: optimistic rollups, which rely on fraud proofs, and ZK rollups, which submit validity proofs.
- Rollups have to publish their transaction data so anyone can rebuild the chain without the operator's help. Since March 2024 that data goes into blobs, and Ethereum's blob target rose to 14 per block on January 7, 2026.
- Fees no longer separate the major networks. On November 5, 2026, the median token swap cost about a third of a cent on Base and about six cents on Ethereum mainnet itself.
- ETHBEAT's stages framework rates how much control operators keep over a network. Six layer 2s held Stage 1 on November 5, 2026, securing about 69% of all layer 2 value, while the four Stage 2 projects held under $1 million between them.
- Base is the largest Ethereum layer 2, with $13.87 billion secured on November 12, 2026. It is rated Stage 1, has no token, and moved off Optimism's OP Stack onto its own software in May 2026.
- Robinhood Chain, built on Arbitrum technology for tokenized stocks, launched on July 1, 2026, and held $2.04 billion on November 12, 2026. ETHBEAT does not yet rate it even Stage 0.
- Polygon PoS and Gnosis Chain are sidechains, and ETHBEAT gives Polygon PoS no stage at all.
- Ethereum's rollup-centric roadmap ended in February 2026. The base layer is scaling again, and at least 11 Ethereum layer 2s shut down or announced closures during 2026.
Why Ethereum Needed Layer 2
Ethereum's main chain can only process a limited number of transactions because every validator executes and stores every one of them. Mainnet averaged about 21 transactions per second in the week to November 4, 2026. Raising that ceiling makes a node more expensive to run, which pushes a network toward fewer and larger operators, the tension at the center of how blockchains are designed.
Scarce space means block space is auctioned. Users bid a gas price to get included, and during the busiest stretches of 2021 and 2022 a single token swap could cost $50 or more. That priced ordinary users out of decentralized finance and NFT markets, and Etherscan's daily transaction chart shows how close to capacity the chain ran in those years.
In October 2020, Vitalik Buterin published a rollup-centric roadmap that set Ethereum's direction for the next five years. Mainnet would stay conservative and act as a settlement and data layer, and rollups would handle most execution. The older plan to split Ethereum into many execution shards was shelved, and the sharding research was redirected into a cheaper way for rollups to publish data, which arrived as blobs in March 2024.
How an Ethereum Layer 2 Works
A rollup splits the work of a blockchain in two. It executes transactions itself and hands settlement and data storage to Ethereum, which is why rollups are described as the execution layer and Ethereum as the settlement layer. Five steps carry a transaction from a wallet to final settlement.
- Sequencing. The transaction goes to the rollup's sequencer, a server that puts incoming transactions in order and returns a confirmation almost instantly. Nearly every major rollup runs a single sequencer operated by the company or foundation behind it.
- Execution. The rollup runs the transaction in its own environment. Most layer 2s are compatible with the Ethereum Virtual Machine, so the same wallets, addresses and smart contracts work with little or no change.
- Batching and posting. The sequencer compresses hundreds or thousands of transactions into a batch and publishes it on Ethereum, along with a state root, a short cryptographic fingerprint of every balance after the batch.
- Proving. The rollup shows Ethereum that the new state root is correct. A ZK rollup attaches a validity proof that a contract on Ethereum checks straight away. An optimistic rollup lets the state root stand unless someone proves fraud during a challenge window.
- Settlement and bridging. Once Ethereum accepts the state root, the batch is final. A contract on Ethereum called the canonical bridge holds the assets users deposited and releases withdrawals only against settled state.
The sequencer's instant confirmation is a promise, and the guarantee only arrives once the data and the result are anchored on Ethereum. Most of the risks of using a layer 2 sit in the gap between those two moments.
Why Rollups Need Data
A rollup is only as trustworthy as the public record of its transactions. If the data is published, anyone can rebuild every balance on the chain from scratch, check the operator's work, and withdraw without asking permission. If the data is withheld, users are left holding a state root they cannot open.
What Data Availability Means
Data availability is the guarantee that the raw data behind a batch was published somewhere anyone can download it. Both rollup families depend on it, for different reasons:
- Optimistic rollups need the data so independent watchers can recompute each batch and build a fraud proof when the operator posts a false result. With no data, nobody can prove fraud.
- ZK rollups prove that a new state is valid, but a proof says nothing about who owns what. Users still need the data to know their balances and to prove ownership when they exit.
Publishing data on Ethereum is the reason rollups can claim Ethereum's security. Once a batch's data sits on mainnet, reversing it would mean reversing Ethereum itself.
From Calldata to Blobs
Until March 13, 2024, rollups posted their data as calldata, the same field ordinary transactions use for their inputs. It competed for block space with every other user and was stored forever, so data was usually the largest single cost of running a rollup.
The Dencun upgrade introduced EIP-4844, also called proto-danksharding, which created blobs. A blob is a packet of about 128 KB that travels alongside a block, has its own fee market, and is deleted by nodes after roughly 18 days, which leaves plenty of time for anyone to download and check it. Each blob is tied to its block by a KZG commitment, a cryptographic fingerprint whose setup Ethereum generated through a public ceremony that ran from January to August 2023 and collected 141,416 contributions. Rollup fees fell by around 90% in the weeks after Dencun.
| Calldata (before March 2024) | Blobs (since March 2024) | |
|---|---|---|
| Where the data lives | In Ethereum's permanent history | Alongside the block, pruned after about 18 days |
| Fee market | Shared with every other transaction | A separate blob fee market |
| Size | Limited by the block gas limit | About 128 KB per blob, with a set number of blobs per block |
| What smart contracts see | The full data | Only a commitment to the data |
PeerDAS, Fusaka, and Blob Parameter Only Forks
The Fusaka upgrade went live on December 3, 2025, and its headline change was PeerDAS (EIP-7594). PeerDAS lets nodes confirm that blob data exists by downloading and checking small random samples, a technique called data availability sampling, so capacity can grow without every node downloading every blob.
Fusaka also introduced blob parameter only forks, small upgrades that raise blob limits without a full network upgrade. The first lifted the target from 6 to 10 blobs per block on December 9, 2025, and the second raised it to 14, with a maximum of 21, on January 7, 2026. Fusaka set Ethereum's default block gas limit at 60 million as well. ETHBEAT's blob tracker shows how much of that space rollups use. The long-term destination is full danksharding, which would extend sampling much further.
The Blob Fee Floor
Cheap blobs created a problem of their own. When blob demand sat below the target, the blob fee could fall almost to zero, and rollups paid next to nothing for the security Ethereum provided. EIP-7918, shipped in Fusaka, ties the blob base fee to Ethereum's ordinary execution fee so blob space always carries a minimum price, roughly one sixteenth of the execution base fee per unit of data.
The floor is the main reason layer 2 fees stopped falling in 2026. Rollups also pay for execution, proof generation, and running their sequencers, so a layer 2 transaction always costs something.
Data Availability Layers Outside Ethereum
Some chains keep their data off Ethereum to cut costs further. They publish it to a data availability committee or to a dedicated data network such as Celestia, EigenDA, or Avail, an approach often called modular blockchain design because it separates execution, settlement, and data into different systems.
- A validium pairs off-chain data with validity proofs.
- An optimium pairs off-chain data with fraud proofs. Celo is one example.
Both add a trust assumption: if whoever holds the data refuses to release it, users cannot rebuild the state to prove what they own. On November 12, 2026, ETHBEAT tracked three validiums and optimiums securing $286.54 million between them, against $27.73 billion on rollups. Mantle shows the trade-off running the other way: it removed its EigenDA data path in April 2026, moved all its data to Ethereum, and was reclassified as a rollup.
Rollups, Validiums, Optimiums, and Sidechains
Plenty of networks market themselves as Ethereum layer 2s while making very different security trade-offs. The dividing line is where the transaction data goes and what Ethereum actually checks.
| Category | Where the data goes | What Ethereum verifies | Extra trust assumption | Examples |
|---|---|---|---|---|
| Optimistic rollup | Ethereum, as blobs | State roots, if someone challenges them in time | At least one honest watcher | Base, Arbitrum One, OP Mainnet |
| ZK rollup | Ethereum, as blobs | A validity proof for every batch | The proof system and its code | Starknet, ZKsync Era, Linea, Scroll |
| Validium | A committee or external data network | Validity proofs | The data holders release the data | App-specific chains using external data networks |
| Optimium | A committee or external data network | Fraud proofs | The data holders release the data | Celo |
| Sidechain | Its own chain | Nothing | Its own validator set | Polygon PoS, Gnosis Chain |
Sidechains cause the most confusion because they were marketed as layer 2s for years. A sidechain is a separate blockchain with its own validators, connected to Ethereum by a bridge. Ethereum's developer documentation on sidechains draws the line clearly: sidechains do not post state changes or transaction data back to mainnet, so they do not inherit Ethereum's security. If a majority of a sidechain's validators turns dishonest, it can freeze or redirect the funds held in its bridge.
Polygon PoS is the best-known case. ETHBEAT files it under "Others", the bucket for chains that do not meet its requirements for the rollup or validium categories, with no stage rating, and it held about $3.9 billion on November 5, 2026. Polygon did run a genuine ZK rollup, Polygon zkEVM, but its sequencer shut down on July 3, 2026, and a claims portal for remaining funds stays open until December 31, 2027. Gnosis Chain, formerly xDai, is another sidechain with its own validator set. On November 12, 2026, the "Others" category held 66 projects and $7.90 billion.
Bitcoin faces the same labeling problem with a different cast of protocols. Ethereum can verify proofs inside a smart contract and Bitcoin's scripting language cannot, which is why Bitcoin's layer 2 landscape grew around payment channels such as the Lightning Network, a split rooted in the different design philosophies of Bitcoin and Ethereum.
Optimistic Rollups Explained
An optimistic rollup assumes every batch is correct and gives anyone a fixed period to prove otherwise. The name comes from that default: the system is optimistic about the operator's honesty and relies on fraud proofs to catch a lie.
How Fraud Proofs Work
The operator posts a batch and a new state root to Ethereum and puts up a bond. A challenge window then opens, usually about seven days on Arbitrum One and OP Mainnet. Any independent watcher who recomputes the batch and gets a different answer can open a dispute.
Most disputes are settled through an interactive game. The two sides repeatedly split the disputed computation in half until they disagree about a single step, and Ethereum executes only that one step to decide who is right. The loser forfeits its bond. Ethereum only ever executes that single instruction, so the process stays cheap even for very large batches.
The security model needs only one honest participant. As long as a single watcher is online, has the data, and can submit a challenge, a false state root cannot survive the window.
Permissionless Proofs and Withdrawal Delays
For years, only whitelisted parties could submit fraud proofs on the largest optimistic rollups, which meant the security model rested on the operators themselves. That changed when OP Mainnet switched on permissionless fault proofs on June 10, 2024, and Arbitrum One activated its BoLD dispute protocol on February 12, 2025, letting anyone defend the chain.
The challenge window is also why withdrawals from an optimistic rollup are slow. Funds leaving through the canonical bridge wait out the whole window, about a week on Arbitrum One and OP Mainnet. Base shortened its own window to five days in June 2026 and added a faster path that combines a zero-knowledge proof with a proof from secure hardware, which finalizes withdrawals in about one day.
The Main Optimistic Rollup Stacks
Most optimistic rollups are built from one of two open-source software kits:
- OP Stack, maintained by the Optimism Collective. It powers OP Mainnet, Kraken's Ink, Uniswap Labs' Unichain, and a group of connected chains known as the Superchain. Base launched on the OP Stack in 2023 and moved to its own software in 2026 while staying compatible with the OP Stack specification.
- Arbitrum Nitro, developed by Offchain Labs. It runs Arbitrum One and Arbitrum Nova, and companies can launch their own chains with it through Arbitrum Orbit, which is how Robinhood Chain was built.
The line between optimistic and ZK designs is starting to blur. Base's fraud proofs now include zero-knowledge components, and several optimistic stacks are adding validity proofs to shorten withdrawals.
Zero-Knowledge Proofs Explained
A zero-knowledge proof lets one party, the prover, convince another, the verifier, that a statement is true without revealing anything beyond the fact that it is true. Shafi Goldwasser, Silvio Micali, and Charles Rackoff introduced the idea in a 1985 paper, and it stayed mostly inside academic cryptography until blockchains gave it a large-scale job.
A classic illustration uses two balls that look identical to a color-blind friend. You claim one is red and one is green. Your friend hides both behind their back, either swaps them or leaves them, shows them again, and asks whether a swap happened. If you answer correctly twenty times in a row, your friend can be confident the balls differ without ever learning which one is red.
Every zero-knowledge proof system has three properties:
- Completeness. An honest prover with a true statement always convinces the verifier.
- Soundness. A false statement cannot convince the verifier, except with a vanishingly small probability.
- Zero knowledge. The verifier learns nothing beyond whether the statement is true.
Why Rollups Use Zero-Knowledge Proofs
The property ZK rollups depend on most is succinctness. A short proof can show that a huge computation was carried out correctly, and checking that proof costs a small fraction of redoing the computation. Ethereum can therefore confirm thousands of rollup transactions with a single contract call.
Most ZK rollups hide nothing, since all their transaction data is published, which is why researchers often prefer the term validity proof. The zero-knowledge label stuck because the same proof systems power privacy tools.
SNARKs vs. STARKs
Two families of proof systems dominate. A zk-SNARK is a Succinct Non-interactive ARgument of Knowledge. A zk-STARK is a Scalable Transparent ARgument of Knowledge, introduced in 2018 by Eli Ben-Sasson and colleagues at StarkWare.
| zk-SNARKs | zk-STARKs | |
|---|---|---|
| Proof size | A few hundred bytes | Tens to a few hundred kilobytes |
| Cost to verify on Ethereum | Low | Higher |
| Trusted setup | Usually required | Not required |
| Underlying cryptography | Elliptic curve pairings | Hash functions |
| Quantum resistance | Not considered quantum-resistant | Considered quantum-resistant |
| Where they show up | Final on-chain verification for many zkEVM rollups | Starknet and many general-purpose proving systems |
A trusted setup generates secret random values that have to be destroyed afterward, because anyone who kept them could forge proofs. Projects run the setup as a multi-party ceremony, which stays secure as long as at least one participant deletes their share. Older SNARKs such as Groth16 need a new ceremony for every program, while newer ones such as PLONK reuse a single universal setup.
STARKs avoid the ceremony entirely by relying on hash functions, which also makes them a stronger candidate for a world with quantum computers. The cost is size: a STARK proof is far larger and more expensive to check on Ethereum. Many teams combine the two, generating a STARK quickly and then wrapping it inside a small SNARK so the final on-chain check stays cheap. Proving one proof inside another is called recursion.
Provers, Verifiers, and zkEVMs
On a ZK rollup, specialized off-chain machines, often clusters of graphics cards, act as the prover. A verifier contract on Ethereum checks each proof before it accepts the new state root. On most ZK rollups the prover is still run by the team, so a prover outage can delay withdrawals even though it cannot forge a balance.
A zkEVM is a ZK rollup that proves the execution of Ethereum Virtual Machine code, so existing Ethereum apps can move across with little change. Designs that match Ethereum exactly are the most expensive to prove, and designs that change some details prove faster at the cost of compatibility. A newer approach, the zkVM, proves programs written for a general-purpose processor and is used by teams that want to prove more than EVM code.
Proving has become fast and cheap enough that Ethereum researchers are now working on proving mainnet's own blocks with zero-knowledge proofs, the groundwork for native rollups.
How Zero-Knowledge Privacy Works
The same mathematics can hide details as well as compress them. A shielded transaction proves that the sender owns unspent funds, that the amounts balance, and that the owner authorized the payment, without revealing the sender, the recipient, or the amount. Zcash brought zk-SNARKs into production for exactly this when it launched in 2016, and the technique sits behind most privacy coins. On Ethereum, Aztec runs a privacy-focused rollup, and ETHBEAT counted it among the four Stage 2 projects on November 5, 2026.
ZK Rollups vs. Optimistic Rollups
ZK rollups and optimistic rollups both batch transactions and publish their data on Ethereum. They differ in how Ethereum becomes convinced that a batch is correct, and that one difference drives their withdrawal times, costs, and security assumptions.
| Optimistic rollups | ZK rollups | |
|---|---|---|
| How a batch is accepted | Assumed valid unless challenged in time | Accepted only with a valid proof |
| Security assumption | At least one honest watcher with access to the data | The underlying math and the correctness of the proof system's code |
| Native withdrawal to Ethereum | About a week, or one to five days on Base | Usually a few hours, depending on how often proofs are posted |
| Proof costs | Low, since proofs are only computed during a dispute | Ongoing, since every batch needs a proof |
| EVM compatibility | Very high | High and improving |
| Main software stacks | OP Stack, Arbitrum Nitro, Base's own stack | Starknet, ZK Stack, Linea, Scroll |
| Largest by value secured (November 12, 2026) | Base, Arbitrum One, OP Mainnet | Mantle, Lighter, Starknet |
Where the Split Stands in 2026
Optimistic rollups still hold most of the money. On November 12, 2026, Base, Arbitrum One, and OP Mainnet together secured $24.10 billion, about 87% of the $27.73 billion ETHBEAT tracked on rollups. ZK rollups are smaller but growing in number, with Mantle at $1.16 billion, Lighter at $1.03 billion, Starknet at $314.78 million, Linea at $297.98 million, and ZKsync Era at $211.67 million. Lighter is an app-specific ZK rollup built for trading perpetual futures on a decentralized exchange, a sign of how far proving costs have fallen.
Most new rollups launched in 2026 chose ZK designs, and the established optimistic chains are absorbing ZK technology. Base's Azul upgrade in May 2026 added a proof path that combines a zero-knowledge proof with a hardware-based proof, cutting its fastest withdrawal route to about one day.
Proof type describes how a rollup verifies its batches, and the stage rating describes who can still change the rules. Mantle and Lighter both use validity proofs and both sat at Stage 0 on ETHBEAT on November 5, 2026, while Starknet was the only ZK rollup among the six Stage 1 networks. A validity proof only protects users while nobody can replace the contract that checks it, and on a Stage 0 rollup the team can still upgrade that contract. For anyone choosing where to hold funds, the stage rating carries more information than the proof type.
The Major Ethereum Layer 2s in 2026
Two networks hold most of the money on Ethereum layer 2s. On November 12, 2026, Base and Arbitrum One together secured $22.68 billion, about 63% of the $35.92 billion ETHBEAT tracked across every category, and no other network held more than a few billion.
| Network | Built by | Type | Software | ETHBEAT stage (November 5, 2026) | Value secured |
|---|---|---|---|---|---|
| Base | Coinbase | Optimistic rollup | Base's own stack | Stage 1 | $13.87B (November 12) |
| Arbitrum One | Offchain Labs, governed by the Arbitrum DAO | Optimistic rollup | Arbitrum Nitro | Stage 1 | $8.81B (November 12) |
| Polygon PoS | Polygon Labs | Sidechain | Polygon PoS | Others, no stage | About $3.9B (November 5) |
| Robinhood Chain | Robinhood | Optimistic, Arbitrum-based | Arbitrum Orbit | Others, not yet Stage 0 | $2.04B (November 12) |
| OP Mainnet | Optimism | Optimistic rollup | OP Stack | Stage 1 | $1.42B (November 12) |
| Mantle | Mantle | ZK rollup | OP Stack with validity proofs | Stage 0 | $1.16B (November 12) |
| Lighter | Lighter | App-specific ZK rollup | Custom | Stage 0 | $1.03B (November 12) |
| Ink | Kraken | Optimistic rollup | OP Stack | Stage 1 | $372.91M (November 12) |
| Starknet | StarkWare | ZK rollup (STARK) | Starknet | Stage 1 | $314.78M (November 12) |
| Linea | Consensys | ZK rollup | Linea zkEVM | Stage 0 | $297.98M (November 12) |
| ZKsync Era | Matter Labs | ZK rollup | ZK Stack | Stage 0 | $211.67M (November 12) |
| Unichain | Uniswap Labs | Optimistic rollup | OP Stack | Stage 1 | $86.86M (November 12) |
| Scroll | Scroll | ZK rollup | Scroll zkEVM | Stage 0 | $34.13M (November 12) |
| Taiko Alethia | Taiko | Based rollup | Taiko | Stage 0 | $11.35M (November 12) |
Value secured is ETHBEAT's count of every token bridged to, minted on, or native to a chain, so it runs well above the decentralized finance deposits that sites such as DeFiLlama report as total value locked. The figures shift week to week, which makes them useful for proportion and less useful as a leaderboard.
A few patterns stand out. Corporate chains now lead the list, with Coinbase, Robinhood, and Kraken all running their own networks. Stage 1 networks hold most of the value, yet several large chains, including Mantle, Lighter, and Robinhood Chain, are rated below it. And the long tail is thin: Scroll, once one of the most closely watched ZK rollups, secured $34.13 million on November 12, 2026.
What Is Base? Coinbase's Ethereum Layer 2
Base is the Ethereum layer 2 built and operated by Coinbase, and it is the largest one by value secured, with $13.87 billion on November 12, 2026, close to 39% of everything ETHBEAT tracks. Coinbase opened a Base testnet in February 2023 and opened mainnet to all users on August 9, 2023. Base is an optimistic rollup that publishes its data to Ethereum as blobs, charges gas in ETH, and has no network token.
How Base Works
Coinbase runs Base's single sequencer. On November 5, 2026, the Denim upgrade moved Base to canonical blocks every 200 milliseconds, five per second, and retired Flashblocks, the stream of faster preliminary confirmations that apps had relied on. In the days after Denim, Base opened an investigation into periodic transaction inclusion delays and an unusually high number of empty blocks, and its upgrade notice was still marked as verifying on November 9.
Base secures its state with fraud proofs that now include zero-knowledge components. Withdrawals through the standard path wait five days, and a second path that pairs a zero-knowledge proof with a proof from secure hardware, known as a trusted execution environment, finalizes in about one day. Base averaged about 94 transactions per second in the week to November 4, 2026, more than four times Ethereum mainnet.
Why Base Left the OP Stack
Base spent its first two and a half years on Optimism's OP Stack as the largest member of the Superchain. On February 18, 2026, it announced a move to a unified stack of its own, pulling its sequencer and node software into a single codebase built on open-source components such as the Reth client. Base said the change would let it ship smaller upgrades more often, keep the codebase simple enough for one engineer to understand, and adopt Ethereum's own upgrades sooner.
| Date | Milestone | What changed |
|---|---|---|
| November 5, 2026 | Denim | 200-millisecond blocks; Flashblocks retired |
| September 30, 2026 | Cobalt | Further protocol and token standard updates |
| June 25, 2026 | Beryl | B20 token standard launched; standard withdrawal window cut from seven days to five |
| May 28, 2026 | Azul | First release on Base's own software; added the zero-knowledge and secure hardware proof path |
| February 18, 2026 | Stack announcement | Base says it will leave the OP Stack |
| August 9, 2023 | Mainnet | Base opens to all users on the OP Stack |
The split also ended Base's revenue-sharing arrangement with the Optimism Collective. Base still follows the OP Stack specification, so wallets and tools built for OP Stack chains keep working, and ETHBEAT still tags it as built on the OP Stack.
How Safe Is Base?
ETHBEAT rates Base Stage 1. Its proof system can be overridden only by a 75% supermajority of its Security Council, and if the sequencer ignores a user, that user can force a transaction through Ethereum with a maximum delay of 12 hours. ETHBEAT's Base project page also lists Base's exit window as none, because the Base Coordinator Multisig and the Security Council can approve contract upgrades with no waiting period. Users holding funds on Base get no advance notice of an upgrade they might want to leave before.
Does Base Have a Token?
Base has no network token as of November 2026, and all gas is paid in ETH. On September 15, 2025, the Base team said it was beginning to explore a token, which set off a wave of speculation and airdrop farming, but nothing has launched since. Any site offering a Base token or airdrop should be treated as a scam unless Base and Coinbase announce one through their own official channels.
What Runs on Base
Base has become one of the main homes for stablecoins and lending outside Ethereum mainnet. In August 2026, Coinbase began issuing tokenized US stocks on Base for eligible non-US users, part of the wider move to put real-world assets on blockchains, and they passed $5 billion in cumulative decentralized exchange volume by November 11, 2026.
| Measure | Value | Date |
|---|---|---|
| Value secured (ETHBEAT) | $13.87 billion | November 12, 2026 |
| Share of all layer 2 value secured | About 39% | November 12, 2026 |
| DeFi total value locked (DeFiLlama) | About $5.66 billion, above Solana's $4.99 billion | November 10, 2026 |
| Stablecoins on Base (DeFiLlama) | $5.01 billion | November 5, 2026 |
| Lending supplied and borrowed | $7.49 billion supplied, $2.65 billion borrowed | November 11, 2026 |
| Median token swap fee (growthepie) | $0.0034 | November 5, 2026 |
| Median ETH transfer fee (growthepie) | $0.0003 | November 5, 2026 |
Robinhood Chain and the Rise of Corporate Layer 2s
Robinhood Chain is an Ethereum layer 2 that Robinhood built on Arbitrum's Orbit technology to trade tokenized stocks and other real-world assets around the clock. It went from announcement to mainnet in about a year:
- June 2025. Robinhood announces the chain, alongside tokenized US stocks for European customers issued on Arbitrum One.
- February 10, 2026. A public testnet opens after six months of private testing.
- July 1, 2026. Public mainnet launches at Robinhood's event in London, removing the whitelist that had limited who could transact.
The chain runs Arbitrum Nitro with the BoLD dispute protocol, produces blocks every 100 milliseconds, and operates as a separate network from Arbitrum One. Robinhood's Stock Tokens trade on it 24 hours a day for eligible users in more than 120 countries, with US persons excluded, and holders can lend them out or post them as collateral in decentralized finance apps. Uniswap deployed a dedicated exchange on the chain from day one. Weekly decentralized exchange volume in tokenized stocks across all chains rose from $627 million to $3.4 billion in the months after the launch.
Robinhood Chain secured $2.04 billion on November 12, 2026, and its ETHBEAT project page shows how differently it is set up from Arbitrum One, a Stage 1 rollup running the same software:
- Only two whitelisted validators can challenge its state. ETHBEAT wants at least five outside parties able to submit challenges before it treats a chain as a rollup, so it files Robinhood Chain under "Others" and does not yet rate it even Stage 0.
- There is no exit window. A Robinhood multisig that needs seven of its eight signers can upgrade the chain's contracts with no delay.
- Force inclusion can be overridden. Robinhood screens transactions, and an authorized filterer can make any transaction fail, including one forced through Ethereum, so ETHBEAT lists the chain's protection against sequencer censorship as "no mechanism." The filter list passed 6,200 entries by November 6, 2026, and ETHBEAT notes that nearly all of them were forced attempts to move funds out of a single frozen wallet.
The same software can produce very different security, depending on who holds the keys and how the chain is configured.
Why Companies Launch Their Own Layer 2s
The biggest names on Ethereum's layer 2 list are now companies. Coinbase runs Base, Kraken runs Ink, Uniswap Labs runs Unichain, and Robinhood runs Robinhood Chain. Owning a chain gives a company several things at once:
- Sequencing revenue. The operator collects the fees users pay and keeps the difference over what it pays Ethereum.
- Product control. The company can tune block times, add features such as Base's B20 token standard with built-in allowlists and pausing, and ship upgrades on its own schedule.
- Distribution. Exchange customers can move funds onto the company's chain with a single withdrawal.
- Ethereum settlement. The chain still publishes its data to Ethereum and taps the liquidity and tooling of the wider Ethereum ecosystem.
The trade-off sits with the user. A corporate chain orders transactions through the company's sequencer and usually lets the company approve upgrades, so its neutrality depends on how far it has moved through ETHBEAT's stages.
How Safe Is Your Money? The Stages Framework
The most useful single measure of an Ethereum layer 2's safety is its stage, a rating of how much power a small group still has to override the chain's code. The framework comes from a November 2022 proposal by Vitalik Buterin, was formalized by L2BEAT in June 2023, and is now maintained by ETHBEAT, the name L2BEAT adopted on November 4, 2026 when it widened its coverage to privacy, interoperability, and staking. Its layer 2 monitoring continues unchanged. In September 2024, Buterin said he would only call a network a layer 2 once it reached Stage 1.
| Stage | Nickname | What it requires | Who can still override the chain |
|---|---|---|---|
| Stage 0 | Full training wheels | Data and state roots posted to Ethereum, source-available software, and a proof system in place | The operator, which can upgrade contracts freely |
| Stage 1 | Limited training wheels | Users can exit without the operator's help, and upgrades outside emergencies give at least seven days to leave | Only a Security Council, with at least 75% of its members agreeing |
| Stage 2 | No training wheels | Anyone can submit proofs, and users get at least 30 days' notice before upgrades | A Security Council, only for bugs that can be proven on-chain |
A Security Council is a group of named individuals and organizations who hold keys to an emergency multisig. On Arbitrum One, for example, the council has 12 members and needs nine to act in an emergency. Those powers get used. In April 2026, nine of the 12 voted to freeze 30,766 ETH, worth about $71 million, linked to the KelpDAO exploit, moving it into a wallet that only Arbitrum governance can release.
What the November 2026 Numbers Show
On November 5, 2026, six layer 2s held Stage 1: Base, Arbitrum One, OP Mainnet, Starknet, Ink, and Unichain. Together they secured about 69% of all layer 2 value. Four projects had reached Stage 2, namely Facet, Aztec, Ethscriptions, and a Cartesi test deployment, and they held under $1 million between them.
The largest networks stop at Stage 1, so most of the money on Ethereum layer 2s still relies on a Security Council acting in good faith. Ratings also move in both directions. Scroll became the first ZK rollup to reach Stage 1 in April 2025, then fell back to Stage 0 on June 1, 2026, when it replaced its independent 9-of-12 Security Council with a 3-of-4 multisig run by its own team.
What a Stage Does Not Measure
A stage describes decentralization. ETHBEAT is explicit that a team could deploy a Stage 2 system with an experimental proof system or unaudited contracts and still be unsafe. Stages also leave out the sequencer, the fees, and the bridge a user picks. A Stage 1 badge answers one question, whether a small group can take user funds, and the rest needs separate checks.
Before depositing meaningful sums on any layer 2, check five things on its ETHBEAT project page:
- Stage. Stage 0, or a place in the "Others" category, means a small group can still change the rules.
- Exit window. The time between an announced upgrade and its activation. Base's is listed as none.
- Sequencer failure. Whether users can force transactions through Ethereum, and with what delay.
- Proposer failure. Whether users can publish state roots themselves if the operator stops.
- Data availability. Whether the data sits on Ethereum or with an outside committee.
Sequencers, Preconfirmations, Withdrawals, and Escape Hatches
Almost every Ethereum layer 2 runs one sequencer, operated by the company or foundation that built the chain. That single operator decides which transactions get in and in what order, so it can delay or censor users, profit from transaction ordering, or simply go offline. Sequencer outages lasting from minutes to hours have hit several major rollups.
Preconfirmations
The instant confirmation a wallet shows after a layer 2 transaction is a preconfirmation, the sequencer's promise to include it. Networks compete on how fast that promise arrives. Base published preliminary Flashblocks for apps until November 5, 2026, when it switched to full blocks every 200 milliseconds, and Robinhood Chain produces blocks every 100 milliseconds. The promise only becomes final once the batch is settled on Ethereum.
Forced Inclusion and Self-Proposing
Well-built rollups give users two escape hatches for when the operator misbehaves:
- Forced inclusion. A user sends the transaction to a contract on Ethereum, and the rollup's rules oblige the sequencer to include it after a set delay. On Base, ETHBEAT lists that maximum delay as 12 hours.
- Self-proposing. If the operator stops posting state roots, users can post them themselves and withdraw against them.
Both only work if the chain's data is on Ethereum and nobody can override them. On Robinhood Chain, an authorized filterer can make even a force-included transaction fail, so ETHBEAT rates its protection against censorship as "no mechanism." On a properly built rollup, the same hatches still leave users a way out after the operator shuts down. When Blast's withdrawal interface closed in October 2026, its users could still recover assets by interacting with its bridge contracts on Ethereum directly, and Polygon zkEVM's claims portal stays open until December 31, 2027.
How Long Withdrawals Take
| Route | Typical time | What you rely on |
|---|---|---|
| Canonical bridge from an optimistic rollup | About a week on Arbitrum One and OP Mainnet; five days on Base, or about one day on its dual-proof path | Ethereum and the rollup's proof system |
| Canonical bridge from a ZK rollup | Usually a few hours | Ethereum and the proof system |
| Third-party bridge | Minutes | The bridge's contracts and liquidity providers |
| Exchange that supports the layer 2 network | Minutes | The exchange |
Third-party bridges get around the wait by paying out immediately from their own pool and collecting the slow withdrawal themselves. That convenience adds a new party to trust, and bridges have been among crypto's most heavily exploited infrastructure, with the Ronin and Wormhole hacks of 2022 alone taking more than $900 million.
Toward Shared and Based Sequencing
Two approaches aim to remove the single sequencer. Shared sequencers let several rollups use one decentralized ordering network. Based rollups go further and hand transaction ordering to Ethereum's own block proposers, which inherits Ethereum's resistance to censorship at the cost of confirmations tied to Ethereum's 12-second slots. Taiko took a based rollup to mainnet in May 2024, and its current network, Taiko Alethia, secured $11.35 million on November 12, 2026. Researchers are building preconfirmations for based rollups so they can match the speed users expect.
How to Move ETH to an Ethereum Layer 2
Getting funds onto a layer 2 takes a few minutes, and most mistakes happen in the setup. These steps work for Base, Arbitrum One, OP Mainnet, and most other EVM layer 2s.
- Check the network on ETHBEAT. Look up its stage, exit window, and where it stores its data before deciding how much to move.
- Pick a route. Withdrawing from an exchange straight to the layer 2 network is usually cheapest, because it skips the Ethereum gas a bridge deposit costs. The network's canonical bridge from Ethereum mainnet is the most direct option for funds already in a self-custody wallet, and deposits typically arrive within minutes. Third-party bridges connect layer 2s to each other and to other chains.
- Add the network to your wallet. EVM layer 2s use the same address format as Ethereum, so your existing address works, but each chain keeps a separate balance. Base uses chain ID 8453, Arbitrum One uses 42161, and OP Mainnet uses 10; take network details from the chain's official documentation.
- Send a small test amount first. Confirm it arrives on the right network before moving the rest.
- Keep some ETH on the layer 2 for gas. Base, Arbitrum One, and OP Mainnet charge gas in ETH, and ETH held on Ethereum mainnet cannot pay for transactions on a layer 2. A few networks use their own gas token, such as MNT on Mantle.
- Check token contract addresses. The same stablecoin can exist on one network in a native version and an older bridged version with a different contract, so confirm the address before swapping or sending.
- Plan the exit before you need it. Know how long a withdrawal takes on that network and which route you will use.
Account Abstraction and EIP-7702
Layer 2s are where smart accounts became practical, because sponsoring a user's gas costs fractions of a cent. A smart account is a wallet controlled by contract code, which allows features such as approving and swapping a token in one step, letting an app pay the gas, paying fees in stablecoins, or signing in with a passkey. ZKsync Era and Starknet build this in natively, so every account on them is a smart contract.
EIP-7702, which arrived with Ethereum's Pectra upgrade on May 7, 2025, lets an ordinary wallet borrow those features by delegating to smart contract code. Phishing sites have abused the same mechanism by tricking users into delegating to malicious contracts, so an unexpected request to "upgrade" a wallet deserves the same suspicion as a request to sign over tokens.
Liquidity Split Across Many Chains
Every layer 2 holds its own pools of tokens. USDC on Base and USDC on Arbitrum One sit in separate markets, so prices can drift between them and large trades move prices more than they would on one shared chain. Wallets and exchanges increasingly route between chains automatically, which hides the split from users without removing it. Buterin named slow progress on interoperability as one of the reasons he rethought the role of layer 2s in February 2026.
Why Layer 2 Networks Are Shutting Down
Cheap blob space solved the user's cost problem and created one for operators. A rollup still pays for its sequencer, bridge maintenance, security audits, engineers, support, and incentive programs, and none of those bills shrink when transaction fees do. Sophon, for example, said it was spending about $3.4 million a year to keep its chain running. With activity concentrating on a few large networks, 2026 brought the biggest wave of layer 2 closures so far, part of a wider shakeout in which more than 60 crypto firms and projects folded in the first seven months of the year.
| Network | Built with | Last day for users | What happened |
|---|---|---|---|
| Silicon Network | Polygon CDK | December 31, 2026 | Deposits stopped on September 2; the team says assets left after the deadline cannot be recovered |
| Abstract | ZK Stack | December 15, 2026 | Announced October 6 by Igloo Inc.; $15.52 million was still on the chain on November 12 |
| Sophon | ZK Stack | November 25, 2026 (chain off November 30) | Announced June 25 after concluding its running costs no longer made sense; apps moving to Base |
| Lisk Chain | OP Stack | October 31, 2026 | Bridging LSK back to Ethereum took about eight days, so the practical deadline came earlier |
| Blast | Optimistic rollup | October 26, 2026 (normal withdrawals) | Announced October 2, saying the chain cost more to run than it earned; later exits go through its bridge contracts on Ethereum |
| Mint Blockchain | OP Stack | October 20, 2026 | NFT-focused chain announced its closure on April 17; assets left after the deadline are lost |
| Zero Network | ZK Stack | July 31, 2026 | Zerion announced the wind-down on May 21 to focus on its wallet; about $1.4 million was secured on it |
| Polygon zkEVM | Polygon zkEVM | July 3, 2026 (sequencer off) | A claims portal for remaining funds stays open until December 31, 2027 |
| Loopring | Custom ZK rollup | June 28, 2026 (exchange and relayer off) | Value locked had fallen from about $760 million in November 2021 to about $8 million |
| Swellchain | OP Stack | June 15, 2026 | Swell closed its Superchain L2 to focus on its Faro product |
| Redstone | OP Stack in Plasma mode | May 15, 2026 | Lattice announced the closure on April 14; afterward only plain wallet balances could be withdrawn, and funds left inside apps such as Uniswap pools were lost |
| Kinto | Arbitrum Orbit | September 30, 2025 | Closed after a $1.55 million exploit in July 2025 and a failed fundraise; balances moved to a claim contract on Ethereum |
| Kroma | OP Stack with ZK fault proofs | June 30, 2025 | Gave users 10 days' notice; ETHBEAT says fund retrieval is not guaranteed after the deadline |
The closures cut across every major software kit, from the OP Stack and Arbitrum Orbit to ZKsync's ZK Stack and Polygon's own technology, so no single design choice kept a chain alive. What users could recover after the final date varied far more. Polygon zkEVM kept a claims portal open into 2027, Kinto deployed a claim contract on Ethereum, and Blast users could still reach their funds through its bridge contracts. Redstone, a Plasma-style chain that kept its data off Ethereum, could only return plain wallet balances, and funds left inside apps were lost.
Other teams pulled back before launching. ENS Labs cancelled its planned Namechain rollup on February 6, 2026, citing a roughly 99% drop in the gas cost of registering a name on Ethereum mainnet, and Synthetix and Superseed paused or dropped their layer 2 plans for the same reason.
Four lessons apply to anyone holding assets on a smaller network:
- Real users do not guarantee survival. Abstract onboarded more than 400,000 of them, and Loopring once held about $760 million.
- A Stage 1 rating did not keep Kinto alive. It held ETHBEAT's Stage 1 rating in August 2025 and closed on September 30, because the stages measure who controls a chain and say nothing about its finances.
- Shutdown deadlines are real deadlines. The easy exit path disappears on a fixed date, and recovering funds afterward takes technical work, where it is possible at all.
- A token and a famous brand say nothing about solvency. Blast and Abstract were both well funded and widely covered.
The End of the Rollup-Centric Roadmap
On February 3, 2026, Vitalik Buterin wrote that the original vision of layer 2s and their role in Ethereum "no longer makes sense." He gave two reasons. Layer 2s had decentralized far more slowly than expected, with Stage 2 and cross-chain interoperability proving much harder than planned, and Ethereum's base layer had started scaling on its own. He told networks that offer a cheaper copy of Ethereum secured by a multisig bridge that they were not scaling Ethereum in the sense the 2020 roadmap meant, and encouraged layer 2s to offer something mainnet cannot.
The Ethereum Foundation set out the new division of labor in March 2026. Mainnet is to be the permissionless, resilient hub for settlement, shared state, liquidity, and decentralized finance, and it is expected to scale by orders of magnitude. Layer 2s keep a role built on specialization and control. ETHBEAT's rebrand in November 2026 followed the same logic: it said scaling is now being pushed at the base layer, while layer 2s look for a purpose beyond it.
How Ethereum Is Scaling Its Own Base Layer
Ethereum's default block gas limit started 2025 at 30 million and reached 60 million with Fusaka in December 2025. Mainnet fees are already low by historical standards: on November 5, 2026, the median token swap on Ethereum cost about six cents.
The next upgrade, Glamsterdam, focuses on the base layer. Its main changes are enshrined proposer-builder separation (EIP-7732), which writes the split between block proposers and block builders into Ethereum's consensus rules, and block-level access lists (EIP-7928), which let clients process transactions in parallel. Glamsterdam went live on the Sepolia test network on October 6, 2026.
Developers postponed its next test network fork, on Hoodi, on October 22 after the auditing firm ChainSecurity found that two gas changes, EIP-8037 and EIP-8038, could break assumptions some existing contracts rely on. The withdrawn Hoodi schedule would also have raised that network's gas limit to 200 million. On November 6, the Ethereum Foundation said Hoodi would wait until the issues are resolved and brought in outside auditors, and as of November 12, 2026, neither Hoodi nor mainnet had a date.
Further out, Ethereum researchers want validators to check a zero-knowledge proof of each block, which would spare them from re-executing every transaction and allow much larger blocks.
Native Rollups
A native rollup would use Ethereum's own execution engine to verify its state, so the chain would no longer need a custom proof system or a Security Council to guard it. Paired with based sequencing, a native rollup would rely on Ethereum for ordering, data, and verification, which is as close as a layer 2 can get to inheriting all of Ethereum's security. Native rollups are still at the research stage.
What Layer 2s Are For Now
With mainnet getting cheaper, the layer 2s gaining ground offer something Ethereum's base layer does not:
- Specialized apps. Lighter runs a whole rollup around perpetual futures trading.
- Corporate products. Base and Robinhood Chain anchor tokenized stocks, stablecoin payments, and lending for their parent companies' customers.
- Speed. Blocks every 100 to 200 milliseconds suit trading and payments.
- Privacy. Aztec builds private transactions into the chain.
- Different virtual machines. Starknet runs its own Cairo language, outside the EVM entirely.
Do Layer 2s Help or Hurt Ethereum?
How much value layer 2s send back to Ethereum is one of the longest-running arguments among people who hold ETH. Base is the clearest example. In the seven days to November 4, 2026, Base users paid about $625,000 in fees, and Base paid Ethereum about $8,600 of that, roughly 1.4%, according to growthepie. Over the same period Base secured close to $14 billion in assets.
The case that layer 2s strengthen Ethereum runs like this:
- They settle tens of billions of dollars on Ethereum and depend on it for data and dispute resolution.
- The largest ones charge gas in ETH, so their users hold and spend ETH every day.
- They brought millions of users into the Ethereum ecosystem at a time when mainnet fees kept most people out.
The case that they weaken it:
- They capture the user relationship and most of the fee revenue while paying little for the security they rely on.
- They split liquidity and users across dozens of separate chains.
- Their commitment to Ethereum is a business decision, and corporate operators can change course quickly, as Base showed when it left Optimism's software stack and revenue-sharing deal in 2026.
Ethereum's developers have answered on several fronts: the EIP-7918 blob fee floor stops blob space from being given away, the push to scale mainnet lets more activity stay on the base layer, and based and native rollups would route ordering and verification back to Ethereum. The argument remains open in late 2026, and Ethereum no longer depends on layer 2s alone for its growth.
Choosing an Ethereum Layer 2
The right network depends on what you are doing and how much trust you are willing to extend to an operator.
| If you are | Reasonable choice | Why |
|---|---|---|
| Trading DeFi with larger sums | Base or Arbitrum One | The deepest liquidity among Ethereum layer 2s, and both rated Stage 1 |
| Making frequent small payments or swaps | Base, OP Mainnet, or another Stage 1 network | Fees of a fraction of a cent per transaction |
| Holding for the long term | Ethereum mainnet | No sequencer, bridge, or upgrade keys between you and your funds, and a swap cost about six cents on November 5, 2026 |
| Trading tokenized stocks | Base or Robinhood Chain, where eligible | Both host stock tokens, with access depending on your country; Robinhood Chain sits below ETHBEAT's rollup requirements |
| Wanting fast withdrawals to mainnet | A ZK rollup, or Base's dual-proof path | Withdrawals finalize once a proof is verified, in hours or about a day |
| Trading perpetual futures on chain | An app-specific chain such as Lighter, with only what you need | It is purpose-built for trading but rated Stage 0, so its team can still upgrade it |
| Trying a smaller or newer network | Check ETHBEAT first | Stage 0 or "Others" means operators control the contracts holding your funds |
| Using an app that exists on one chain only | That chain, with only what you need | Bridging assets you are not using adds risk with no benefit |
Two habits cover most of the risk. Check a network's stage and exit window before depositing, and treat assets on a layer 2 as working capital you are actively using. The networks that closed in 2026 gave users weeks of notice, which only helped people who were still paying attention.
Conclusion
Ethereum layer 2s execute transactions away from mainnet and settle them on Ethereum, and in November 2026 they carry most of Ethereum's day-to-day activity. Rollups publish their data in blobs so anyone can rebuild them, and they prove their batches with either fraud proofs or zero-knowledge proofs. Fees on the major networks have settled at fractions of a cent, held up by the blob fee floor, and Ethereum mainnet has become cheap enough for many everyday transactions too.
The category is consolidating around a few large networks, several of them run by large crypto and trading companies. Base and Arbitrum One hold about 63% of all layer 2 value, Robinhood Chain arrived in July 2026, and a string of smaller rollups closed during the year as Ethereum stepped back from treating layer 2s as its main path to scale.
For anyone moving money, a network's stage and exit window on ETHBEAT say more about safety than any fee comparison. A rollup at Stage 1 or above that posts its data to Ethereum leaves users a way out even if the operator disappears. On a Stage 0 chain, a validium, or a sidechain, that way out depends on the people running it.






