Layer 1 vs layer 2: what is the difference in crypto?

On 24 October 2026 the median transaction on Ethereum cost about 2.2 US cents. On Base, a network that runs on top of Ethereum, it cost about 0.17 cents, and Base handled roughly 7.5 million transactions that day against 1.7 million on Ethereum itself, according to growthepie data. That gap is the short answer to layer 1 vs layer 2: Ethereum is the base layer, Base is a layer 2 that batches many users’ transactions and writes the result back to the base layer. The longer answer is about who actually protects your money on each side, how long it takes to leave, and why the bridges between the two have lost billions.

This lesson follows our comparison of USDT and USDC, the two largest dollar stablecoins, and stablecoins are a good place to start: the same USDC or USDT can sit on Ethereum, on Arbitrum, on Base or on Polygon, and the network you pick changes the fee, the waiting time and the risk. Picking the wrong one when you withdraw from an exchange can cost you the whole amount.

Layer 1 vs layer 2 in crypto: who settles your transaction

A layer 1 is a blockchain that runs on its own. It has its own validators or miners, its own consensus rules and its own record of who owns what. Bitcoin, Ethereum and Solana are layer 1s. When you send ether on Ethereum mainnet, Ethereum’s validators order the transaction, put it in a block and make it final, with nobody else involved. If you want the mechanics of blocks and consensus first, our lesson on how a blockchain works covers them, and what Ethereum is explains the chain this lesson uses as its main example.

The catch is capacity. Ethereum mainnet can process roughly 15 transactions per second, according to ethereum.org. When more people want in, they bid against each other for block space and gas fees rise. Raising capacity on the base layer itself is slow, because every node in the world has to keep up.

Then comes the layer 2, a separate system that executes transactions away from the base chain and then posts a compressed record, or a cryptographic proof, back to it. Ethereum.org describes a layer 2 as a separate blockchain that “extends Ethereum and inherits the security guarantees of Ethereum”. L2BEAT, the research site most people use to compare these networks, defines layer 2 as “a category of technical solutions aimed to scale the base layer in a trust minimized way” (L2BEAT FAQ). The two definitions point at the same idea: the layer 2 does the work, the layer 1 is the judge.

Well-known Ethereum layer 2s include Base (built by Coinbase), Arbitrum One, OP Mainnet, Starknet, ZKsync Era, Scroll and Linea. On Bitcoin, the best-known layer 2 is the Lightning Network, which works very differently and gets its own section below. Solana, Avalanche or BNB Chain are layer 1s: they compete with Ethereum rather than build on it.

There is also a third floor. A layer 3 is a chain that settles on a layer 2 instead of on Ethereum directly. L2BEAT tracked 13 of them as of 25 October 2026, the largest being Hyperliquid’s chain, which it lists on top of Arbitrum One. For a beginner the rule is the same at every floor: the higher you go, the more pieces of software stand between you and the base layer.

How a layer 2 inherits the security of its layer 1, and when it does not

“Secured by Ethereum” is the phrase you will read on almost every layer 2 website. It is true only under conditions, and those conditions are not met by most networks today. Ethereum.org itself warns that many layer 2 projects “are still young and somewhat experimental” and that none is as battle tested as Ethereum mainnet.

Three conditions: published data, a working proof, an exit

First, the transaction data must be published on Ethereum. If it is, anyone can rebuild the layer 2’s balances from scratch, even if the company running the network disappears. L2BEAT is strict on this point: if the data is stored somewhere other than a permissionless layer 1, the project “is not a rollup, but rather a validium or an optimium”.

Second, Ethereum must be able to check that the layer 2 is telling the truth. A smart contract on Ethereum holds the funds deposited into the layer 2 and only releases them against a valid claim. That check is done by a proof system, and the two families of rollups use two different kinds of proof.

Third, you need a way out that nobody can block. If the operator refuses your transactions or pushes a harmful upgrade, you should be able to force your withdrawal through Ethereum and have time to leave before the change takes effect. Without that exit, the operator’s keys matter more than Ethereum’s validators.

Optimistic and zero-knowledge rollups in two paragraphs

An optimistic rollup (Arbitrum One, OP Mainnet, Base) posts its results to Ethereum and assumes they are correct. Anyone watching the chain can dispute a wrong result with a fraud proof during a challenge period of roughly seven days, according to ethereum.org’s page on optimistic rollups. If nobody wins a challenge, the result becomes final. The design is cheap to run, and the price is the waiting time on withdrawals.

A zero-knowledge rollup (Starknet, ZKsync Era, Scroll, Linea) sends a validity proof with each batch: a piece of mathematics, a zero-knowledge proof, that Ethereum’s contract checks before accepting the new state. Producing the proof costs computing power, but once it is accepted there is nothing to wait for. How these proofs are built and how the challenge games run will come in a later lesson of this path.

L2BEAT stages: 10 of 82 layer 2s reach Stage 1 or 2

L2BEAT grades rollups on a scale it calls stages. Stage 0 means the project publishes its data and state on Ethereum and has a proof system, but its operators or a Security Council can still override it. Stage 1 requires that the only way, bugs aside, to block or fake a withdrawal is to compromise at least 75% of a Security Council, and that upgrades by anyone else give users at least seven days to exit. Stage 2 requires permissionless fraud proofs, at least 30 days to exit before an unwanted upgrade, and a Security Council limited to fixing bugs that can be proven on-chain.

The data shows how far the sector is from the slogan. As of 25 October 2026, L2BEAT’s summary listed 82 layer 2s. Only 4 were Stage 2 and 6 were Stage 1. Another 55 were listed at Stage 0, and 17 were filed under “Others” with no stage, because L2BEAT does not consider their proof systems fully functional.

Source: L2BEAT scaling summary data, 25 October 2026, about USD 39.2 billion secured across the 82 layer 2s. The value secured (TVS) moves with token prices: it was about USD 44.1 billion a month earlier, on 25 September 2026, as ether traded lower in between. A stage is an assessment on a given date and can change with any upgrade.

Most of the money sits on the handful of Stage 1 networks, about 70% of the total on that date, led by Base at about USD 15 billion and Arbitrum One at about USD 10 billion. And the Stage 2 networks, the ones closest to the textbook promise, hold almost nothing. A label such as “ZK rollup” says which proof is used, not how mature the network is: Mantle and Lighter are both filed as ZK rollups and both sit at Stage 0.

Even Ethereum’s co-founder has stopped selling the old picture. On 3 February 2026 Vitalik Buterin wrote on X that “the original vision of L2s and their role in Ethereum no longer makes sense” and called for a new path, as quoted by The Block. He argued that layer 2s had decentralised far more slowly than expected while Ethereum’s own base layer had advanced, and said layer 2s should still reach Stage 1 “at the minimum”.

Fees and speed: what a layer 2 changes, dated

The saving comes from sharing. Writing to Ethereum is expensive, so the layer 2 collects thousands of transactions, compresses them and pays for one write that all those users share. The biggest cost in that write used to be the data itself, posted as ordinary transaction data (calldata) that every Ethereum node keeps forever.

That changed on 13 March 2024, when the Dencun upgrade activated on Ethereum mainnet at 13:55 UTC, according to the Ethereum Foundation’s announcement. Dencun introduced EIP-4844, which added “blobs”: packets of 128 KiB that layer 2s can post at a separate, usually much lower, price. Ethereum nodes only keep blobs for 4,096 epochs, about 18 days, long enough for anyone to check the data and rebuild the layer 2’s state. Since then Ethereum has raised blob capacity three times.

Dates and figures from the Ethereum Foundation’s announcements for Pectra and Fusaka. Fusaka also brought PeerDAS, which lets nodes check that blob data is available by sampling pieces of it instead of downloading everything. More blob space per block means more room for layer 2 data, so more room before blob prices rise.

The result shows in daily data. On 24 October 2026 the median transaction cost was a fraction of a cent on every large layer 2 tracked by growthepie, while Ethereum mainnet stayed in the cents.

Source: growthepie fundamentals data, values for 24 October 2026. Medians hide spikes: Ethereum’s own median was close to 6 cents two days earlier, on 22 October. These fees are network fees only. An exchange charges its own withdrawal fee on top, and that is often the larger number for a small transfer.

Speed: a quick receipt, a slower final settlement

On a layer 2, a single company usually runs the sequencer, the machine that orders transactions. It answers quickly, so a payment or a swap looks done almost at once in your wallet. That answer is a promise from the operator. The transaction is settled in the Ethereum sense only after the batch is posted to Ethereum and, for an optimistic rollup, after the challenge period has passed. For everyday use inside the layer 2 the difference rarely shows. It shows the day you want to move money back to Ethereum.

Withdrawing from a layer 2 can take a week

Depositing into a layer 2 through its official bridge takes minutes. Leaving is another matter. On OP Mainnet, Optimism’s documentation says that a withdrawal to Ethereum takes approximately 7 days and three transactions: you start it on the layer 2, prove it on Ethereum, then finalise it once the challenge period is over. The docs add that this waiting period “is a core part of the security model of the OP Stack and cannot be circumvented”. Other chains built on the same software can set their own period, as Blast did when it cut its own to one day in October 2026. On Arbitrum the dispute window is about 6.4 days, according to Arbitrum’s docs.

Zero-knowledge rollups do not have this built-in delay. Ethereum.org notes that exits are executed once the rollup’s contract on Ethereum has verified the validity proof, so the wait depends on how often the operator posts proofs.

Most people do not wait a week. Third-party “fast bridges” and exchanges give you the money on Ethereum straight away, from their own liquidity, and collect your layer 2 funds later. You pay a fee for that, and you add a new party you have to trust.

The exit also protects you from censorship. On Base, according to L2BEAT, if the sequencer stops including your transactions you can send them through Ethereum and they must be included, with a delay of up to 12 hours. On Arbitrum One the delay can reach one day. Starknet, by contrast, lets users log a stuck transaction on Ethereum but not force it: a minority of its Security Council has to step in.

When the operator holds the keys: sequencers, upgrades and Blast’s shutdown

Two kinds of keys explain most layer 2 risk. The sequencer key decides which transactions get in and in what order. The upgrade keys, often held by a multisig wallet, can replace the contracts that hold everyone’s deposits on Ethereum. L2BEAT’s data on 25 October 2026 showed no exit window at all for unwanted upgrades on Base, Arbitrum One and Starknet, because their contracts can be upgraded instantly. On a Stage 1 network such an upgrade needs a Security Council’s approval: on Base, according to L2BEAT, both the Base Coordinator Multisig and the Base Security Council must sign.

Blast shows what happens when a layer 2 decides to stop. On 2 October 2026 Blast said it was winding down because the chain cost more to run than it earned, Bankless reported. The team said withdrawals would pause for about a week while Blast unwound its positions on Lido, then reopen with the usual wait cut to 24 hours, and it gave users until 26 October 2026 to move funds back to Ethereum through its regular interface. Funds left on Blast after that date are not lost, Bankless wrote, but getting them out means calling Blast’s bridge contracts on Ethereum directly, which is not a beginner’s task. Anyone still holding funds there should follow the instructions on Blast’s own channels.

Blast’s page on L2BEAT shows the rest of the picture. A 3-of-5 multisig controlled the system’s upgrades with no delay, the chain had no working state validation, and on 14 October 2026 an upgrade cut the finalisation period for withdrawals from 7 days to 1 day. L2BEAT’s page says Blast “is not even a Stage 0 project”, and the value it secured was down to about USD 44 million on 25 October 2026. At its peak, on 6 June 2024, the staking or liquidity, measured at current market prices. It is the main size metric in DeFi.">value locked in applications on Blast was about USD 2.26 billion, according to DefiLlama data. In this case the exit worked because the team kept the bridge open. A network that only reaches Stage 0 gives users no guarantee of that.

Bridges: where the biggest losses happened

A bridge moves value between two chains. Most work by locking your tokens in a contract on one chain and issuing a matching token on the other, or by burning on one side and minting on the other. Every rollup has an official bridge, often called the canonical bridge: it is the deposit contract on Ethereum described above, and its security is exactly the security of the rollup, stage included. Third-party bridges connect many chains at once, including layer 1s that have nothing to do with Ethereum, and each adds its own validators, signers or messaging layer.

Ethereum.org’s page on bridges distinguishes trusted bridges, where you rely on an operator who holds custody, and trustless ones that rely on smart contracts, while noting that every bridge carries smart contract risk. History backs the warning. In a post dated 2 August 2022, the day after the Nomad hack, Chainalysis estimated that USD 2 billion had been stolen across 13 separate cross-chain bridge hacks, and that bridge attacks made up 69% of all funds stolen in 2022 up to then (Chainalysis).

For Nomad, Mandiant’s analysis puts the loss at more than USD 190 million, taken by a crowd of copycats once the flaw was public. On Ronin, the FBI said in a statement on 14 April 2022 that it had confirmed Lazarus Group and APT38, cyber actors associated with North Korea, were responsible for the theft of USD 620 million in ether reported on 29 March, as reported by BleepingComputer. The Kelp DAO case is recent and still disputed. According to The Block, an attacker released about 116,500 rsETH from Kelp’s LayerZero-based bridge at 17:35 UTC on 18 April 2026, and Kelp’s emergency multisig paused the contracts at 18:21 UTC. Two days later LayerZero said, with what it called preliminary confidence, that North Korea’s Lazarus Group was likely behind it, and blamed Kelp’s choice of a single verifier, CoinDesk reported. That attribution is LayerZero’s allegation, not a finding by a court or a law enforcement agency, and Kelp replied the same day, The Block reported, that its single-verifier setup was the default LayerZero shipped for new deployments.

None of these four was a rollup’s canonical bridge to Ethereum. They were cross-chain bridges with their own keys and signers, which is the main reason the official route is usually the safer one when it exists. Our lesson on whether a blockchain can be hacked explains why the base chains themselves held up while the software around them failed.

For the reader, bridging safely comes down to a few habits. These are safety checks, not recommendations of any network.

• Use the bridge linked from the network’s own documentation, or withdraw from your exchange directly to the layer 2 when the exchange lists that network.
• Type the bridge address yourself or use a bookmark. Fake bridge and claim sites are a common way to drain wallets, as our lesson on fake airdrops and giveaways shows.
• Send a small test amount first and wait until it arrives before moving the rest.
• Approve only the amount you are bridging, not an unlimited allowance, and revoke approvals you no longer need.
• Check the network’s page on L2BEAT before putting a large sum on it: stage, who holds the upgrade keys, and how long an exit takes.

If funds do go missing after a bridge transfer or a signature on a fake site, the transaction hashes and addresses are what investigators and exchanges will ask for first. Our lesson on what to do after crypto is stolen lists the steps and the places to file a report.

Why a sidechain such as Polygon PoS is not a layer 2

A sidechain is a separate blockchain connected to Ethereum by a bridge. It has its own validators and its own consensus, and, in the words of ethereum.org, sidechains “do not post state changes and transaction data back to Ethereum Mainnet”. If a majority of its validators turned dishonest, Ethereum would have no way to notice or undo the damage. That is the line between a sidechain and a layer 2: Ethereum does not check a sidechain’s work.

Polygon PoS, the network most people simply call Polygon, is the best-known example. L2BEAT lists it in its scaling tables but under “Others”, with no stage, because it is not a rollup. Its L2BEAT page, as of 25 October 2026, shows state validation “None”, data kept available by an external proof-of-stake network of 104 validators (with a cap of 105 and permissioned entry), and contracts that can be upgraded instantly. Gnosis Chain gets the same treatment: its L2BEAT page says that “Ethereum contracts do not validate Gnosis Chain state transitions”, and withdrawals through its bridges need signatures from 4 of 7 bridge validators.

This does not make Polygon PoS useless or unsafe by default. It processed about 4.9 million transactions on 24 October 2026, more than Ethereum mainnet, and held about USD 3.8 billion according to L2BEAT. It means the security of your funds there rests on Polygon’s validators and bridge, not on Ethereum.

Between the two sits a middle category. A validium uses validity proofs checked on Ethereum, like a ZK rollup, but keeps its transaction data off Ethereum; an optimium does the same with an optimistic design. L2BEAT calls these “light” layer 2s. If the data holders withhold the data, users may be unable to prove what they own, even though every state change was proven.

Bitcoin’s layer 2: the Lightning Network

Lightning is a layer 2 in a different sense from Ethereum rollups. Two people lock bitcoin in a shared address with one on-chain transaction, which opens a payment channel. They then exchange signed balance updates between themselves, as many as they like, without touching the blockchain. When they close the channel, the final balance is written to Bitcoin in a second on-chain transaction. Payments can hop across several channels to reach someone you have no direct channel with.

Security works by deterrence. If one side tries to close the channel with an old balance that favours them, the other side has a time window to publish the newer signed state and take the cheater’s funds as a penalty. There is no batch posted to Bitcoin and no proof checked by Bitcoin, which is why Lightning is not a rollup and has no L2BEAT stage. The trade-off is that you, or a service you trust, must stay watchful while a channel is open.

The network is smaller than its reputation. The latest snapshot in mempool.space’s public data, dated 30 August 2026, counted about 32,700 public channels between about 16,200 nodes, with a total capacity of about 3,800 BTC. The same data put capacity at about 5,290 BTC on 26 October 2023, with about 62,300 public channels, and about 5,530 BTC in mid-January 2026. Public capacity is lower than three years earlier, and the number of public channels has roughly halved. These figures cover public channels only: private channels used by many wallets do not show up.

In practice, Lightning is used for small and frequent payments. Sending a large sum depends on finding channels with enough liquidity along the way, and each channel needs on-chain transactions to open and close, which cost ordinary Bitcoin fees.

Choosing a network as a US or UK user: checks before you send

For American and British readers, the regulated party in any of this is the exchange or broker, not the chain. Coinbase, which builds Base, or Kraken in the US, and FCA-registered firms in the UK, decide which networks you can withdraw to, and they publish the list in their apps. Layer 2 networks themselves are software. We found no specific rule from the SEC, the CFTC or the FCA on layer 2 networks or bridges as such, as of October 2026, so the protections you get depend mostly on the platform you use at each end.

Taxes follow the general rules. The IRS treats digital assets as property, and HMRC applies capital gains tax to disposals, but as of October 2026 we found no specific rule from either on bridging or on moving funds to a layer 2. Whether a bridge transfer is a move between your own wallets or an exchange of one token for another can matter, so keep records of each transfer and read our basics of crypto taxes before asking an adviser.

• Check that the receiving wallet or exchange supports the exact network you choose. An Ethereum address looks the same on Ethereum, Base, Arbitrum and Polygon, but funds sent on the wrong network may not arrive where you expect, and the exchange may be unable to recover them.
• Keep a little ether on the layer 2 to pay its gas: most Ethereum rollups charge fees in ETH, while Polygon PoS uses its own token.
• Look up the network’s chain ID on its official documentation before adding it to a wallet, rather than trusting a pop-up from a website.
• Before a large deposit, read the network’s L2BEAT page or ethereum.org’s network list, which labels networks by maturity, from Stage 0 up.

Before moving money to any layer 2, look up two lines on its L2BEAT page: who can upgrade the contracts, and how a withdrawal works if the operator stops. Blast’s users found out in October 2026 why the second line matters. Later lessons of this path cover rollup proofs and bridging step by step.