On 10 November 2026, Base processed about 8.5 million transactions. Ethereum, the network Base is built on, processed about 2.1 million the same day. The median fee on Base was about 0.18 of a US cent; on Ethereum it was about 3 cents, according to growthepie. Five days earlier Base had cut its block time from 2 seconds to 200 milliseconds. Yet every one of those 8.5 million transactions ends up recorded on Ethereum, the slower and dearer chain.
That is the trick behind every layer 2. Layer 2 explained from the inside comes down to one transaction: you sign it in your wallet, a server orders it, a batch carries it to Ethereum, and a contract on Ethereum keeps the money it moves. Who touches it on the way, where the fee goes and what you are trusting at each step are the questions below, with figures of 10 and 11 November 2026.
Layer 2 explained in one transaction, from your wallet to Ethereum
Two earlier lessons set the stage, and this one does not repeat them. Our lesson on why blockchains need scaling shows why Ethereum cannot simply make its blocks ten times bigger: every node would have to re-run and store everything. Our comparison of layer 1 and layer 2 gives the definitions, the three conditions a real layer 2 must meet (its data published, a working proof, a way out) and the withdrawal delays. How gas and blob fees are priced on Ethereum itself is in our guide to Ethereum gas fees.
Here we open the machine. Picture a 5-dollar USDC payment sent on Arbitrum One. Your wallet signs it and sends it to a single server, the sequencer. Within a fraction of a second the sequencer puts it in order and your wallet shows it as done. A few minutes later, your payment and thousands of others are squeezed into one compressed package and posted to Ethereum. Then a claim about the new state of the whole chain lands on Ethereum too, and a proof system stands ready to reject that claim if it is false. Your money, meanwhile, never left an Ethereum contract.
Each of those steps has a name and a weak point, and each one shows up in the fee you pay.
What counts as a layer 2, and who decides
There is no official register. ethereum.org, the community site, describes a layer 2 broadly as a separate chain that extends Ethereum, and says that rollups “bundle (or ‘roll up’) hundreds of transactions into a single transaction on layer 1”. Under that reading, almost any network that settles on Ethereum qualifies. Our glossary entry on the rollup gives the short version.
The reference most people check is stricter. ETHBEAT (formerly L2BEAT), the research group that tracks these networks, now publishes from ethbeat.org; the old l2beat.com address redirects there. Its new name reflects a wider scope than layer 2s, but its stages framework and risk pages are unchanged. As of 11 November 2026 it lists 80 active layer 2s. Only 20 of them get a proper rollup label: 11 ZK rollups and 9 optimistic rollups. Two are validiums and one is an optimium, designs that keep their data off Ethereum. The other 57 sit in a bucket called “Others”: projects that call themselves layer 2s but lack something ETHBEAT considers essential, often a working proof system.
That gap is the real debate on what counts. A chain can post its data to Ethereum and still let a small group decide, with nothing to stop them, which state is correct. Marketing says “layer 2”; the tracker asks whether you could get your money back if the operator vanished or lied. In this lesson, “layer 2” means a rollup unless stated otherwise. Validiums, optimiums and sidechains get their own lesson later in the path, sidechains vs rollups. Networks built on top of a layer 2, sometimes called layer 3s, follow the same logic one floor up.
How a rollup works in five steps
Every rollup, optimistic or zero-knowledge, follows the same five steps. The names change from one project to the next; the jobs do not.
Step 1: your wallet signs, the sequencer puts transactions in order
A rollup is a full blockchain with its own blocks, its own state and, on most large networks, the same Ethereum Virtual Machine as Ethereum. Your wallet signs a transaction exactly as it would on Ethereum, only with a different chain ID, and sends it to the network’s RPC endpoint.
There it meets the sequencer, the machine that decides the order of transactions and builds the layer 2 blocks. On almost every rollup today there is one sequencer, run by the company behind the chain. Arbitrum’s documentation says it plainly: the sequencer of Arbitrum One and Nova “is run by Offchain Labs” (Arbitrum docs). Base’s sequencer is run by Coinbase.
Because one machine decides, confirmation is fast. On Base, a new block has come every 200 milliseconds since the Denim upgrade went live on 5 November 2026 at 18:00 UTC, down from 2 seconds, as we reported when the change was scheduled. Your wallet shows the transaction as confirmed almost at once.
That confirmation is a promise, not a settlement. Arbitrum calls it soft finality, “soft” because it “depends on the Sequencer keeping its promises”. The sequencer could in theory reorder or drop your transaction before it reaches Ethereum. In practice large sequencers have a reputation to keep, but the guarantee only becomes hard at step 2.
Step 2: the batcher compresses the data and posts it to Ethereum
Every few minutes, a component called the batcher (Arbitrum calls it the batch poster) gathers the transactions the sequencer has ordered, compresses them and sends them to Ethereum in one go. Arbitrum compresses its batches with Brotli, a general-purpose algorithm also used by web browsers, and posts them as blobs when the chain operator has enabled them, otherwise as ordinary transaction data called calldata.
Blobs are packets of data that Ethereum validators keep for about 18 days and then delete, cheaper than calldata, which every node keeps forever. Each blob holds 128 kilobytes. Eighteen days is enough for anyone who wants a copy to download it and rebuild the rollup’s state, which is the point: once the data is on Ethereum, nobody needs the sequencer’s permission to know who owns what.
This is the step that makes a rollup a rollup. Ethereum does not re-run your USDC payment. It only stores the compressed bytes that describe it. Storing is far cheaper than computing, and that difference is where most of the saving comes from.
Step 3: a state commitment lands on Ethereum
Data alone does not tell Ethereum what the result is. So a second actor, the proposer, regularly posts a short commitment: a single hash, called a state root or output root, that sums up every balance and every contract of the rollup after a given block. One hash of 32 bytes can stand for millions of accounts.
The rollup’s contracts on Ethereum record that claim. Withdrawals are later checked against it: to take ETH out, you show that your withdrawal is part of a state that Ethereum has accepted. So the question becomes simple and serious. Who checks that the claim is true?
Step 4: a proof decides who is right
Here the two families of rollups part ways. An optimistic rollup, such as Arbitrum One or OP Mainnet, assumes the claim is correct and leaves a window of several days during which anyone can challenge it with a fraud proof. If nobody proves fraud, the claim stands. A ZK rollup, such as Starknet or ZKsync Era, attaches a validity proof, a piece of cryptography that Ethereum checks before accepting the claim at all.
Optimistic rollups are simpler and cheaper to run; ZK rollups confirm withdrawals faster but cost more to prove. Some now mix the two. Base, which left the OP Stack for its own software in February 2026, switched on its Azul upgrade on 28 May 2026, according to Base’s documentation. Since then a claim can be backed by an attestation from a secure hardware enclave, by a ZK proof, or by both. With both, it is settled after 1 day. With only one, Base first kept the classic 7-day wait, then cut it to 5 days with its Beryl upgrade on 25 June 2026, per the Beryl notes; ETHBEAT’s Base page describes the same system. How fraud proofs work in detail is the subject of the next lesson of this path, optimistic rollups explained; ZK proofs get their own lesson, ZK rollups explained.
Step 5: the bridge contract on Ethereum holds the money
Your ETH or USDC on a layer 2 is, in the end, a claim on tokens locked in a contract on Ethereum: the rollup’s canonical bridge. When you deposit, the contract locks your tokens on Ethereum and the layer 2 credits you, usually within minutes. When you withdraw through that same bridge, you start the exit on the layer 2, wait until the state that includes it is accepted on Ethereum, then claim your tokens from the contract.
On an optimistic rollup that wait is the challenge window. ETHBEAT puts the total exit delay on Base at 5 days and 12 hours as of 11 November 2026: 12 hours of possible sequencer delay, a 20-minute checkpoint and 5 days for the state to be confirmed with a single proof. On Arbitrum One the challenge period alone lasts about 6.4 days, per the Arbitrum documentation, and OP Stack chains such as OP Mainnet take about a week. Third-party bridges and exchanges shorten the wait by lending you funds on the other side for a fee, a separate market covered in our guide to cross-chain bridges and liquidity.
Why a rollup transaction costs a fraction of a cent
On 10 November 2026 the median transaction on Ethereum cost $0.0298 and the median on Base $0.0018, about 16 times less, by growthepie’s count. The saving is not magic. It comes from paying Ethereum for storage only, and splitting that bill among thousands of users.
Two fees in one: execution on the layer 2 and a share of Ethereum’s data cost
Every rollup fee has two parts. The first pays for running your transaction on the layer 2 itself: gas used multiplied by the layer 2’s own base fee and tip, the same formula as on Ethereum, only with a much lower price per unit because the rollup has room to spare. The second part is your share of the cost of posting data to Ethereum.
The OP Stack, used by OP Mainnet, Ink and many smaller chains, adds both together, plus an operator fee on chains that set one. It estimates the size of your transaction after compression (with an algorithm called FastLZ) and prices it from Ethereum’s current base fee and blob fee, according to the Optimism documentation. You cannot cap that data part in your wallet: if Ethereum gets busier between your click and the batch, it moves. Arbitrum works the same way, with an L1 data fee charged on every transaction that comes in through the sequencer, on top of the layer 2 gas.
On Arbitrum One, priority fees (tips) go 97 percent to the Arbitrum DAO treasury and 3 percent to a developer guild, per the Arbitrum documentation. The data charge, the same documentation says, “is paid to the Sequencer to cover its costs”, which means the bill for posting batches to Ethereum.
Compression and batching, with a worked example
Take a live reading. At about 09:44 UTC on 11 November 2026, Ethereum block 26,390,869 carried 266 transactions and 8 blobs, read from a public Ethereum node. The blob base fee in the blocks around it was about 3.4 million wei per unit of blob gas. A blob uses 131,072 units, so one blob cost about 0.00045 ETH, or roughly 93 US cents with ether at $2,065.98 at 08:00 UTC that day. These are approximate figures from our own arithmetic.
Now an illustration, not a measurement: if a batch of 2,000 small, compressed transfers fits into one blob, each user’s data share is under 0.05 of a cent. Add the layer 2 execution fee and the operator’s margin, and you land in the range of the medians above. The more users a rollup has, the more transactions share each blob, and the cheaper each one gets, up to the point where Ethereum’s blob space itself fills up.
Blobs: the change of March 2024
Before 13 March 2024, rollups posted their data as calldata and paid the same gas price as everyone else, which kept their fees at tens of cents or more. The Dencun upgrade added blobs with their own fee market, and fees on rollups fell sharply. Ethereum has since raised the blob count several times; since 7 January 2026 it targets 14 blobs per block, with a maximum of 21. The dates and the blob price formula are in our gas fees guide; the history of blobs gets a lesson of its own later in this path, blobs and EIP-4844.
Fees and activity on 10 November 2026
The table below gives growthepie’s figures for one full UTC day. Medians hide spikes, and they count network fees only, not what an app or exchange adds on top.
| Network (10 Nov 2026) | Median fee (USD) | Transactions | Daily active addresses | Fees paid by users (USD) |
|---|---|---|---|---|
| Ethereum (layer 1) | 0.0298 | 2,064,096 | 664,428 | 349,656 |
| Base | 0.0018 | 8,468,276 | 348,168 | 130,798 |
| Arbitrum One | 0.0048 | 1,289,367 | 107,461 | 21,143 |
| OP Mainnet | 0.000025 | 1,859,879 | 393,928 | 2,826 |
| Starknet | 0.0037 | 93,767 | 1,663 | 2,340 |
| Linea | 0.0167 | 15,659 | 3,220 | 876 |
| Scroll | 0.0022 | 16,313 | 2,004 | 215 |
Ethereum’s own median moved a lot that week, from $0.0521 on 5 November to $0.0129 on 7 November, while Base stayed between $0.0018 and $0.0020 every day and Arbitrum One between $0.0044 and $0.0048. That steadiness is the other half of the layer 2 offer: the layer 2 part of the fee barely moves when Ethereum gets busy, and the data part is shared.
The main layer 2 networks in November 2026
Value secured, the measure ETHBEAT uses, counts the tokens held by a rollup’s contracts and on the chain. It moves with the price of ether: the total across all projects tracked fell from about $43.0 billion on 12 October 2026 to about $36.1 billion on 11 November, while ether fell from about $2,535 to about $2,066. That is mostly a price effect, not money leaving.
| Network (11 Nov 2026) | Type per ETHBEAT | Stage | If the sequencer ignores you | Value secured (USD bn) |
|---|---|---|---|---|
| Base | Optimistic rollup (own stack, TEE + ZK proofs) | 1 | Force via Ethereum, up to 12 h | 13.82 |
| Arbitrum One | Optimistic rollup (Arbitrum) | 1 | Force via Ethereum, up to 24 h | 9.03 |
| Robinhood Chain | Others (Arbitrum stack) | 0 | No mechanism | 2.06 |
| OP Mainnet | Optimistic rollup (OP Stack) | 1 | Force via Ethereum | 1.43 |
| Mantle | ZK rollup (OP Stack) | 0 | Force via Ethereum | 1.17 |
| Ink | Optimistic rollup (OP Stack) | 1 | Force via Ethereum | 0.37 |
| Starknet | ZK rollup | 1 | Request logged on Ethereum, cannot be forced | 0.32 |
| Linea | ZK rollup | 0 | No mechanism | 0.30 |
| ZKsync Era | ZK rollup | 0 | Queue via Ethereum | 0.21 |
| Scroll | ZK rollup | 0 | Force via Ethereum | 0.03 |
Source: ETHBEAT API, read at 08:00 UTC on 11 November 2026. Two names are missing on purpose. Polygon PoS, with about $3.74 billion, is filed under “Others” and is a sidechain with its own validators; Hyperliquid, about $6.14 billion, is listed as a layer 3. Neither fits the rollup model described here.
Two networks dominate. Base and Arbitrum One together secure about $22.85 billion of the roughly $36.3 billion held by the 80 active layer 2s, about 63 percent. Base, launched by Coinbase in 2023, also leads by activity, as the fee table shows. Arbitrum, built by Offchain Labs and governed by the Arbitrum DAO, is the oldest of the large optimistic rollups. Optimism runs OP Mainnet and maintains the OP Stack, the open-source software Ink and many other chains are built with. Base, the largest chain launched on it, left the OP Stack and the Superchain group in February 2026 to run its own software.
Most of the ZK rollups are far smaller by value. Starknet, ZKsync Era, Linea and Scroll each secured well under half a billion dollars on 11 November 2026. They prove every batch with cryptography, which is stronger in principle, but the stage column shows that stronger proofs do not by themselves remove the human controls around them.
Stages 0, 1 and 2 in plain words
ETHBEAT grades rollups on one question: how much do you depend on the people who run it? Its stages page, last updated on 23 July 2025, sets three levels.
Stage 0 is a rollup with training wheels fully on. It posts state roots and data to Ethereum and its software is public, so anyone can rebuild the state, but the operators can still overrule the proof system or change the contracts at will. Fifty-five of the 80 layer 2s were at Stage 0 on 11 November 2026, including Linea, ZKsync Era, Scroll, Mantle and Robinhood Chain.
Stage 1 means a working proof system decides, with a Security Council of outsiders as a backstop. The only way, bugs aside, to block a withdrawal for good is by “compromising ≥75% of the Security Council”. If someone outside that council can upgrade the contracts, users must get at least 7 days to leave first. Six networks were at Stage 1 on 11 November 2026: Base, Arbitrum One, OP Mainnet, Ink, Starknet and Unichain. Together they held about $25.1 billion, some 69 percent of all value on layer 2s.
Stage 2 removes the remaining hand on the brake. Fraud proofs must be open to anyone, users must have “at least 30 days to exit in case of unwanted upgrades”, and the Security Council may act only on errors detected on chain. On 11 November 2026, three small projects met it: Facet, Cartesi’s PRT Honeypot v2 and Aztec Network v6, with under $1 million between them. None of the large networks has reached Stage 2.
A simple way to remember it is the walkaway test. ETHBEAT’s Base page says Base passes it: if the team walked away tomorrow, users could still get their funds out through Ethereum. Passing that test is the core of Stage 1. Stage 2 adds the promise that nobody can change the rules under your feet faster than you can leave.
What you trust on a layer 2 today
A layer 2 at Stage 1 is far from a bank, and far from Ethereum itself too. Four things still depend on people.
The sequencer can delay you
A single sequencer can ignore your transaction. ETHBEAT notes that Base’s sequencer gives “no real-time censorship resistance”. What you get instead is a delayed escape: you can send your transaction to the rollup’s contract on Ethereum and force it in. On Base, the sequencer has up to 12 hours to include it. On Arbitrum One, a well-behaved sequencer picks up such messages after about 10 minutes; the maximum is 24 hours, after which, the documentation says, “anyone can force it to be promoted to the main inbox”. On Linea and Robinhood Chain there was no such mechanism at all as of 11 November 2026.
So the sequencer cannot steal your funds, but it can make you wait, and a few hours of delay in a market crash is real money. If it simply goes down, you wait too: sequencer outages have happened on several networks, a topic for the later lesson on layer 2 risks and ETHBEAT.
Upgrade keys with no exit window
Rollups are sets of upgradeable contracts on Ethereum, and someone holds the right to upgrade them. On Base, all contracts sit behind an admin controlled by two multisigs that must both sign: the Base Coordinator Multisig (3 of 6 signers) and the Base Security Council (8 of 11). There is no delay before an upgrade takes effect, per ETHBEAT. The same “no exit window” applies to Arbitrum One, OP Mainnet and Starknet. In plain words: a large enough group of key holders could change the rules, including the bridge, before you could react. The safeguard is who those people are, not code.
The withdrawal wait and the bridge itself
The canonical bridge is the only exit that relies on Ethereum alone, and it is slow, as step 5 showed. Faster routes add a third party. And because all the money of a rollup sits in one bridge contract, that contract is the prize for attackers: the largest crypto hacks of past years hit bridges, a history told in our layer 1 vs layer 2 lesson.
Rollups can also be shut down by their own teams. Blast, an Ethereum layer 2 that is closing, set a withdrawal deadline; when it passed in October 2026, about $40 million was still sitting in its bridge, although the money could still come out through the contracts, as we reported on 26 October. A closure is easier on users who move early than on those who never hear about it.
Ethereum keeps the data and the money; small groups of key holders still keep the right to change the contracts. A network’s stage and its exit delay are listed on its ETHBEAT page, and both are worth a look before moving a large sum.
Using a layer 2 in practice
For a user, a layer 2 looks like another network in the same wallet. Most wallets that support Ethereum can add Base, Arbitrum One or OP Mainnet with a click; the network name, chain ID and RPC address come from the network’s own documentation, not from a random website. Your address stays the same, because the layer 2 uses the same account format as Ethereum. Your balances do not: ETH on Base and ETH on Ethereum are separate, and the wallet shows them network by network.
On most large Ethereum layer 2s, including Base, Arbitrum One and OP Mainnet, fees are paid in ETH. A few use their own token. So you need a little ETH on the layer 2 itself before you can do anything, even move USDC. At the medians of 10 November 2026, a few dollars of ETH covers hundreds of ordinary transactions on Base.
To get funds there, you have three routes in general terms. Some exchanges let you withdraw directly to a layer 2: pick the network in the withdrawal form and check it twice, because the same address exists on every network and a wrong choice can strand funds. You can deposit through the rollup’s canonical bridge, which takes minutes in and days out. Or you can use a third-party bridge, faster but with its own risk. A step-by-step guide comes later in this path, how to bridge to layer 2. We do not recommend any bridge or app; send a small amount first, wait for it to arrive, then the rest.
Finally, know which kind of network you are on. A rollup at Stage 1 and a network filed under “Others” can look identical in a wallet. ETHBEAT’s summary page shows the difference, and the stage of any network changes over time, up or down.
Layer 2s and your taxes in the United States and the United Kingdom
Moving to a layer 2 does not change the tax rules that apply to what you do there. A swap, a sale or a payment on Base or Arbitrum is treated like the same operation on Ethereum. In the United States the IRS states that “digital assets are considered property, not currency”. Its digital assets page, which explains the yes-or-no digital asset question on the federal return, says a transfer between wallets or accounts you own or control is not a digital asset transaction, unless you paid a transaction fee with digital assets. Layer 2 fees are tiny, and paid in ETH they still belong in your records.
In the United Kingdom, HMRC taxes disposals of cryptoassets under capital gains rules, and keeping a record of every fee and every network helps when you work out your gains. The country rules, thresholds and forms are in our guide to crypto taxes in the United States and our guide to crypto taxes in the United Kingdom.
