What Layer-2 Networks Change About Ethereum Fees

A layer-2 network (L2) lowers fees because it executes transactions away from Ethereum's main chain and posts only compressed data, and in some designs a proof, back to Ethereum. Every transaction in a batch shares the expensive resource, space on Ethereum. Fees do not disappear, though: an L2 fee still includes a share of the cost of publishing data to Ethereum, and the lower price comes with its own trade-offs.

Key takeaways

A rollup batches many transactions off the main chain and stores their data on Ethereum, so the cost of one Ethereum data post is split across the whole batch.
Since the Dencun upgrade on March 13, 2024, rollups can post that data as blobs, which have their own fee market. Since Fusaka (December 3, 2025), the blob fee has a floor tied to Ethereum's regular base fee.
An L2 fee has two main parts: an execution fee for running the transaction on the L2, and an L1 data fee for publishing its data to Ethereum.
Cheaper fees come with trade-offs that vary by network: withdrawal delays on optimistic rollups, sequencer control over ordering, upgrade powers, and bridge risk.
In the hypothetical worked example below, the data cost per transaction stays under one cent in every scenario.

What a layer 2 does differently

Ethereum's main chain is called layer 1 (L1). According to ethereum.org's layer-2 page, Ethereum acts as a settlement layer for these newer networks, and networks that store their data on Ethereum are often called rollups. The same page explains that making the main chain itself faster would require larger nodes and more specialized hardware, which would weaken decentralization.

A rollup has three jobs. A sequencer, the operator that receives user transactions, orders and executes them quickly on the L2. It then bundles them into a batch. Finally, the batch's data lands on Ethereum so that anyone can reconstruct the L2's state from it.

Optimistic and ZK rollups compared

Rollups differ in how Ethereum learns that a batch is correct, as ethereum.org's pages on optimistic rollups and ZK-rollups explain.

Feature	Optimistic rollup	ZK-rollup
How batches are accepted	Assumed valid unless someone submits a fraud proof	A validity proof is checked by an Ethereum contract before the new state is accepted
Withdrawal to Ethereum	Waits out a challenge period, roughly seven days per ethereum.org; each network sets its own	No challenge period; exits complete once the batch's proof is verified on Ethereum, so timing depends on how often the network posts proofs
Security assumption	At least one honest node watches and submits fraud proofs when needed	Rests on the cryptographic proof system; some proof systems need a trusted setup
Main cost drawback	Delay for native withdrawals	Generating and verifying proofs is costly (verification alone uses roughly 500,000 gas, per ethereum.org)

Optimistic rollups do offer a faster route out: ethereum.org notes that a liquidity provider can take over a pending withdrawal and pay the user on L1 in exchange for a fee.

How blobs changed the cost of data

Before March 2024, rollups posted their data as calldata, the same transaction input space ordinary Ethereum transactions use, and paid regular gas for it. Ethereum's upgrade history records the Dencun upgrade as going live on March 13, 2024. It added EIP-4844, which introduced data "blobs" that let rollups post data to mainnet for a short period of time.

What a blob is

Under EIP-4844, a blob holds 4,096 field elements of 32 bytes each, or 131,072 bytes (128 KiB), and costs 131,072 units of blob gas. Blob gas has its own base fee, which adjusts on a rule similar to EIP-1559 and was originally independent of regular gas. Since Fusaka, EIP-7918 keeps the blob base fee from settling below about one-sixteenth of the regular base fee, so very cheap blob space now depends partly on a quiet main chain.

How blob capacity has grown

Capacity is set by a target and a maximum number of blobs per block.

Change	Date (UTC)	Blob target / max per block
Dencun (EIP-4844)	March 13, 2024	3 / 6
Pectra (EIP-7691)	May 7, 2025	6 / 9
Fusaka BPO1	December 9, 2025	10 / 15
Fusaka BPO2	January 7, 2026	14 / 21

EIP-7691 raised the original 3 / 6 to 6 / 9 and changed how fast the blob fee adjusts. Fusaka went live on December 3, 2025. Its headline feature, PeerDAS, lets nodes check that blob data is available by sampling instead of downloading whole blobs. The Ethereum Foundation's mainnet announcement scheduled two "blob-parameter-only" (BPO) forks that raised the target and maximum, as shown in the table. As of October 23, 2026, the Foundation's blog had announced no later blob change, and ethereum.org listed the next upgrade, Glamsterdam, without a date.

More capacity does not by itself set a price. The blob base fee still rises when demand runs above the target, so L2 fees can rise when blob demand is high.

A worked example of one batch

Illustrative example. Every input below is an assumption chosen for round arithmetic, not a measured figure. A rollup batches 1,000 transactions, each averaging 110 bytes after compression. ETH is priced at $3,200 for the example. The batch fits in one blob, and the L1 transaction that carries the blob has an assumed fixed overhead of 80,000 gas at a 4 gwei regular base fee.

batch data      = 1,000 tx x 110 bytes = 110,000 bytes
blob fill       = 110,000 / 131,072 = 83.9%
blob cost (ETH) = 131,072 blob gas x blob base fee (gwei) x 10^-9
overhead (ETH)  = 80,000 gas x 4 gwei x 10^-9 = 0.00032 ETH = $1.02
cost per tx     = (blob cost + overhead) x $3,200 / 1,000

A blob costs 131,072 blob gas whether or not it is full. The lowest row uses 0.25 gwei because, under EIP-7918, that is the floor level for a 4 gwei base fee (4 x 8,192 / 131,072).

Assumed blob base fee	Blob cost for the batch	Batch total with overhead	Per transaction
0.25 gwei	$0.10	$1.13	about $0.0011
1 gwei	$0.42	$1.44	about $0.0014
20 gwei	$8.39	$9.41	about $0.0094

For comparison, a plain ETH transfer on L1 uses 21,000 gas, as ethereum.org's gas page states. At the same assumed 4 gwei base fee that is 0.000084 ETH, about $0.27.

Three lessons follow from the arithmetic. First, even in the 20 gwei blob scenario, the data share is about $0.0094 per transaction, far below the $0.27 L1 transfer, because 1,000 transactions share one blob and one overhead payment. The L2's own execution fee comes on top of that share. Second, at low blob fees the fixed overhead makes up most of the batch cost, so extra blob space alone barely lowers the figure. Third, the older calldata route costs more. Before Pectra, at 16 gas per non-zero byte, the rate in the Optimism docs and applied here to every byte as a simplification, the 110,000 bytes would need 1,760,000 gas, about $22.53. Since Pectra, EIP-7623 sets a floor of 40 gas per non-zero byte for data-heavy transactions, which makes it 4,400,000 gas for the data alone, about $56.32, or $0.056 per transaction.

The example leaves out proof costs, the sequencer's margin and the fact that real transactions vary in size.

What a user still pays on a layer 2

The fee a wallet shows on an L2 is not just the blob cost divided up. The Optimism documentation describes OP Mainnet's fee as an execution gas fee plus an L1 data fee. The execution fee works as on Ethereum: gas used times the base fee plus an optional priority fee. The L1 data fee covers publishing the transaction's data to Ethereum and, on chains that post blobs, depends largely on the blob base fee. Since the Isthmus upgrade, an operator fee can also be added.

Arbitrum's documentation splits the fee in a similar way: one part for the child chain's computation and storage, and one for the parent-chain cost of posting the transaction's data. It estimates a transaction's data footprint by compressing it and charging 16 gas per compressed byte, then adjusts the price over time to track what batch posters actually pay.

So a busy L2 raises the execution part, while a rise in Ethereum's base fee or blob base fee raises the data part. L2BEAT's costs page tracks what L2s pay Ethereum for data and proofs (September 23 to October 23, 2026 when opened) and notes these are costs incurred by the L2 itself, not fees paid directly by users. Live fees change constantly, so none are quoted here.

Trade-offs and risks to weigh

Who controls the network

L2BEAT's stages framework (last updated July 23, 2025) rates a rollup's decentralization, not its security. L2BEAT itself warns that a Stage 2 project can still be unsafe. Stage 1 requires that, bugs aside, the only way to indefinitely block an L2-to-L1 message is to compromise at least 75% of a Security Council. Upgrades from outside that council must give users at least a 7-day exit window. Stage 2 requires a permissionless fraud-proof system and at least 30 days to exit before unwanted upgrades.

On L2BEAT's scaling summary as opened on October 23, 2026, Arbitrum One, Base, OP Mainnet and Starknet were listed at Stage 1, Aztec Network at Stage 2, and Linea, Scroll and ZKsync Era at Stage 0. Labels change as projects upgrade; they show the range and are not endorsements. Separately, ethereum.org's ZK-rollup page lists centralized sequencers that can influence transaction ordering among that design's drawbacks.

Withdrawal delays and bridges

Moving funds between L1 and an L2 goes through a bridge. Ethereum.org's bridge page lists smart contract risk, counterparty risk for bridges that rely on outside validators, and open questions about how bridges behave under congestion or attack. It says bridges account for the top three biggest hacks in DeFi. For a dated figure, Chainalysis reported on August 2, 2022 that $2 billion had been stolen across 13 cross-chain bridge hacks, 69% of all funds stolen that year up to publication. Those incidents involved cross-chain bridges in general, not only rollup bridges.

Frequently asked questions

Are blobs stored permanently on Ethereum?

No. Under EIP-4844, nodes must serve blob data for 4,096 epochs, which the specification puts at around 18 days. Ethereum.org's ZK-rollup page calls this a serving obligation rather than a deletion deadline and says blobs are not permanent archival storage. Anyone who needs the data longer depends on archival services.

Does a layer 2 make my transactions private?

Not by itself. Ethereum.org's ZK-rollup page says that publishing transaction data is what makes permissionless verification possible and is not a privacy mechanism, and that validity proofs alone do not make transactions confidential.

Can I set a maximum on the L1 data fee?

On OP Mainnet, no. The Optimism fee documentation states that it is currently not possible to limit the maximum L1 data fee a transaction is willing to pay. Other networks may handle this differently, so check each network's documentation.

What if a sequencer refuses to include my transaction?

For ZK-rollups, ethereum.org says users who believe the operator is censoring them can submit the transaction directly to the rollup contract on Ethereum and force an exit without the operator's permission. How that escape route works on a particular network is a design detail to read in that network's documentation.

Is there an official Ethereum layer 2?

No. Ethereum.org states there is no "official" Ethereum layer 2, since anyone can create one. It adds that safety depends on a network's technology, smart contract security and maturity, and suggests starting with small transactions.

Checking a layer 2 before using it

Find out whether the network is an optimistic or ZK rollup and whether it posts its data to Ethereum. Native withdrawals from optimistic rollups take about a week unless a liquidity provider covers the exit for a fee.
Look up its stage and risk summary on L2BEAT and note the date you checked, since stages change.
Read the fee documentation to see the parts of the fee and whether the L1 data fee can be capped.
Check what the bridge you plan to use relies on, and start with a small amount, as ethereum.org suggests.
Compare the wallet's fee with the same action on L1 on the day, since both parts of an L2 fee move with demand.