Network profile

Arbitrum: an optimistic rollup with deep DeFi liquidity

Arbitrum executes off Ethereum and posts compressed data back. The security assumptions sit in the bridge and the sequencer, not in the marketing.

Layer 2 · Optimistic rollupChecked 8 Nov 2026

Arbitrum is an optimistic rollup: it executes transactions away from Ethereum, publishes the result back, and asks the settlement layer to trust the answer unless someone proves it wrong in time.

Native token
ETH for fees; ARB exists as a governance token and is not needed to transact
Type
Layer 2 optimistic rollup over Ethereum, with Nova as a cheaper variant
Consensus
No independent consensus: a sequencer orders transactions and Ethereum settles disputes
Fees
Execution gas on the rollup plus the cost of posting compressed data to Ethereum
Settlement
Ethereum, after a challenge window during which a state root can be disputed
Ecosystem
Deep DeFi liquidity, perpetual exchanges, gaming, derivatives, infrastructure
Explorer
arbiscan.io
Documentation
docs.arbitrum.io

How it works

A rollup moves execution off the base chain but keeps the data on it. Arbitrum runs an EVM-compatible environment, produces blocks quickly, and periodically posts a compressed batch of its transactions to Ethereum. Anyone can reconstruct the rollup state from that data, which is what makes the system more than a separate chain with a marketing arrangement.

«Optimistic» describes the default: a posted state root is assumed correct. If it is not, a challenger can open a dispute during a challenge window and force the claim to be resolved by the settlement layer. The design assumes at least one honest party is watching and can afford the bond. That is a real assumption: if nobody watches, an invalid state root stands.

Withdrawals to Ethereum therefore wait for the challenge window: that delay is the price of the security model, and it is why third-party «fast bridges» exist, fronting the money and taking the timing risk for a fee.

Nova uses a different data availability arrangement: a committee stores the data instead of posting everything to Ethereum. That is cheaper and it introduces a committee you have to trust, which is exactly the trade.

The operational centre is the sequencer, which orders transactions and gives users fast confirmations. It can go down, and it can, in principle, order transactions as it likes. The escape hatch is the delayed inbox on Ethereum: a transaction can be forced in from the base layer if the sequencer refuses, which turns censorship into a delay rather than a block.

Fees: the mechanism

Two costs, added together:

  • L2 execution. Gas for computation inside the rollup, priced far below Ethereum because the work is not replicated by thousands of nodes.
  • L1 data. The cost of putting your transaction's bytes on Ethereum, which is the part that historically dominated and the part that upgrades to Ethereum's data pricing changed the most.

This is why rollup fees track Ethereum's data market rather than the rollup's own popularity. A quiet week on Ethereum lowers fees here even if the rollup is busy; a spike in blob demand raises them. Wallets usually show one number, but the mechanism is two markets.

Priority fees on Arbitrum exist for ordering within the sequencer's queue, not for outbidding other rollups.

What is actually built there

Arbitrum's identity is DeFi depth. Perpetual exchanges, lending markets, options, yield strategies and aggregators run here because execution is cheap enough for frequent trading and liquidity is thick enough to absorb size. For a trader, that combination is the product.

Gaming and consumer applications exist in quantity on Nova and on the main rollup, usually because fees are negligible for small actions. Infrastructure — oracles, indexers, account abstraction tooling — is well developed because the code is EVM-compatible and teams port easily.

Governance matters here: token holders and a security council control upgrades, and an upgrade can change the rules of the bridge. That is normal for rollups at this stage and it is still a trust assumption.

Risks and limits

  • The sequencer is centralised. Fast confirmations are a service, not a protocol guarantee, and downtime is a real scenario.
  • Challenge windows delay exits. A withdrawal to Ethereum is not instant, and fast bridges substitute their own risk for that wait.
  • Fraud proofs need watchers. The model is only as good as the incentives and the people running validation software.
  • Upgrades are governed, not automatic. A security council can act quickly, which is useful in an emergency and uncomfortable in principle.
  • Bridge risk is separate. The canonical bridge is part of the rollup; third-party bridges and wrapped assets are not.
  • Liquidity is concentrated in a few protocols, so a single exploit can dent the ecosystem's depth more than the chain's uptime.

Official sources