Arbitrum is a family of Ethereum layer 2 networks built on optimistic rollup technology: it executes transactions off the Ethereum main chain, posts compressed transaction data back to Ethereum, and relies on fraud proofs to guarantee that the resulting state is correct. In Blockchain scaling, it is one of the most widely used examples of moving computation away from layer 1 while inheriting Ethereum security.
Arbitrum was developed by Offchain Labs, and its flagship chain, Arbitrum One, opened to the public in 2021. The ecosystem also includes Arbitrum Nova, a lower-cost chain with a different data model, and Arbitrum Orbit, a framework for launching custom chains. Governance passed to the Arbitrum DAO, whose members vote with the ARB token.
How Arbitrum processes and secures transactions
Arbitrum works by having a sequencer order user transactions, execute them in an Ethereum-compatible environment, and periodically publish batches of compressed data to Ethereum. Anyone with that data can recompute the chain state, which is the property that makes Arbitrum a rollup rather than a sidechain.
Validity is assumed optimistically. Validators post assertions about the new state to a contract on Ethereum, and any honest party can dispute a wrong assertion during a challenge period of about one week. Instead of re-executing the whole disputed batch on Ethereum, Arbitrum uses an interactive bisection protocol: the two sides repeatedly split the disputed computation until they disagree on a single step, which Ethereum then executes to decide the winner.
The table below lists the main components and the role each plays in the system.
| Component | Role | Where it runs |
|---|---|---|
| Sequencer | Orders transactions and gives fast soft confirmations | Off chain, operated by a designated party |
| Batch poster | Publishes compressed transaction data | Writes to Ethereum calldata or blobs |
| Rollup contracts | Store assertions, bridge funds, resolve disputes | Ethereum layer 1 |
| Validators | Check assertions and open challenges | Off chain, connected to Ethereum |
| Delayed inbox | Lets users force transactions past the sequencer | Ethereum layer 1 |
The Nitro upgrade in 2022 rebuilt the execution layer around a fork of the Go Ethereum client and compiles it to WebAssembly for dispute resolution. This gave Arbitrum high EVM equivalence, so most Solidity contracts and developer tools work without modification.
Arbitrum One, Nova and Orbit chains
The Arbitrum family contains chains that share the same technology stack but make different data availability choices. That choice determines both the cost of a transaction and the trust assumptions a user accepts.
| Chain type | Data availability | Typical use | Main trade-off |
|---|---|---|---|
| Arbitrum One | Full data posted to Ethereum | DeFi, high-value applications | Highest security, higher cost |
| Arbitrum Nova | AnyTrust data availability committee | Gaming, social, high-volume apps | Cheaper, adds committee trust |
| Orbit rollup | Data posted to a parent chain | Application-specific chains | Operator controls configuration |
| Orbit AnyTrust | Custom data availability committee | Low-cost dedicated chains | Security depends on committee honesty |
AnyTrust assumes that at least two members of the committee are honest; if the committee fails to sign, the chain falls back to posting full data on its parent chain. Orbit chains can settle to Arbitrum One instead of Ethereum, which is why they are sometimes called layer 3 chains.
A further extension, Stylus, lets developers write contracts in languages that compile to WebAssembly, such as Rust, C and C++, alongside ordinary Solidity contracts. It targets compute-heavy workloads where WebAssembly execution is significantly cheaper than equivalent EVM code.
Arbitrum compared with other scaling approaches
Arbitrum is best understood as one design point among several scaling strategies. Its closest relative is Optimism, another optimistic rollup, while ZK-rollups and sidechains such as early Polygon represent different security models.
| Property | Arbitrum | ZK-rollup | Sidechain |
|---|---|---|---|
| Proof method | Fraud proofs, interactive | Validity proofs | Own consensus, no proofs to layer 1 |
| Withdrawal to Ethereum | About seven days via native bridge | Minutes to hours after proof | Depends on bridge design |
| Security source | Ethereum plus one honest validator | Ethereum plus correct cryptography | Sidechain validator set |
| EVM compatibility | Very high | Varies by implementation | Usually high |
| Proving cost | Low unless disputed | High, continuous proof generation | None |
Compared with Optimism and Base, which share the OP Stack, Arbitrum historically differed by using multi-round interactive proofs instead of single-round re-execution. Its BoLD dispute protocol made validation permissionless, so that any participant can defend the correct state within a bounded time, a step toward stronger decentralization.
Fees, finality and practical use
An Arbitrum transaction fee has two components: an execution charge for computation on layer 2 and a data charge for the share of the batch posted to Ethereum. Since the Dencun upgrade introduced EIP-4844 blobs in 2024, the data charge has fallen sharply, and ordinary transfers or swaps usually cost a small fraction of the layer 1 price.
Users experience several levels of finality, and choosing the right one matters for exchanges, bridges and payment services.
| Finality level | Approximate time | Who relies on it |
|---|---|---|
| Sequencer confirmation | Under one second | Wallets and most applications |
| Batch posted on Ethereum | Minutes | Exchanges and cautious services |
| Ethereum finality of the batch | About 15 minutes after posting | High-value transfers |
| Assertion confirmed | About one week | Native bridge withdrawals |
Users who cannot wait a week usually rely on third-party bridges or liquidity providers, which advance funds on the destination chain for a fee. This is faster but adds counterparty and smart contract risk that the native bridge does not have.
Limitations and common misconceptions
Arbitrum inherits much of Ethereum’s security, but not all of it. Several centralizing elements remain, and understanding them prevents overestimating the guarantees the network provides.
- Centralized sequencer: a single operator orders transactions, so it can delay or reorder them briefly; the delayed inbox lets users force inclusion after a waiting period of roughly a day.
- Upgradeable contracts: the DAO and a security council can upgrade core contracts, so users trust the governance process in addition to the code.
- Not instant withdrawals: the one-week challenge window is a security feature, not a temporary bug.
- Nova is not equivalent to One: AnyTrust chains carry an additional committee trust assumption.
- Fees are not fixed: during congestion, layer 2 execution fees rise, and data costs follow Ethereum blob demand.
Independent risk frameworks classify rollups by how much they still depend on trusted operators. Arbitrum One is generally regarded as one of the more mature rollups, yet it remains a system whose safety depends on correct upgrades and honest validators being present.
A plain-language summary
Ethereum is secure but busy, and busy roads are slow and expensive. Arbitrum is like an express lane built next to the highway. Cars travel quickly on the express lane, and every so often a summary of all the trips is written into the official record on the main road.
The clever part is how mistakes get caught. Arbitrum assumes the summaries are honest, but for about a week anyone can raise an alarm and prove a summary was wrong. Because one honest watcher is enough, people get much lower fees while still relying on Ethereum as the final judge. The price of that design is patience: moving money back to Ethereum through the official route takes about a week.
An example from practice
A small online marketplace that pays freelance designers in a dollar stablecoin processes about 4,000 payouts per month. On Ethereum layer 1, each transfer cost it between one and several dollars during busy periods, which ate into payments of 20 to 50 dollars. The team moves its treasury to Arbitrum One through the native bridge and sends all payouts there.
After the switch, a typical payout costs a few cents, and designers see the payment in their wallets within seconds thanks to the sequencer confirmation. For accounting, the marketplace waits until each batch is posted to Ethereum before marking a payout as settled. Once a quarter, when it needs funds back on layer 1, it plans the one-week native withdrawal in advance rather than paying a fast-bridge fee, so its monthly transaction costs fall by more than 95 percent without moving to a separate, independently secured chain.