Arbitrum

Arbitrum lowers data costs with AnyTrust committee certificates

Arbitrum lowers batch publication costs through AnyTrust when a data availability committee supplies the transaction data nodes need for verification. Committee members store compressed batches and sign certificates, while Nitro handles execution and settlement through the parent chain. Rollup mode publishes the batch data directly on that chain. Data availability means nodes can obtain the inputs needed to reconstruct execution. AnyTrust adds a committee honesty assumption, and its cost advantage can shrink when full-data fallback becomes necessary.

Bottom line: AnyTrust’s lower publishing costs depend on committee signatures, so full-data fallback belongs in the same cost comparison.

Lower data overhead for repeated contract updates

AnyTrust’s lower publication overhead can benefit applications generating frequent contract updates, provided their chain can obtain enough committee signatures. The sequencer orders transactions, and the batch poster groups their data for publication. Compact certificates keep the full batch out of the normal parent-chain payload. Nodes retrieve the underlying bytes from data availability servers, then process the transactions under Nitro’s rules. This separates the cost of making inputs available from the computation contracts perform. A workload with many updates can therefore reduce its publication burden without changing its contract logic solely for the data model.

Execution still consumes resources. Choosing committee storage doesn’t remove gas charges, application-level costs, or the parent chain’s role in settling results. Withdrawal timing follows the chain’s settlement and exit configuration.


Certificate signers and honest committee members

An AnyTrust committee secures data availability through signed storage promises, with the applicable quorum set by the chain’s registered keyset. A Data Availability Certificate, or DACert, identifies a batch through its hash. It also records an expiry and evidence of the participating signatures. The committee’s servers hold the bytes matching that hash. Hashes let nodes check they retrieved the intended data; they don’t supply missing bytes themselves.

The number of signatures required and the number of members assumed honest describe different parts of the trust model.

In the standard model assuming at least two honest members, a committee with N members needs at least N - 1 signatures. Omitting only one member leaves at least one honest signer if the assumption holds. That signer must actually store and provide the batch during the promised retention period. Having two honest members doesn’t imply the batch poster can collect enough signatures from two servers. Certificate production needs the signing quorum, while the availability argument needs an honest signer within it.

Dedicated chains can configure a different honesty assumption. If H counts the assumed honest members, the required signatures follow K = N + 1 - H. Reducing the signature requirement by raising H strengthens the honesty assumption. The live keyset establishes the chain’s parameters; another network’s committee size doesn’t establish them.

Keep publishing when committee signatures fall short

If the batch poster can’t collect enough committee signatures, it can post the complete batch to the parent chain. Nodes can read either the full-data format or a valid certificate and its corresponding batch. This fallback lets the poster publish new batches during a temporary committee signing outage. It still needs a functioning batch poster and usable parent-chain publication. Higher posting costs can return during the fallback period.

A missing signature and unavailable data behind an accepted certificate pose different problems. Fallback handles insufficient signatures before the poster commits a certificate. It doesn’t turn a dishonest storage promise into accessible data afterward. If the committee violates the chain’s honesty assumption and withholds the certified bytes, independent verification can fail. The additional trust therefore remains material even though a fallback path exists. Parent-chain settlement can’t reconstruct inputs it never received.


A committee outage changes the publication path

In this hypothetical AnyTrust setup, a committee has 18 members and assumes at least two honest members, so a certificate needs 17 signatures. Assume available members store and serve batch data and return valid signatures. The batch poster and parent chain remain operational throughout this comparison.

With one member offline, the remaining 17 members can meet the threshold. The poster can publish a certificate, and the committee supplies its matching batch data. If two members go offline, only 16 signatures remain available. They don’t meet the requirement. The poster can then use full-data publication, replacing the certificate’s storage promise with the batch bytes on the parent chain.

Both cases supply the data Nitro needs, under those assumptions, but their parent-chain payloads differ. A Rollup chain uses parent-chain batch publication from the outset. Its data path has no committee-signature threshold. AnyTrust reduces normal publishing overhead while introducing a dependency whose failure can change that overhead. A budget built solely around certificate posting misses this fallback exposure.

Fee inputs beyond data storage

An AnyTrust fee estimate still reflects the execution a transaction consumes and the chain’s policy for recovering publication costs. Execution demand, minimum fee settings, and batch posting costs can all affect the charge. Nitro accounts for execution and parent-chain publication separately. Publication charges apply to sequencer-batched transactions when the chain’s fee settings enable them. Committee certificates reduce the data sent upward in normal operation; they don’t freeze the execution price. A contract performing more computation can cost more than a simpler call on the same chain. Comparing identical operations makes the data model’s contribution easier to understand.

Ethereum blobs also make Rollup data publication cheaper than an older comparison based only on calldata would suggest. AnyTrust’s advantage has no permanent multiplier.

When publication fees are enabled for full-data batches, parent-chain prices and the transaction’s estimated compressed size affect the publication charge. Dedicated chains can set their own economic policies, so a protocol-level saving needn’t pass through as an identical user discount. Committee operators also need storage and serving infrastructure. Their operational costs remain part of running the chain, even when users don’t see a separate committee charge. For predictable budgeting, execution demand and possible full-data posting deserve separate estimates.

Separate AnyTrust configuration from Nova’s maintenance mode

Dedicated Arbitrum chains can select AnyTrust as a data availability option while configuring other aspects of their operation independently. Choosing it means arranging committee servers, registering the relevant keyset, and configuring the batch poster and nodes consistently. Data storage, signing availability, and key management become ongoing operating responsibilities. A chain’s sequencing and validation arrangements require their own settings. Sharing Nitro doesn’t give every deployment identical control structures or service commitments. AnyTrust can suit workloads that value lower publication overhead and accept this extra committee dependency.

Illustration: Separate AnyTrust configuration from Nova’s maintenance mode (Arbitrum)

Open full-size image

Arbitrum Nova launched as an AnyTrust chain, and the ArbitrumDAO approved a minimization plan for it in 2026. Nova entered its minimized state by September 2026. The plan specifies normal batch posting through Ethereum blobs, a passive committee, and reduced infrastructure and support. Under that design, dormant committee members can remain registered while the poster publishes data directly to Ethereum. The plan preserves the network in maintenance mode rather than fully shutting it down. These changes concern Nova’s operating model; they don’t discontinue AnyTrust as a configuration option for dedicated chains.

A chain publishing full-data batches still incurs their publication costs even if its configuration retains committee keys. Registered membership alone doesn’t establish an active certificate-based data path.

Common questions

Can an AnyTrust chain rotate its committee members?

An authorized chain administrator can replace or rotate committee members by registering a new keyset. Older keysets remain valid until the administrator removes them. The keyset identifies the signing keys and establishes the required quorum. Registering or removing a keyset changes which certificates the chain accepts, without automatically reconfiguring every server. The batch poster and nodes need compatible settings for the revised committee.

What does a DACert’s expiry mean for historical data?

A DACert’s expiry marks the end of the committee’s promised data retention period for that batch. It doesn’t guarantee permanent historical storage. Archive data availability servers preserve expired batches so later nodes can reconstruct history. A chain relying only on servers that discard expired data can lose that historical retrieval path, even after earlier nodes processed the batch.

Does AnyTrust keep transaction data confidential?

AnyTrust doesn’t automatically make transaction data confidential. Public data availability servers expose batch retrieval, which lets nodes reproduce execution. A dedicated deployment may impose data access controls, but those are separate configuration choices. Full-data fallback also publishes batch bytes on the parent chain, so offchain storage alone can’t establish a confidentiality guarantee.

Is a DACert proof of correct smart contract execution?

A DACert certifies a signed data availability promise. Its signature doesn’t independently establish that a contract’s execution or output was valid. Nodes still need the underlying transactions and Nitro execution rules to reconstruct state, and validators assess state assertions. The committee’s storage role and the validation role protect different parts of the system.

Which tokens can pay gas on a dedicated AnyTrust chain?

A dedicated AnyTrust chain can use ETH or a compatible custom ERC-20 gas token. The operator chooses and configures the native fee currency for that deployment. AnyTrust doesn’t require a particular custom token, and compatibility requirements still apply. Custom gas tokens aren’t exclusive to AnyTrust; Rollup chains can support them with their required configuration.

Can a mirror DAS supply a missing committee signature?

A read-only mirror Data Availability Server (DAS) serves stored batches and doesn’t add a committee signature. The quorum counts signatures from the configured committee membership, not extra copies of the same batch. Mirrors can improve retrieval capacity and resilience, and archive mirrors can preserve historical batches. The signing quorum still needs the required committee members to return valid signatures.

Do AnyTrust chains have to settle directly on Ethereum?

Dedicated AnyTrust deployments can run as layer 3 chains on a supported layer 2 parent. A layer 2 deployment can settle directly to Ethereum, while a layer 3 deployment uses a supported layer 2 parent. Committee certificates and full-data fallback target that parent’s inbox. The added layer changes the settlement relationship; it doesn’t remove the committee’s data availability assumption.

Last updated -