The Future of AVSs Depends on Data Availability, Not Just Consensus

Modern rollups have outsourced consensus to L1s like Ethereum, making execution and settlement layers increasingly homogeneous. The true competitive edge for AVSs (e.g., shared sequencers, oracles, coprocessors) is secure, low-latency access to high-fidelity data from diverse sources.

• Execution AVSs need mempool data for MEV capture and fair ordering.
• Oracle AVSs need reliable real-world data feeds with cryptographic attestation.
• Coprocessor AVSs need verifiable access to historical state for complex computations.

General-purpose DA like Ethereum is expensive and slow for high-throughput AVSs. The future is a modular stack where AVSs plug into purpose-built DA layers like Celestia, EigenDA, or Avail for cost efficiency, then use Ethereum for ultimate settlement security.

• Cost: DA can be ~99% cheaper than calldata on Ethereum L1.
• Throughput: Dedicated DA layers offer 10-100x higher data bandwidth.
• Security: The security budget is right-sized for the application, from opt-in economic security to Ethereum-level guarantees.

Raw data is useless without verifiability. Next-gen AVSs will require data to arrive with intrinsic proofs—ZK proofs of validity, TEE attestations, or cryptographic signatures—creating trust-minimized pipelines from source to consumer.

• **Projects like Brevis, Lagrange, and Risc Zero enable ZK coprocessors that prove arbitrary state transitions.
• **Oracles like Pyth and Chainlink CCIP provide cryptographically signed data with on-chain verification.
• This shifts the security model from "trust the operator" to "verify the proof accompanying the data."

AVSs that control their data supply chain will outperform generic competitors. We'll see the rise of vertically integrated stacks like Espresso Systems (sequencing + DA), EigenLayer (restaking + DA + AVS ecosystem), and Near DA (chain + data availability).

• Moats are built by owning the data pipeline, not just the consensus client.
• Performance is unbounded by generic layer constraints.
• Economic models are captured from data sourcing to service delivery, creating sustainable fee generation.

Modularity's core promise—decoupling execution from consensus and DA—creates a critical failure domain. If Celestia or any external DA layer halts or censors, every rollup built on it freezes. This isn't a single-chain halt; it's a cascade failure across hundreds of sovereign chains, turning a modular advantage into a systemic vulnerability.

• Risk: A single DA layer outage can brick $10B+ in aggregated TVL.
• Reality: Validators agree on a block, but no one can prove what's in it.

EigenDA's security is slashed from Ethereum's consensus via restaking, but its data availability guarantees are probabilistic and limited by bandwidth. It's optimized for high-throughput, low-cost posting, not for ensuring every byte is retrievable forever. Under sustained spam attacks or targeted data withholding, proofs can fail, making fraud proofs impossible and forcing honest validators into inactivity.

• Throughput ≠ Guarantees: 10 MB/s target throughput doesn't ensure 100% data retrievability.
• Slashing is Reactive: Penalties occur after the fact, after chains may have already halted.

Data being 'available' on a DA layer is meaningless if it's not retrievable within the challenge window. Networks like Celestia and EigenDA rely on peer-to-peer networks for data dissemination. If a malicious actor controls enough nodes to withhold data fragments, they can prevent timely reconstruction, causing soft finality failures and stalling L2 state progression. This is a direct attack on the light client bridge to Ethereum.

• Attack Vector: Targeted data withholding against fraud proof challengers.
• Consequence: 7-day challenge windows become useless if data is unreachable.

Cross-chain messaging protocols like LayerZero, Axelar, and Wormhole assume the underlying chains are live and their states are verifiable. A DA failure on a source chain means no cryptographic proof can be generated for a cross-chain message. This doesn't just isolate one chain; it severs the entire interconnected rollup ecosystem, freezing asset bridges and composable DeFi across the modular stack.

• Domino Effect: One DA failure can halt billions in cross-chain liquidity.
• Architectural Flaw: Interop stacks have no fallback for a total DA blackout.

The only hedge against a single DA layer failure is redundancy—posting data to multiple providers like EigenDA AND Celestia. This immediately negates the cost savings that modular DA promises. If rollups must pay for 2-3x the data posting fees for safety, the economic argument for leaving Ethereum's monolithic security collapses. The market will converge on the simplest, safest option: Ethereum blobspace.

• Dilemma: Choose between systemic risk or 2x+ cost overhead.
• Result: Ethereum L1 becomes the dominant DA by default.

Solana, Monad, and Ethereum (as a rollup DA) provide a brutal truth: tight integration of execution, consensus, and DA eliminates coordination failure. The security model is unified and the data is inherently available to all validators. For high-value DeFi and stablecoin issuance, this simplicity is non-negotiable. The future may be multi-chain, but the highest-value state will remain on chains with monolithic security guarantees.

• Guarantee: Data availability is a cryptoeconomic certainty, not a service.
• Market Reality: >90% of stablecoin supply remains on Ethereum/Solana.

Celestia decouples consensus and data availability (DA), creating a modular stack. This proves that specialized, high-throughput DA is a primary scaling vector, not an afterthought.

• Enables sovereign rollups with their own execution and governance.
• Reduces L1 data costs by ~99% compared to monolithic chains like Ethereum for calldata.
• Creates a new security market where AVSs can choose DA based on cost and speed.

EigenDA leverages Ethereum's consensus and restaked security to provide high-throughput, low-cost DA. It's the canonical AVS for data availability within the EigenLayer ecosystem.

• Leverages Ethereum's trust via restaking, avoiding new trust assumptions.
• Targets 10 MB/s throughput, sufficient for hundreds of rollups.
• Integrates directly with major rollup stacks like OP Stack and Arbitrum Orbit.

AVS nodes need the latest state to validate operations. Without cheap, fast DA, they either fall out of sync or become prohibitively expensive to run, centralizing the network.

• High sync times (>1 hour) create liveness vulnerabilities and high hardware costs.
• Monolithic L1s like Ethereum charge a premium for data, making AVS economics unsustainable.
• Forces trade-offs between security (full nodes) and scalability (light clients).

Specialized data availability layers (Celestia, EigenDA, Avail) separate the data publishing and verification problem from consensus. This is the core infrastructure for scalable, secure AVSs.

• Guarantees data is published so any honest node can reconstruct state.
• Enables light nodes to securely verify data with cryptographic proofs (e.g., Data Availability Sampling).
• Unlocks horizontal scaling where execution and DA scale independently.

Restakers securing an AVS are ultimately securing its ability to read and write correct data. The DA choice is the primary risk vector after the consensus layer itself.

• Slashing risk is tied to data withholding or incorrect data attestation.
• Due diligence must shift from "who's the operator?" to "what DA layer do they depend on?".
• Yield will correlate with the cost and reliability of the underlying DA solution.

The winning AVS client software will be DA-agnostic, allowing operators to plug into Celestia, EigenDA, or Ethereum based on cost and latency requirements for specific applications.

• Builders will select DA like a cloud service (AWS vs. GCP).
• Clients like Polymer and Succinct are building interoperability layers for multi-DA.
• Creates a competitive market driving innovation and lower costs for all AVSs.