What Is Data Availability (DA): Complete Modular Blockchain Guide (2026)

If you have heard people talk about modular blockchains, rollups, or the future of Ethereum scaling, you have probably stumbled across a strange three-letter acronym: DA. Data Availability sounds boring on the surface, like something only protocol engineers should care about. The truth is the opposite. Data Availability is the invisible backbone that lets cheap, fast Layer 2 networks remain secure. Without it, every rollup on the planet collapses into something closer to a centralized database.

In 2026, Data Availability has become its own multi-billion-dollar category. Celestia mainnet has been live for over two years, EigenDA processes serious throughput for restaking-secured rollups, Avail has its own L1 token, NearDA offers the cheapest published bytes in the market, and Ethereum itself ships EIP-4844 blobs that anchor Arbitrum and Optimism. Five serious DA layers compete for the same job: storing rollup data long enough for anyone to challenge invalid state transitions.

This guide explains Data Availability in plain English. We cover what DA actually means, why rollups die without it, how Data Availability Sampling lets a phone verify gigabytes of data, the five major DA layers and their prices, which rollups use which DA, the failure modes that keep researchers up at night, and where the technology is heading. By the end you will understand why DA is the single most important primitive in modern blockchain scaling.

What Is Data Availability (DA)?

Data Availability is the property that a block of transaction data has been fully published to the network and that anyone who wants to download that data can do so. That is the entire definition. Data is available if you can retrieve it. Data is unavailable if some validator or sequencer is hiding part of it.

This sounds trivial but it is one of the deepest problems in distributed systems. When a block producer announces a new block, all they really publish is a small header containing a hash commitment to the block contents. The actual data, which can be megabytes of transactions, may or may not be sitting on their server somewhere. If they refuse to release the bytes behind that hash, no one can verify what happened. They can claim the block contains transaction X, but X is a hash. You cannot read a hash.

In a monolithic chain like early Bitcoin or pre-rollup Ethereum, this problem is solved brute-force. Every full node downloads every block in full. If a block producer hides data, no validator will accept the block. The cost is that running a full node requires storing the entire chain forever, which is why Bitcoin nodes today need hundreds of gigabytes and Ethereum archive nodes need terabytes.

Modular blockchains and rollups try to escape that cost. A rollup posts a compressed summary of its activity to an external DA layer instead of forcing every user to download every byte. But the compression only works if the underlying bytes are guaranteed to exist somewhere that anyone can fetch. That guarantee is Data Availability, and the system that provides it is the DA layer.

Why DA Matters for Rollups

Rollups are the dominant scaling solution in 2026. Optimistic rollups like Arbitrum and Optimism, and ZK rollups like zkSync and Starknet, all share one fundamental design: they execute transactions off-chain and post the results to a settlement layer. The settlement layer, usually Ethereum, only stores enough information to verify the rollup state is correct. This is why a rollup can run at 50x the throughput of Ethereum L1 for a fraction of the cost.

But the entire security model collapses if rollup data is unavailable. Consider what happens if Arbitrum's sequencer posts a state root to Ethereum claiming "the new balance of address 0xABC is 1000 ETH" but refuses to publish the transactions that justify that claim. No one can verify whether those transactions were legitimate. No one can construct a fraud proof. The rollup has effectively become a private database run by the sequencer with Ethereum providing nothing but rubber-stamp finality.

This is why every rollup must publish its transaction data to some DA layer that anyone in the world can read. For optimistic rollups, the data is needed so watchers can detect fraud and submit fraud proofs during the challenge window. For ZK rollups, the data is needed so users can reconstruct their own balances if the sequencer goes offline. Without DA, you cannot exit a rollup safely. You are trusting the operator, which is exactly what crypto is supposed to eliminate.

The technical term is "DA guarantee." A rollup with strong DA guarantees is one where the DA layer has been proven, cryptographically and economically, to make data retrievable. A rollup with weak DA guarantees is essentially a custodial service with extra steps. The choice of DA layer therefore directly determines how trustless a rollup is. Use Ethereum 4844 blobs and you inherit Ethereum's security. Use a centralized data committee and you inherit the trust assumptions of that committee.

The Modular Blockchain Stack

To understand where DA fits, you need to see the full modular stack. Modular blockchains break the monolithic design into four specialized layers, each optimized for one job. This separation is the entire thesis behind the Celestia roadmap and the modular movement.

In a monolithic chain, all four jobs are performed by the same validator set. In a modular stack, each job runs on the network best suited for it. Execution layers optimize for throughput and EVM compatibility. Settlement layers optimize for security and bridge finality. DA layers optimize for cheap, high-bandwidth data publishing. Application layers optimize for user experience. The result is a system where each component does one thing well rather than all things badly.

Notice that Data Availability sits at the bottom of the stack. It is the foundation. Every other layer depends on DA being honest. If the DA layer fails, execution cannot be verified, settlement cannot resolve disputes, and applications cannot guarantee their state. This is why DA is sometimes described as "the most important layer no one talks about."

Data Availability Sampling (DAS): How Light Clients Verify

The killer feature of a dedicated DA layer is Data Availability Sampling, usually shortened to DAS. DAS solves a problem that seems mathematically impossible: how can a small device, like a phone or a Raspberry Pi, verify that a 100 MB block has been fully published, without downloading the 100 MB?

The answer is statistics. A light client running DAS does not download the full block. Instead, it downloads a tiny number of random pieces, called samples. If the block producer is honest, every sample will be retrievable. If the block producer is hiding even a small portion of the data, every random sample has a high probability of hitting the hidden region, exposing the dishonesty.

The math is elegant. If a block producer publishes more than 50% of the data, you can mathematically reconstruct the rest using erasure codes (we get to those next). So hiding any data means hiding more than 50% of it. The probability that 30 random samples all land in the published 50% is roughly one in a billion. So with 30 samples, a light client gets nearly perfect certainty about the availability of a megabyte-scale block by downloading only a few kilobytes.

DAS is what makes Celestia, Avail, and (eventually) Ethereum's full Danksharding design viable. It is the reason DA layers can scale to gigabytes per second without forcing every user to run server-grade hardware. A mobile phone can sample a Celestia block in less than a second and reach the same security conclusion as a node with a 10 gigabit fiber connection.

Reed-Solomon Erasure Coding: The Math Foundation

For DAS to work, the data has to be encoded in a way that hiding any small piece becomes equivalent to hiding a huge piece. This is done with Reed-Solomon erasure coding, the same family of codes used in QR codes, satellite communications, and RAID storage.

The idea is to take the original data, treat it as a polynomial, and evaluate that polynomial at many extra points. The resulting encoded data is twice as large as the original, but you can reconstruct the entire original from any 50% of the encoded version. If a malicious block producer wanted to make even 1% of the original data permanently unavailable, they would have to withhold more than 50% of the encoded data. Hiding 50% is loud. It is statistically impossible to do quietly under random sampling.

Celestia uses a 2D Reed-Solomon scheme. The block is arranged as a square matrix, each row and each column is independently erasure-coded, and the commitments to each row and column are placed in the block header. This lets a light client cheaply verify which row or column is missing data if the producer cheats, and submit a compact fraud proof to the rest of the network. The 2D structure is critical because it makes fraud proofs constant-size regardless of block size, which is a property single-dimension erasure codes do not have.

EigenDA and Avail use different encoding strategies (KZG polynomial commitments for Avail, modular variants for EigenDA), but the core insight is the same: encode the data so partial withholding becomes equivalent to total withholding, then let light clients sample randomly. The cryptographic glue varies. The economic security model does not.

Ethereum 4844 Blobs: The Original DA

Before Celestia and the dedicated DA chains, Ethereum itself launched the first production-scale DA solution: EIP-4844, known as proto-danksharding, activated with the Dencun upgrade in March 2024. EIP-4844 added a new transaction type called blob-carrying transactions, which let rollups post data as "blobs" that are temporarily stored on Ethereum.

Each blob is 128 KB of raw data, and a block can carry up to 6 blobs by default, with a target of 3 per block. Blob data is not stored forever like normal calldata. Validators only keep it for around 18 days, which is more than enough for fraud proof windows on optimistic rollups and validity proof generation on ZK rollups. After 18 days, the data is dropped from the consensus layer but anyone interested can archive it elsewhere.

The economic impact was immediate and dramatic. Before 4844, Arbitrum and Optimism were posting transaction data as calldata, which competed for the same gas market as every other Ethereum transaction. Rollup transaction fees often spiked to over $1 during gas wars. After 4844, the same data goes into the dedicated blob market, which has its own EIP-1559 style fee mechanism. Costs dropped 10-100x. A typical Arbitrum swap that used to cost $0.30 now costs around $0.01-0.05.

EIP-4844 is secured by the full Ethereum validator set, which makes it the strongest DA available in terms of economic security. Roughly 1.1 million ETH-equivalent validators (about $4 billion in staked capital as of 2026) attest to blob data being available. This is why almost every major optimistic rollup (Arbitrum, Optimism, Base, Blast, Linea) defaults to blobs as their DA. The tradeoff is throughput. Ethereum's current blob capacity is around 0.375 MB per slot, which limits total rollup data to roughly 4 MB per minute across the entire ecosystem.

Celestia: The First Dedicated DA Chain

Celestia launched its mainnet in October 2023 as the first blockchain designed from the ground up to do nothing but Data Availability. It removed execution and smart contracts from its base layer entirely. Celestia validators do not run an EVM. They do not process transactions in the traditional sense. They publish data and produce DA attestations. That is the entire job.

This radical specialization gives Celestia massive throughput. The current Celestia mainnet supports 2 MB blocks at a 12-second block time, providing roughly 8 MB per minute of DA bandwidth. The roadmap targets 8 MB blocks (32 MB per minute) by late 2026 and eventually 1 GB blocks once full DAS networks are stable. That is two to three orders of magnitude more capacity than Ethereum blobs.

Celestia uses the TIA token for paying for blob space and securing the network through proof-of-stake. The pricing is denominated in TIA per byte, which translates to roughly $0.002-0.05 per MB depending on demand and TIA market price. This is substantially cheaper than Ethereum blobs in most market conditions. Detailed economics live in our Celestia and TIA guide.

Celestia pioneered the production-scale rollout of Data Availability Sampling. Anyone can run a Celestia light node on a laptop, a Raspberry Pi, or even some smartphones, and the light node performs DAS independently. This decentralizes verification in a way no other DA layer has achieved at scale. The rollups using Celestia include Manta Pacific, Eclipse, Movement Labs, and dozens of smaller sovereign rollups built with frameworks like Rollkit and Sovereign SDK.

EigenDA: EigenLayer-Secured DA

EigenDA is built on top of EigenLayer restaking and went live in 2024 as the first major Actively Validated Service (AVS). The pitch is straightforward: instead of bootstrapping a new validator set and a new token, EigenDA borrows Ethereum's existing validators via restaked ETH. Operators run EigenDA nodes alongside their Ethereum validators, putting their restaked ETH at risk if they fail to publish data correctly.

The architecture is high-throughput by design. EigenDA targets 10-15 MB per second of DA bandwidth in 2026, which dwarfs both Celestia and Ethereum blobs in raw capacity. It achieves this by skipping the consensus-on-data-availability step that other DA layers perform. Instead, EigenDA operators sign attestations off-chain that the data is available, and these signatures are aggregated and submitted to Ethereum L1 as a single proof.

The tradeoff is the security model. EigenDA does not currently implement full DAS with light client verification. Instead, it relies on a quorum of restaked operators to honestly report data availability. If those operators collude, they can lie about availability and only get slashed if a separate party submits a custody proof. This is a different trust model than Celestia or Ethereum blobs, and it has been the subject of intense debate in the modular research community.

EigenDA is used by Cyber, Mantle (for some workloads), Celo's L2 plans, and several emerging L3s that need very high throughput at low cost. Pricing on EigenDA is among the lowest in the market, often quoted at fractions of a cent per MB. The combination of low cost and Ethereum-aligned security (via restaking) makes it attractive for high-volume rollups that want to stay in the Ethereum ecosystem.

Avail: The Polygon Spinoff

Avail started as a research project inside Polygon Labs and spun out as an independent network in 2024, launching its own mainnet and AVAIL token. The chain runs on a Substrate-based stack with KZG polynomial commitments for erasure coding, and it ships full Data Availability Sampling with verifiable light clients from day one.

Avail's main competitive angle is interoperability. It bundles three products: Avail DA (the data layer), Avail Nexus (a cross-rollup messaging layer), and Avail Fusion (a security aggregation layer that lets Avail inherit security from multiple sources including Bitcoin and Ethereum). For rollup builders, this stack is the only one offering native cross-rollup messaging baked into the DA layer itself.

Polygon's CDK (Chain Development Kit), which is the framework used to build the Polygon zkEVM and dozens of OP-Stack alternative chains, supports Avail as a first-class DA backend. This gives Avail a built-in pipeline of customers from the Polygon ecosystem. Rollups like Sophon and a number of gaming chains have launched on Avail DA throughout 2025 and 2026.

Pricing on Avail is competitive with Celestia, typically in the $0.005-0.02 per MB range depending on AVAIL token price and demand. The light client implementation in Avail is one of the most aggressive in the space, with native browser-based DAS clients that let dApps verify availability directly in the user's browser without any external infrastructure.

NearDA: The Cheapest Tier

NearDA is a DA service built on top of the Near Protocol blockchain. Unlike Celestia or Avail, Near was not designed primarily for DA. It is a general-purpose smart contract chain with sharded execution. But Near's sharded design happens to produce abundant cheap blockspace, and the team realized they could repurpose that bandwidth into a competitive DA offering.

NearDA targets the cheapest end of the market. Pricing routinely lands at $0.0001-0.001 per MB, which is roughly 10-100x cheaper than Celestia and 100x-1000x cheaper than Ethereum blobs. This makes NearDA attractive for rollups where DA cost is the dominant expense, particularly gaming chains, social chains, and any application that posts large volumes of data with low per-transaction value.

The tradeoff is the security model. Near's validator set is smaller and less battle-tested than Ethereum's, and NearDA does not yet ship full DAS for external rollups. Rollups that pick NearDA are typically optimizing for cost over security, which is a perfectly reasonable tradeoff for many use cases but unsuitable for high-value financial applications.

Aurora (Near's EVM L2) is the most prominent NearDA user, but several gaming rollups including some on the Caldera and Conduit rollup-as-a-service platforms have adopted Near DA for cost reasons. The Near team is actively working on DAS integration that would close the security gap with Celestia, though the timeline is uncertain.

5-DA Market Comparison

Here is how the five major DA layers stack up in 2026. Prices are approximate ranges that vary with token markets and demand.

The right DA layer depends on what the rollup is optimizing for. High-value financial rollups that need maximum security pick Ethereum blobs even at higher cost. High-throughput general-purpose rollups balance Celestia or Avail for the cost-security tradeoff. Gaming and social rollups that prioritize cost above all else pick NearDA or EigenDA. There is no single winner because DA is fundamentally a tradeoff space.

Which Rollups Use Which DA

The mapping below shows which production rollups currently use which DA layers. This landscape changes constantly as rollups switch DA providers, sometimes adopting multi-DA strategies for redundancy.

Notice that the largest rollups by TVL all sit on Ethereum 4844. This is not an accident. Liquidity follows security, and Ethereum's validator set provides the strongest DA security guarantee available. Smaller and newer rollups, especially those focused on gaming, social, and consumer applications, can stomach the security tradeoff of an alternative DA layer in exchange for cost savings of 10-100x.

DA Pricing Economics: Cost per MB

To understand why DA choice matters economically, consider a high-throughput rollup processing 50 transactions per second with an average data size of 200 bytes per transaction. That works out to 25 MB per hour, or 600 MB per day, or roughly 18 GB per month of DA usage.

At Ethereum blob pricing of $0.10 per MB, that rollup spends roughly $1,800 per month on DA. At Celestia pricing of $0.01 per MB, the same workload costs $180. At NearDA pricing of $0.0005 per MB, it costs $9. The spread between most-expensive and cheapest DA is roughly 200x for an identical workload.

This is why DA choice is one of the first decisions a rollup team makes. A rollup expecting consumer-scale traffic of millions of transactions per day cannot afford Ethereum blobs at scale. Those numbers blow past what users will pay in fees. Switching to Celestia or EigenDA can take the same rollup from economically infeasible to highly profitable.

The flip side is that DA cost is passed to users in transaction fees. A rollup paying $0.10 per MB for DA must charge users enough per transaction to cover that cost plus sequencer revenue. A rollup paying $0.001 per MB can offer transactions for sub-cent fees and still earn a margin. This is why the gas fees on Celestia-secured rollups like Manta are usually 10-50x lower than equivalent transactions on Arbitrum or Base.

DA Failure Modes

DA layers are critical infrastructure, which means they have critical failure modes. Understanding these failure modes is essential for evaluating which DA layer to trust with what kind of value.

Data withholding attacks are the canonical DA failure. A malicious block producer publishes the block header (with a commitment to the data) but refuses to release the underlying bytes. Without DAS, full nodes that haven't yet downloaded the block cannot detect this. They see a valid-looking header. Honest light clients sample the data and quickly discover the withholding because their samples come back empty. They emit a fraud proof, the network rejects the block, and the producer gets slashed. Without DAS, this attack succeeds.

Light client convergence problems happen when different light clients see different fragments of the data and cannot agree on whether the full block is available. This is a known issue in 2D Reed-Solomon schemes: a producer can release just enough data that some samples succeed and some fail. Honest light clients then split into groups that believe different things. The Celestia consensus protocol handles this with "bad encoding proofs" that any honest party can submit if the commitments themselves are malformed, but the implementation is subtle and has been the subject of multiple rounds of formal verification.

Validator collusion is the failure mode for DA layers that don't ship full DAS. If a quorum of validators (typically 67% in proof-of-stake systems) colludes to publish false DA attestations, light clients have no way to detect the lie. EigenDA's current security model depends on this assumption holding: restaked operators are economically rational and won't risk slashing for short-term profit. The question is how much value can sit on top of EigenDA before this assumption breaks.

Bridge desync is a special failure mode for rollups using external DA. If the rollup's bridge to Ethereum (or whatever its settlement layer is) depends on reading attestations from the DA layer, and the DA layer pauses or forks, the bridge can desync. Users may be unable to withdraw funds even though both the rollup and the settlement layer are functioning. This is why DA layer reliability matters at the operational level, not just the security level.

The Future: ZK-DA, Danksharding, Native DAS in Ethereum

The DA landscape in 2026 is the early innings. Three big technical evolutions are coming over the next 18-36 months that will reshape the market.

First, full Danksharding is Ethereum's endgame for DA. EIP-4844 was only proto-danksharding. The full version targets 1.3 MB per slot of blob capacity (around 30 MB per minute), with native DAS implemented at the consensus layer so any light client can sample blob data without trusting full nodes. This would put Ethereum on equal footing with dedicated DA chains for throughput while preserving its security advantage. The timeline depends on the rollout of PeerDAS, which is on most current Ethereum roadmaps for 2027.

Second, ZK-DA is an emerging design where ZK proofs are used to certify availability without erasure coding. Instead of sampling, light clients verify a succinct cryptographic proof that the data has been distributed across a quorum of operators. This removes the need for 2D Reed-Solomon and reduces light client bandwidth even further. Polygon, Starkware, and several research groups are actively developing ZK-DA implementations.

Third, multi-DA is becoming the production norm. Rather than picking one DA layer, sophisticated rollups post data to multiple DA layers in parallel, accepting whichever attestation arrives first. This provides redundancy: if one DA layer goes down, the rollup keeps operating from the others. We expect multi-DA to become the default by 2027, especially for high-value financial rollups that cannot tolerate any DA downtime.

Risks

Even with all the cryptographic guarantees, DA carries real risks. A DA layer compromise where a majority of validators collude or a bug in the consensus client could expose every rollup that depends on that DA layer to data withholding. The blast radius depends on how many rollups are using the affected DA at the time.

Validator failures are a softer risk. If 33% of a DA layer's validators go offline simultaneously (during a major cloud outage, for example) the chain may halt. Rollups using that DA cannot post new data, which means their sequencers cannot produce new blocks. Users see frozen transactions. This is not theoretical: every major chain has experienced multi-hour halts at some point.

Token-based DA layers like Celestia, Avail, and Near also have token price risk. If TIA or AVAIL collapses 90%, the security budget of the DA layer collapses proportionally. A formerly-secure DA layer can become economically attackable if its token bleeds for long enough. This is why some research suggests the safest DA over very long time horizons remains Ethereum blobs, whose security scales with ETH market cap.

Finally, regulatory risk is hovering. DA layers store data on behalf of third parties. Some jurisdictions could classify DA operators as data processors or even as financial intermediaries, depending on how they interpret the technology. The legal status of running a DA node in Europe under the upcoming MiCA Phase 2 framework is still being clarified. None of this is acute today, but it is on the horizon.

FAQs

Is Data Availability the same as data storage?

No. Data Availability only requires that data be retrievable for a limited window, typically long enough for rollup fraud proof or finality periods (days to weeks). After that window, the DA layer is allowed to forget the data. Long-term storage is a separate concern handled by archive services like Filecoin, Arweave, or simply paid storage providers. DA is "is this data retrievable right now?" Storage is "is this data retrievable in 10 years?"

Can a rollup switch DA layers later?

Technically yes, but it is a major upgrade that requires coordinated bridge changes and usually a governance vote. Once a rollup is live and holds substantial TVL, switching DA layers is painful because the bridge contract on the settlement layer must be updated to read from the new DA's attestation mechanism. We have seen rollups announce switches (Mantle moved from Ethereum calldata to EigenDA) but it remains rare.

Do I need to care about DA as a regular user?

Indirectly, yes. The DA layer your favorite rollup uses determines your transaction fees, your security guarantees, and your ability to withdraw funds if the sequencer fails. If you hold significant value on a rollup, knowing which DA layer it uses tells you who you are ultimately trusting. Rollups using Ethereum 4844 have the strongest user guarantees. Rollups using alt-DA layers offer cheaper fees but trade off varying amounts of security.

What is the difference between DA and consensus?

Consensus is about agreeing on the order of transactions. DA is about making the transaction bytes retrievable. A chain can have strong consensus and weak DA (validators agree on a hash but hide the data), or strong DA and weak consensus (data is published but validators disagree on order). Modular blockchains separate these concerns intentionally. The scalability trilemma can be partly broken by treating DA and consensus as independent dimensions.

Why does Vitalik talk so much about Data Availability?

Because DA is the bottleneck for Ethereum's rollup-centric scaling roadmap. Vitalik's essays have consistently framed Ethereum as a "rollup hub" where the main chain provides settlement and DA, while execution moves to rollups. The success of this thesis depends entirely on DA scaling. Without 4844, the rollup-centric roadmap would have stalled. Without future Danksharding, rollups will hit a hard throughput ceiling. DA is the variable that determines how big the rollup ecosystem can grow.

Are DA tokens (TIA, AVAIL) good investments?

This guide does not give financial advice, but the structural argument for DA tokens is that they capture fees from every byte every rollup publishes. As the rollup ecosystem grows, DA demand grows proportionally. The risks are competition (five major DA layers fighting for the same market), Ethereum 4844 absorbing demand at higher security, and execution risk on each DA team's roadmap. Position-size accordingly and always do your own research.

Conclusion

Data Availability is the silent foundation of modular blockchains. It does not capture headlines like new L2 launches or token unlocks, but every L2 launch and every token unlock depends on a DA layer doing its job invisibly in the background. Get DA right and the rollup ecosystem scales gracefully. Get DA wrong and rollups become custodial databases pretending to be blockchains.

In 2026, the DA market has consolidated around five real players: Ethereum 4844 for maximum security, Celestia for dedicated sovereign rollup support, EigenDA for restaking-aligned throughput, Avail for cross-rollup interoperability, and NearDA for ultra-cheap workloads. Each has its niche. Each is being battle-tested by real economic value sitting on top. The competition between them is one of the most important storylines in blockchain infrastructure.

Whether you are a developer choosing a DA layer for a new rollup, a user evaluating where to deposit funds, or just a curious observer trying to understand modular scaling, the takeaway is the same: ignore DA at your peril. The next several years of blockchain scaling will be defined by which DA layers win, which fail, and which new designs emerge. Pay attention to rollup frameworks, sampling implementations, and on-chain bridge security, because that is where the action is happening. Data Availability is no longer a niche research topic. It is the layer that makes everything else possible.

Related Guides

• What is Celestia (TIA)? Modular Data Availability Explained
• What Is a Modular Blockchain: Celestia and the New Stack (2026)
• What Is IoTeX (IOTX)? Modular DePIN Platform for IoT + AI Explained in 2026
• What Is Aleo? The Zero-Knowledge Privacy Layer 1 Blockchain Explained in 2026
• NEAR vs Solana: Layer 1 Blockchains Compared (2026)

Frequently Asked Questions

What is data availability in blockchain?

Data availability refers to the guarantee that the data behind a block has been published and can be retrieved by anyone who needs to verify it. It is important so that participants can check the validity of transactions and state.

What is a data availability layer?

A data availability layer is a part of a modular blockchain stack dedicated to making transaction data available for verification. Rollups and other systems can publish their data to such a layer instead of storing everything on a single chain.

What is data availability sampling?

Data availability sampling is a technique where nodes check small random pieces of block data to gain confidence that the whole dataset is available. It allows lightweight participants to verify availability without downloading everything.

How does data availability relate to modular blockchains?

Modular blockchains separate functions such as execution, settlement and data availability into specialized layers. Dedicating a layer to data availability can help rollups scale while keeping their data verifiable.