Data availability is the hard limit. A rollup's throughput is not defined by its execution speed but by how cheaply and quickly its data can be posted to a secure base layer like Ethereum. This is the scalability ceiling.
zk-rollups-the-endgame-for-scaling
Blog
Why Data Availability Layers Will Dictate Rollup Scalability
An analysis of how the bandwidth, pricing, and security of underlying data availability layers—Ethereum, Celestia, Avail, EigenDA—form the fundamental bottleneck for all rollup ecosystems.
introduction
THE DATA
The Hidden Bottleneck
Rollup scalability is ultimately constrained by the cost and speed of publishing transaction data to a base layer.
Execution is cheap, data is not. A rollup sequencer can process 100k TPS internally, but publishing that data to Ethereum L1 at 80 KB per block costs more than the execution itself. This creates a cost-per-transaction floor.
The DA layer dictates economics. Solutions like EigenDA, Celestia, and Avail compete by offering cheaper data blobs, directly lowering transaction fees for rollups like Arbitrum and Optimism. This is a commoditization race.
Evidence: Posting 1 MB of data to Ethereum costs ~0.1 ETH. Posting the same data to Celestia costs ~$0.01. This 1000x cost differential is the primary scaling vector for the next generation of rollups.
key-insights
THE BOTTLENECK IS DATA
Executive Summary
Rollup throughput is no longer constrained by execution; it's limited by the cost and speed of publishing transaction data to a base layer.
01
The Problem: Ethereum as a Data Sink
Publishing data to Ethereum L1 is the single largest cost for rollups like Arbitrum and Optimism. This creates a hard ceiling on scalability and user fees.\n- Cost: ~80% of a rollup's operational expense is L1 data posting.\n- Throughput: Limited by Ethereum's ~80 KB/s data bandwidth.
80%
of Rollup Cost
80 KB/s
Ethereum Cap
02
The Solution: Dedicated DA Layers
Specialized data availability layers like Celestia, EigenDA, and Avail decouple data publishing from consensus/execution. They offer orders of magnitude more bandwidth at a fraction of the cost.\n- Scale: Celestia targets ~100 MB/s, a 1000x+ increase.\n- Cost: Sub-cent transaction data costs vs. dollars on L1.
1000x+
More Bandwidth
>99%
Cost Save
03
The Trade-Off: Security vs. Sovereignty
Choosing a DA layer is a fundamental security and political decision for a rollup. It defines whose crypto-economic security backs the chain's data.\n- Ethereum DA: Maximum security, higher cost, shared sovereignty.\n- External DA: Lower cost, modular flexibility, new trust assumptions.
High
Ethereum Security
Low
External DA Cost
04
The Future: DA as a Commodity
DA is becoming a hyper-competitive, commoditized service. The winning layers will be those that optimize for prover efficiency (for zk-rollups like zkSync and Starknet) and light client verifiability.\n- Key Metric: Cost per MB/GB of data posted.\n- Endgame: Rollups will multi-home across DA layers for redundancy and cost optimization.
$/GB
Pricing Metric
Multi-Home
Rollup Strategy
thesis-statement
THE BOTTLENECK
The First-Principles Argument
Rollup scalability is fundamentally limited by the cost and speed of publishing transaction data.
Data availability is the bottleneck. A rollup's execution speed is meaningless if its data cannot be posted to a base layer for verification. The cost of this data posting, not computation, is the dominant expense for rollups like Arbitrum and Optimism.
Execution scales, data does not. A rollup can process 10,000 TPS internally, but it is throttled by the ~80 KB/s data bandwidth of Ethereum. This creates a hard ceiling that execution-only scaling cannot solve.
DA layers break the ceiling. Dedicated data availability layers like Celestia, EigenDA, and Avail decouple data publishing from Ethereum's constraints. They provide orders of magnitude more bandwidth at a lower cost, directly determining a rollup's feasible throughput.
Evidence: The cost to post 1 MB of data on Ethereum L1 is ~$500. On Celestia, it is ~$0.01. This 50,000x cost differential dictates which applications are economically viable on a rollup.
THE BOTTLENECK IS DATA
DA Layer Benchmark Matrix: The Throughput & Cost Ceiling
Comparative specs of leading Data Availability layers that set the hard limits for rollup scalability and cost.
| Core Metric / Feature | Ethereum (Blobs) | Celestia | Avail | EigenDA |
|---|---|---|---|---|
Current Throughput (MB/s) | ~0.75 MB/s | ~14 MB/s | ~6 MB/s | ~10 MB/s |
Cost per MB (Current Est.) | $20-80 | $0.10-0.50 | $0.30-1.00 | $0.05-0.20 |
Data Availability Sampling (DAS) | ||||
Proof System | KZG Commitments | Namespaced Merkle Trees | KZG + Validity Proofs | Restaking + KZG |
Settlement & Consensus Coupling | Tight (L1 Consensus) | Loose (Modular) | Loose (Modular) | Loose (Restaked Security) |
Time to Finality (Data) | ~20 min | ~15 sec | ~20 sec | ~5 min |
Max Blob Size / Block | ~1.3 MB | ~8 MB | ~2 MB | Dynamic (Sharded) |
Primary Use Case | Ethereum-aligned L2s | Sovereign & High-TPS Rollups | General-Purpose Modular Stack | High-Security, Ethereum-Native Apps |
deep-dive
THE DATA BOTTLENECK
The Modular DA War: Security vs. Scale
Rollup scalability is fundamentally constrained by the security and throughput of the underlying data availability layer.
The DA layer is the bottleneck. Rollups execute transactions, but they must publish transaction data for verification and state reconstruction. The speed and cost of this data posting determines the rollup's effective throughput and user cost.
Security models dictate architecture. Ethereum's consensus-level security (via calldata or EIP-4844 blobs) provides maximal liveness guarantees but imposes hard bandwidth limits. EigenDA and Celestia offer higher throughput by separating data availability consensus from execution, trading some security assumptions for scale.
The trade-off is verifier decentralization. High-throughput DA layers rely on a smaller set of data availability committees or light clients. This creates a data availability sampling challenge; networks like Avail and Celestia optimize for this, but the security floor is lower than Ethereum L1.
Evidence: Ethereum's full 30 blobs per block provides ~0.375 MB/sec. Celestia's mainnet targets 8 MB/block, a 20x+ throughput increase, which directly translates to lower costs for rollups like Arbitrum Orbit and OP Stack chains that integrate it.
protocol-spotlight
DATA AVAILABILITY
Architectural Trade-offs in Practice
Rollup scalability is bottlenecked by the cost and speed of publishing data to a secure base layer. The DA layer choice is the primary architectural decision.
01
Ethereum's Blobs: The Security Anchor
Using Ethereum as the DA layer provides maximal security but introduces a hard cost floor. This is the baseline for L2s like Arbitrum and Optimism.
- Key Benefit: Inherits Ethereum's full consensus and validator set security.
- Key Trade-off: ~$0.01-$0.10 per blob cost, creating a variable but non-zero fee for every transaction batch.
~128KB
Per Blob
~$0.05
Avg. Cost
02
Celestia: The Modular Disruptor
A specialized DA layer that decouples data publishing from execution, offering lower costs by design. Adopted by rollups like Manta and Eclipse.
- Key Benefit: ~$0.0001 per MB data posting, orders of magnitude cheaper than Ethereum blobs.
- Key Trade-off: Introduces a new security assumption, relying on Celestia's validator set instead of Ethereum's.
100-1000x
Cheaper/Blob
New Trust
Assumption
03
EigenDA: The Restaking Security Play
Leverages Ethereum's economic security via restaked ETH from EigenLayer, aiming for a middle ground between cost and security.
- Key Benefit: Security backed by $15B+ in restaked ETH, a stronger cryptoeconomic guarantee than a new PoS chain.
- Key Trade-off: More complex security model; inherits slashing risks from EigenLayer operators, not Ethereum core consensus.
$15B+
TVL Secured
Hybrid
Security Model
04
The Validium Compromise
Solves for ultra-low cost by posting only data availability proofs (or nothing) off-chain, used by applications like Immutable X and dYdX v4.
- Key Benefit: Near-zero on-chain fees; ideal for high-throughput, low-value transactions (e.g., gaming).
- Key Trade-off: Users must trust a Data Availability Committee (DAC) or a Proof-of-Stake chain; introduces liveness assumptions and potential censorship.
~$0.001
Per TX Goal
Off-Chain DA
Trust Assumption
05
Avail: Data Availability as a Primitive
Aims to be a sovereign, general-purpose DA layer with advanced data sampling (via validity proofs) and interoperability features.
- Key Benefit: Enables sovereign rollups that settle their own state, not just execution layers.
- Key Trade-off: Early-stage network; security and adoption are still being proven versus established alternatives.
Validity Proofs
For DA
Sovereign
Rollup Focus
06
The Interoperability Tax
Choosing a non-Ethereum DA layer fragments liquidity and composability. Bridges like LayerZero and Axelar become critical, adding latency and trust layers.
- Key Problem: A rollup on Celestia cannot be natively composed with an app on an Ethereum DA rollup without a bridging step.
- Key Reality: The "modular stack" often trades off unified liquidity for lower costs, creating a new layer of infrastructure complexity.
+2-10s
Bridge Latency
Fragmented
Liquidity
counter-argument
THE DATA DILEMMA
The Validity Proof Counter: "We Need Less Data!"
Validity proofs shift the scalability bottleneck from execution to data availability, making efficient data layers the primary constraint for rollups.
Validity proofs compress execution. A single SNARK or STARK proof verifies millions of transactions, but the sequencer must still publish the transaction data for reconstruction and fraud detection, making data availability the new bottleneck.
The cost is in the calldata. Rollups like Arbitrum and Optimism historically posted all transaction data to Ethereum L1 as expensive calldata. This creates a direct cost correlation where L1 gas fees dictate L2 transaction costs, limiting scalability.
Data Availability (DA) layers are the escape hatch. Solutions like Celestia, EigenDA, and Avail provide specialized, cheaper data publishing. Rollups post data blobs to these layers, and their validity proofs to Ethereum, decoupling security from expensive L1 storage.
Modular architecture wins. This separation creates a stack specialization: execution on the rollup, settlement and proof verification on Ethereum, and data availability on a purpose-built layer. This is the core thesis behind projects like Eclipse and Saga.
Evidence: After EIP-4844 (proto-danksharding), Arbitrum's cost to post data to Ethereum dropped by over 90% when using blobs versus calldata, proving that cheaper DA directly lowers L2 fees.
future-outlook
THE BOTTLENECK
The 2024-2025 Inflection Point
Data availability costs, not execution, become the primary constraint for rollup scalability and economics.
Data is the bottleneck. Rollups compress execution but must post transaction data for verification. The cost of this data availability (DA) on Ethereum L1 dominates transaction fees, limiting throughput and user cost reductions.
DA layers are the new battleground. The competition shifts from execution engines (Arbitrum, Optimism) to DA providers like Celestia, EigenDA, and Avail. Each offers trade-offs between cost, security, and Ethereum alignment.
Modularity fragments liquidity. Rollups using external DA (e.g., a zkRollup on Celestia) create new bridging challenges. Users face a mesh of sovereign chains, increasing reliance on interoperability protocols like LayerZero and Axelar.
Evidence: Starknet's planned transition to a validium using external DA will cut fees by ~99%. This move validates the economic imperative but introduces new security assumptions for users.
takeaways
DATA AVAILABILITY
Architect's Checklist
Rollup scaling is not a compute problem; it's a data publishing problem. The DA layer you choose dictates your chain's security, cost, and finality.
01
The Ethereum Security Trap
Publishing data to Ethereum Mainnet provides maximum security but creates a hard cost ceiling. Each byte of calldata costs ~$0.10-$1.00, making high-throughput applications economically impossible.
- Cost Bottleneck: >90% of rollup transaction fees are DA costs.
- Throughput Limit: Capped at ~0.5-1 MB/s by Ethereum's gas limits.
>90%
Fee Overhead
~1 MB/s
Max Throughput
02
Celestia's Modular Gambit
Separates consensus and execution, offering a sovereign DA layer with data availability sampling (DAS). This enables rollups to scale independently of any execution chain's constraints.
- Cost Arbitrage: DA costs are ~100-1000x cheaper than Ethereum.
- Sovereignty: Rollups control their own governance and upgrade paths.
100-1000x
Cheaper DA
~$0.0001
Per KB Cost
03
EigenDA's Restaking Security
Leverages Ethereum's economic security via restaked ETH, creating a high-throughput DA layer that inherits crypto-economic guarantees. This is the core thesis behind the EigenLayer ecosystem.
- Security Inheritance: Backed by $10B+ in restaked ETH.
- High Bandwidth: Targets 10-100 MB/s data throughput.
$10B+
Securing TVL
10-100 MB/s
Target Throughput
04
Avail's Validity Proof Focus
Builds a DA layer optimized for ZK-rollups, using validity proofs (KZG commitments) and erasure coding to guarantee data is available. This is critical for light client bridges and trust-minimized interoperability.
- ZK-Native: Eliminates fraud proof windows for faster finality.
- Light Client Proofs: Enables secure cross-chain bridges without full nodes.
~2s
Finality Time
Zero Delay
For ZK Proofs
05
The Interoperability Tax
Choosing a non-Ethereum DA layer introduces a bridging tax for assets and messages. You trade lower fees for increased complexity in connecting to the Ethereum DeFi ecosystem via bridges like LayerZero and Axelar.
- Liquidity Fragmentation: Assets are no longer natively on Ethereum L1.
- Security Stack: You now trust the DA layer and the bridge.
2+ Layers
Trust Assumptions
~30-60s
Bridge Latency
06
The Blob Future: EIP-4844 & Danksharding
Ethereum's proto-danksharding (EIP-4844) introduces blob-carrying transactions, a dedicated data space that is ~10-100x cheaper than calldata. This is a direct response to external DA competition.
- Cost Reduction: Immediate ~10-100x DA cost drop for L2s.
- Roadmap Alignment: The path to full danksharding and 16 MB/s+ throughput.
10-100x
Cost Reduction
16 MB/s+
Future Throughput
ENQUIRY
Get In Touch
today.
Our experts will offer a free quote and a 30min call to discuss your project.
NDA Protected
Confidentiality24h Response
Time to ValueDirectly to Engineering Team
No Sales Fluff10+
Protocols Shipped
$20M+
TVL Overall