Data availability (DA) is the new bottleneck. High-throughput L2s like Arbitrum and Optimism generate more data than the Ethereum base layer can store economically, creating a fundamental scaling limit.
the-modular-blockchain-thesis-explained
Blog
The Future of Data Availability: Beyond On-Chain Storage
Data Availability layers are not just cheap storage. They are becoming verifiable data publishing networks, fundamentally separating proof of publication from historical data persistence. This evolution is the core of the modular stack.
introduction
THE DATA AVAILABILITY FRONTIER
Introduction
The core scaling bottleneck for blockchains has shifted from execution to ensuring data is provably available for verification.
On-chain storage is a cost trap. Storing all transaction data permanently on a monolithic chain like Ethereum imposes unsustainable costs, which rollups like zkSync and StarkNet directly pass to end-users.
The future is modular separation. Dedicated DA layers, including Celestia, EigenDA, and Avail, decouple data publishing from consensus and execution, enabling specialized, cost-optimized scaling.
Evidence: Celestia's mainnet publishes data for ~$0.01 per MB, orders of magnitude cheaper than equivalent Ethereum calldata, proving the economic imperative for dedicated DA.
key-trends
BEYOND ON-CHAIN STORAGE
Executive Summary: The DA Shift
Data Availability is the new bottleneck. The future is a modular stack separating execution from secure, verifiable data publishing.
01
The Problem: Full Nodes Are a Scaling Dead End
Requiring every node to store all data creates a centralizing force. The Ethereum full node count has stagnated at ~6,000 while state size grows exponentially. This is the core constraint for monolithic L1s and expensive rollups.
~6K
Ethereum Nodes
10TB+
State Growth/Yr
02
The Solution: Data Availability Sampling (DAS)
Light clients can probabilistically verify data availability by sampling small, random chunks. This enables secure scaling without trusting a central operator. It's the cryptographic primitive powering Celestia, EigenDA, and Ethereum's own Proto-Danksharding.
- Key Benefit: Security scales with node count, not data size.
- Key Benefit: Enables ~1,000x throughput for rollups.
1,000x
Throughput Gain
<$0.001
Target Cost/Tx
03
The Trade-Off: Security vs. Cost Spectrum
DA is not binary. Projects choose a point on the security-cost continuum.
- Ethereum Consensus (High Sec): ~$1.2K/TB, maximal security.
- EigenDA (Modular Sec): ~$50/TB, Ethereum staking economic security.
- Celestia (Sovereign Sec): ~$1/TB, independent consensus & light client security.
$1.2K vs $1
Cost/TB Range
3 Models
Security Tiers
04
The Next Layer: Proving & Dispersal
DA isn't just storage; it's about proving data was published. This is shifting from simple Merkle roots to KZG commitments (EIP-4844) and zk-proofs of DA. Avail and Near DA use validity proofs to compress verification. Dispersal networks like EigenDA leverage restaked ETH for cryptoeconomic security.
KZG / zkProofs
Proof Tech
$16B+
Restaked Sec
05
The Endgame: DA as a Commodity
Like cloud storage, DA will become a low-margin, high-reliability utility. Competition will be on latency, geodistribution, and integration. The winning stack will be the one that makes DA invisible to developers, abstracted away by rollup frameworks like Rollkit, Optimism Stack, and Arbitrum Orbit.
<2s
Target Finality
~0
Dev Overhead
06
The Risk: Fragmentation & Interop Hell
Multiple DA layers (Celestia, EigenDA, Avail, Near) create data silos. Cross-rollup composability breaks if they use different DA. Solutions like LayerZero V2 (Omnichain), Polymer, and shared settlement layers (Espresso, Fuel) are emerging to rewire liquidity across DA boundaries.
4+
Major DA Layers
High
Fragmentation Risk
thesis-statement
THE DATA
The Core Thesis: Proof, Not Storage
The future of data availability is defined by cryptographic proof of data existence, not by the physical location of the data itself.
Data availability is a verification problem. The core requirement for a rollup is to prove that transaction data was published and accessible, not to store it forever on a monolithic chain like Ethereum. This distinction separates the consensus layer from the archival layer.
Validity proofs shift the security model. With systems like zkSync Era and Starknet, the L1 only needs the validity proof and a tiny data commitment. The security guarantee becomes cryptographic, not economic, reducing the need for persistent, full on-chain data blobs.
EIP-4844 proto-danksharding is a transitional tool. It provides cheap blob storage as a temporary bandwidth subsidy, but the endgame is data availability sampling (DAS). Networks like Celestia and EigenDA architect for this from first principles, using light nodes to probabilistically verify data availability without downloading it.
Evidence: Ethereum's full historical state is ~15TB and growing linearly. A Celestia light node verifies data availability for 100MB blocks using just ~1KB of data, demonstrating the asymptotic efficiency of proof-based systems.
BEYOND ON-CHAIN STORAGE
DA Layer Architecture Comparison
A technical breakdown of leading data availability solutions, contrasting their core architectures, performance, and economic models for CTOs and architects.
| Feature / Metric | Monolithic L1 (e.g., Ethereum) | Modular DA (e.g., Celestia, EigenDA) | Validium (e.g., StarkEx, zkPorter) | Hybrid / Emerging (e.g., Avail, Near DA) |
|---|---|---|---|---|
Core Architecture | Execution + Consensus + DA on one chain | Specialized DA chain with light clients | DA off-chain with on-chain proofs | DA with embedded validity proofs or sharding |
Data Availability Guarantee | 100% (Full L1 Security) | Cryptoeconomic (Fraud/Validity Proofs) | Committee/Guardian-based | Cryptoeconomic + Light Client Bridges |
Throughput (MB/s) | ~0.06 MB/s (1.5 MB/block) | 10-100 MB/s | 100-1000+ MB/s | 1-10 MB/s (scaling with shards) |
Cost per MB (Est.) | $1000 - $5000 | $0.10 - $1.50 | < $0.01 | $0.50 - $5.00 |
Settlement Latency | 12-15 seconds (Ethereum) | ~2 seconds (optimistic) to ~15s (finalized) | Instant (off-chain) to ~30m (ZK proof gen) | 2-5 seconds |
Interoperability Model | Native (within L1) | Light Client Bridges (IBC, Ethereum bridge) | Proprietary Bridge to L1 | Universal Light Clients (e.g., KZG + zk) |
Data Pruning / Historical Access | Full nodes required | Data Availability Sampling (DAS) enables light nodes | Relies on Data Availability Committee | DAS + Archival networks (e.g., BitTorrent) |
Ethereum Security Inheritance |
deep-dive
THE POST-BLOCK PARADIGM
The Mechanics of Verifiable Publishing
Data availability is shifting from monolithic block storage to a modular system of cryptographic proofs and decentralized sampling.
Verifiable publishing decouples consensus from storage. Layer 2s like Arbitrum and Optimism no longer need to post full transaction data to Ethereum. Instead, they commit a cryptographic fingerprint and make raw data available elsewhere, slashing costs by 90%+.
The core innovation is data availability sampling (DAS). Clients like those on Celestia or EigenDA download random chunks of data to probabilistically guarantee its publication. This scales throughput linearly with the number of light nodes, unlike monolithic chains.
Proof systems like KZG and validity proofs are the final guarantee. A KZG commitment proves a blob's existence, while a validity proof (e.g., from a zkVM) confirms correct state execution. This creates a trust-minimized data pipeline from sequencer to settlement layer.
The endgame is a competitive DA marketplace. Rollups will choose between cost-optimized options like Celestia, security-maximized Ethereum blobs via EIP-4844, and high-throughput alternatives like Avail. This commoditizes a core blockchain function.
risk-analysis
BEYOND ON-CHAIN STORAGE
The Bear Case: Risks of Decarding
Decoupling execution from data availability introduces new, systemic risks that threaten the security assumptions of the entire modular stack.
01
The Data Unavailability Attack
A sequencer can withhold data, preventing fraud proofs and allowing invalid state transitions to finalize. This is the core liveness failure mode of modular designs.
- Ethereum's Danksharding mitigates this with data availability sampling (DAS).
- Celestia and Avail use 2D Reed-Solomon encoding for similar guarantees.
- Without robust sampling, the system reverts to honest-majority-of-validators security.
1-of-N
Honest Node Required
~10 min
Challenge Window
02
The Interoperability Fragmentation Trap
Multiple competing DA layers (EigenDA, Celestia, Avail) create liquidity and state fragmentation, breaking composability.
- Bridges and cross-chain apps like LayerZero and Axelar must now secure multiple light client connections.
- This increases the attack surface and complexity for protocols like Uniswap or Aave operating across rollups.
- The winner-take-most economics could recentralize power around a single DA provider.
5-10x
More Light Clients
$B+
Bridged Value at Risk
03
The Long-Term Recentralization of Sequencers
DA is a commodity; the real power and profit accrues to the sequencer/block producer who orders transactions.
- Projects like Espresso Systems and Astria are building shared sequencer networks to combat this.
- Without enforceable decentralization, rollups become glorified sidechains with centralized operators.
- The modular stack risks recreating the validator centralization problems of L1s.
>50%
Sequencer Profit Margin
~3
Major Providers
04
The Cost Illusion & Subsidy Cliff
Today's low DA costs are driven by token subsidies and unproven demand. At scale, marginal costs converge.
- EigenDA leverages restaked ETH, creating circular security dependencies with EigenLayer.
- If usage spikes, costs could approach those of Ethereum calldata, negating the primary rollup value proposition.
- This creates a fragile economic model for high-throughput apps like dYdX or Sorare.
100-1000x
Potential Cost Increase
$0
Current Token Cost
future-outlook
BEYOND STORAGE
Future Outlook: The DA Stack in 2025
Data Availability will evolve from a simple storage layer into a competitive market of specialized services and integrated execution.
DA becomes a commodity market. The core function of publishing and attesting to data will be a low-margin, high-volume service. Competition between Celestia, EigenDA, and Avail will drive costs toward the marginal price of bandwidth and storage, decoupling DA from execution.
Specialized DA services emerge. Generic blob storage is insufficient. We will see services optimized for high-frequency state diffs (for gaming), privacy-preserving proofs (using zk-tech), and real-world data oracles, creating vertical-specific DA solutions.
The execution-DA boundary blurs. Protocols like Fuel and Sovereign rollups will use DA layers not just for data but for verifiable state transitions. The DA layer becomes the settlement and consensus backbone, with execution clients built directly on top.
Evidence: EigenDA's throughput target of 10 MB/s and sub-$0.10 per MB cost sets the benchmark for 2025's commodity pricing, forcing all competitors to match or specialize.
takeaways
THE FUTURE OF DATA AVAILABILITY
TL;DR for Builders
On-chain storage is a bottleneck. The next wave of scaling demands new architectures for data availability that separate security from execution.
01
The Problem: Ethereum's Blob-Carrying Capacity is Finite
Ethereum's ~0.375 MB per slot blob capacity is already a contested resource. As L2 adoption grows, this creates a predictable fee market and congestion for data.\n- Blob fees are volatile, making L2 cost predictions unreliable.\n- Throughput is capped, limiting the total number of cheap L2 transactions.
~0.375 MB
Per Slot Cap
High Volatility
Fee Risk
02
The Solution: Modular DA Layers (Celestia, EigenDA, Avail)
Specialized data availability layers decouple DA from execution, offering orders of magnitude more bandwidth at a fraction of Ethereum's cost.\n- Costs slashed by 100-1000x vs. Ethereum calldata.\n- Enables sovereign rollups with their own governance and fork choice.\n- Uses data availability sampling (DAS) for lightweight, scalable verification.
100-1000x
Cheaper
MB/s
Throughput
03
The Trade-off: Security vs. Sovereignty Spectrum
Not all DA is equal. Builders must choose a point on the spectrum between Ethereum's maximal security and modular chains' cost efficiency.\n- Ethereum DA: Maximum security, highest cost, integrated settlement.\n- EigenDA (Restaked Security): Ethereum security pool via EigenLayer, medium cost.\n- External DA (Celestia): Independent security, lowest cost, requires a separate settlement layer.
Ethereum
Max Security
Celestia
Max Sovereignty
04
The Next Frontier: Volitions and Hybrid DA
Applications won't choose one DA source. Volitions (from StarkWare) and hybrid models let users select their preferred DA for each transaction.\n- High-value tx: Can opt for secure, on-chain Ethereum DA.\n- Low-cost tx: Can use cheap, external modular DA.\n- This creates a dynamic, user-centric cost structure within a single app.
User Choice
Per TX
Hybrid
Architecture
05
The Risk: Fragmented Security and Bridging Complexity
Modularity introduces new attack vectors. The security of an L2 is now the weakest link in its DA, settlement, and execution stack.\n- Data withholding attacks become possible if DA layer security is weak.\n- Cross-chain messaging (e.g., via LayerZero, Hyperlane) between different DA environments adds complexity and trust assumptions.
Weakest Link
Security Model
New Vectors
Attack Surface
06
The Builder's Playbook: How to Choose
Your DA choice dictates your stack, costs, and security model. Ask: What are you optimizing for?\n- Optimize for Security/Ecosystem: Use Ethereum DA (Blobs) or EigenDA.\n- Optimize for Cost/Throughput: Use a modular DA layer like Celestia or Avail.\n- Optimize for Flexibility: Build a Volition or support multiple DA backends.
Security
vs. Cost
Modular Stack
Required
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