On-chain data is the bottleneck. ZK-rollups compress execution but must post validity proofs and transaction data to L1. The proof cost is fixed, but the data cost scales linearly with usage. This creates a fundamental subsidy drain.
zk-rollups-the-endgame-for-scaling
Blog
Why On-Chain Data Availability is a Ticking Cost Bomb for ZKRs
Ethereum's calldata and blobs are a temporary, unsustainable cost center. This analysis explains why ZK-rollups must adopt modular DA layers like Celestia and EigenDA to survive long-term scaling.
introduction
THE DATA BOMB
The Hidden Subsidy is Running Out
The unsustainable economics of on-chain data availability are the primary long-term cost threat to ZK-rollup scalability.
Ethereum calldata is not cheap. Storing 1 MB of data on Ethereum currently costs ~0.3 ETH. For a rollup processing millions of transactions, this L1 data fee dominates operational costs and limits sustainable throughput.
The DA subsidy is ending. Rollups like Arbitrum and zkSync have relied on cheap calldata. Post-EIP-4844 and full danksharding, costs drop but remain variable and tied to L1 demand. This eliminates the predictable subsidy model.
Evidence: The Starknet calculus. StarkWare's analysis shows that with high throughput, over 90% of a user's fee pays for L1 data posting, not proof generation or L2 execution. The marginal cost per transaction is a DA fee.
key-trends
WHY ON-CHAIN DA IS A TICKING COST BOMB
The Three Inevitable Cost Pressures
The promise of ZK-Rollups is cheap execution, but their reliance on Ethereum for data availability creates a fundamental and growing cost contradiction.
01
The Blob Fee Volatility Problem
Ethereum's EIP-4844 blobs are a variable fee market, not a fixed cost. As L2 adoption grows, blob demand will spike, making ZK proof verification a rounding error compared to data posting costs. This directly undermines the economic model of high-throughput ZKRs like zkSync, Starknet, and Polygon zkEVM.
- Blob gas is subject to congestion and speculation.
- Base fee can spike 10-100x during network events.
- Creates unpredictable operating costs for sequencers.
10-100x
Fee Spikes
>80%
Cost is Data
02
The State Growth Tax
Every byte of calldata or a blob committed to Ethereum L1 is permanent state growth. The Ethereum community is increasingly resistant to this, viewing it as a subsidy to L2s. Future EIPs will inevitably make this more expensive, directly taxing ZKR scalability.
- Dencun's blob discount is temporary political will.
- Long-term, full nodes bear the cost of L2 data.
- Solutions like EIP-4444 (history expiry) will shift cost burden.
Permanent
State Burden
EIP-4444
Future Pressure
03
The Throughput vs. Cost Deadlock
A ZKR's scalability is gated by its data availability layer. To increase TPS, you must post more data per second to Ethereum, hitting the ~0.375 MB/sec blob throughput limit and paying exponentially higher fees. This creates a hard economic ceiling before any technical limits are reached.
- Throughput scaling increases marginal cost linearly.
- Limits mass adoption for data-heavy apps (GameFi, Social).
- Forces trade-offs between decentralization (on-chain DA) and cost (off-chain DA).
~0.375 MB/s
Blob Cap
Linear
Cost Growth
deep-dive
THE COST CURVE
The Math of the DA Bomb
On-chain data availability costs scale linearly with transaction volume, creating an unsustainable economic model for ZK-rollups.
On-chain DA is a linear cost function. Every transaction's calldata must be posted to Ethereum L1, making ZKR scaling costs directly proportional to usage. This linear scaling contradicts the non-linear scaling benefits promised by ZK-proof compression.
The DA cost dominates the fee structure. For a ZKR like zkSync Era, over 80% of a user's transaction fee pays for L1 data posting, not proof generation or execution. This makes fee reduction asymptotically impossible as activity grows.
EIP-4844 blunts but does not defuse the bomb. Proto-danksharding introduces cheaper blob data, but blob capacity is finite. At mass adoption, blob space becomes the new scarce, auction-based resource, recreating the cost problem.
Evidence: Starknet's transaction cost is ~$0.12, with ~$0.10 covering DA. Without a cheaper DA layer, a 100x increase in TPS multiplies this L1 data cost 100x, erasing any profit margin for the sequencer.
ZK-ROLLUP ECONOMICS
DA Cost Projection: On-Chain vs. Modular
Comparative analysis of data availability (DA) cost structures for ZK-Rollups, highlighting the unsustainable scaling trajectory of on-chain posting versus modular alternatives like Celestia, Avail, and EigenDA.
| Cost & Performance Metric | On-Chain DA (e.g., Ethereum Calldata) | Modular DA (e.g., Celestia, Avail) | Hybrid/Volition Mode (e.g., zkSync, StarkNet) |
|---|---|---|---|
Cost per Byte (Current, Est.) | $0.24 | < $0.001 | $0.24 (L1) / <$0.001 (L2) |
Cost per 100k Txs (Projected at Scale) | $3,000+ | $15 - $75 | User-selectable |
Cost Scaling Trajectory | Tied to L1 gas; Quadratic with adoption | Sub-linear; Decouples from L1 settlement | Bimodal; depends on user/application choice |
Throughput Cap (Data-Only) | ~80 KB/s (Ethereum blob target) | 10-100 MB/s | 80 KB/s (L1) or 10-100 MB/s (L2) |
Censorship Resistance | Weak to Moderate (Probabilistic) | ||
Time to Finality (Data) | ~20 min (Ethereum blob finality) | ~2-10 seconds | ~20 min (L1) or ~2-10 sec (L2) |
Requires Native Token for Security | Varies (false for L1, true for L2) | ||
Architectural Lock-in |
counter-argument
THE COST BOMB
The Ethereum-Maximalist Rebuttal (And Why It's Wrong)
Relying on Ethereum for data availability creates a permanent, non-scalable cost structure that defeats the purpose of ZK-Rollups.
Ethereum's DA is a cost anchor. ZK-Rollups must post compressed transaction data as calldata to Ethereum L1 for security. This cost scales linearly with usage, creating a hard floor for transaction fees that cannot be reduced by ZK-proof efficiency alone.
The modular trade-off is false. The 'modular' argument posits Ethereum for security and execution elsewhere. However, security is not binary; validiums using Celestia or EigenDA offer probabilistic security with 10-100x lower costs, a trade-off most applications accept.
Proofs and data are decoupled. A ZK-proof verifies execution integrity. Data availability ensures state reconstruction. High-cost DA on Ethereum is an architectural choice, not a requirement. Projects like zkSync and Starknet face this cost reality, pushing them toward validium or volition models.
Evidence: Ethereum's average calldata cost is ~$0.10 per 100 bytes. A simple ETH transfer in a ZKR costs ~$0.03-0.05 just for DA, a permanent tax that makes microtransactions and high-frequency DeFi on Arbitrum or Optimism economically impossible at scale.
protocol-spotlight
COST STRUCTURE ANALYSIS
The Modular DA Escape Hatch
On-chain data availability is the single largest and most volatile cost center for ZK rollups, threatening long-term viability.
01
The L1 DA Tax: A 90% Cost Burden
For a ZK-rollup, posting data to Ethereum L1 consumes 90%+ of total transaction costs. This is a fixed, non-negotiable tax that scales with L1 gas prices, not rollup efficiency.
- Cost Driver: Every byte of calldata is priced at volatile L1 gas rates.
- Scalability Ceiling: Throughput is capped by L1's data bandwidth, creating a ~80 KB/s bottleneck.
90%+
Of Tx Cost
80 KB/s
L1 Bottleneck
02
EigenDA: The Cost Arbitrage Play
EigenDA uses a network of cryptoeconomically secured nodes to provide data availability at a ~99% discount versus Ethereum calldata. It's the pragmatic, production-ready escape hatch.
- Mechanism: Data blobs are dispersed and attested to by restaked ETH validators.
- Adoption Signal: Major rollups like Arbitrum, Optimism, and Manta are already integrating.
-99%
Vs. L1 Calldata
10+
Rollups Integrating
03
Celestia: The Sovereignty Argument
Celestia decouples execution from consensus and DA, offering rollups a dedicated data layer. This enables sovereign rollups that control their own governance and upgrade paths.
- Key Trade-off: Introduces a new security and liveliness assumption outside of Ethereum.
- Market Fit: Ideal for app-chains and ecosystems prioritizing maximal independence over Ethereum alignment.
Sovereign
Chain Control
Dedicated
DA Layer
04
The Avail & Near DA Trilemma
New entrants like Avail (Polygon) and Near DA compete on a trilemma of cost, decentralization, and proof speed. They use validity proofs and novel consensus to challenge incumbents.
- Avail: Focuses on light client verifiability and a robust peer-to-peer network.
- Near DA: Leverages Nightshade sharding for high throughput, positioning as a high-performance alternative.
Trilemma
Optimization
Sharded
Architecture
05
The Blob Market: A Race to Zero
EIP-4844 (blobs) was a temporary relief valve. The emerging modular DA market will create a commoditized, competitive landscape where price is the primary differentiator.
- Endgame: DA becomes a low-margin utility, with rollups dynamically routing to the cheapest secure provider.
- Risk: Over-optimization for cost can lead to fragmented security and increased systemic complexity.
Commoditized
Market
Dynamic
Routing
06
The Strategic Imperative: DA Abstraction
Forward-thinking ZK stacks like zkSync, Starknet, and Polygon zkEVM are building DA abstraction layers. This allows applications to choose their DA provider without protocol-level changes.
- Developer UX: Unlocks "DA-as-a-Service" where cost and security profiles are selectable parameters.
- Future-Proofing: Insulates dApps from the coming DA wars and ensures cost predictability.
Abstracted
Layer
Selectable
Security/Cost
takeaways
THE DA COST BOMB
TL;DR for Protocol Architects
On-chain data availability is the single largest and most volatile cost center for ZK rollups, threatening long-term viability.
01
The Problem: Ethereum as a $1M/Month Ledger
Publishing ZK proof data to Ethereum L1 is a raw, linear cost. For a chain with ~100 TPS, this can exceed $1M/month in pure calldata fees. This scales directly with usage, creating a perverse incentive to limit adoption.
$1M+
Monthly Cost
Linear
Cost Scaling
02
The Solution: Off-Chain DA Layers (Celestia, EigenDA, Avail)
Decouple data publication from settlement. Dedicated DA layers like Celestia and EigenDA offer ~100x cheaper data blobs. This transforms the cost structure from variable OpEx to fixed, predictable infrastructure spend.
- Key Benefit 1: Sub-cent transaction costs become feasible.
- Key Benefit 2: Enables sustainable hyper-scalability (>10k TPS).
~100x
Cheaper DA
Fixed Cost
Pricing Model
03
The Trade-off: The Security <> Cost Frontier
Moving data off Ethereum introduces a new security assumption: the liveness of the external DA layer. This creates a spectrum from Ethereum (max security, max cost) to Validium (high throughput, minimal cost). The architect's choice defines the chain's trust model.
- Key Trade-off: You are trading Ethereum's consensus security for economic scalability.
Spectrum
Security Model
Validium
Max Scale
04
The Architecture: Modular vs. Monolithic Stacks
This cost bomb forces a fundamental architectural decision. Monolithic chains (Solana) bundle execution, settlement, consensus, and DA. Modular stacks (using Rollkit, Eclipse, Polygon CDK) let you mix-and-match best-in-class components for each function, optimizing for your specific cost/security profile.
Modular
Winning Stack
Mix & Match
Components
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