Execution is not the bottleneck. Rollups execute transactions off-chain, but their finality depends on posting data to a base layer like Ethereum. This data availability (DA) layer determines the maximum sustainable throughput.
the-ethereum-roadmap-merge-surge-verge
Blog
Why Data Availability Controls Rollup Throughput
The promise of rollups is infinite scaling. The reality is a hard cap set by data availability. This is the fundamental bottleneck for transaction throughput, and the entire Surge roadmap is about breaking it.
introduction
THE DATA BOTTLENECK
The Scaling Lie: Throughput Isn't Free
Rollup scalability is ultimately constrained by the cost and speed of publishing transaction data to a base layer.
Throughput is a function of cost. Each byte of calldata posted to Ethereum L1 costs gas. Rollups like Arbitrum and Optimism must batch transactions to amortize this cost, creating a direct link between L1 gas prices and L2 user fees.
Cheaper DA enables higher TPS. Solutions like EigenDA, Celestia, and Avail offer lower-cost data availability, allowing rollups to post more data per second. This is the core scaling mechanism for zkSync and Starknet on these networks.
Evidence: Ethereum's current calldata capacity caps a single rollup at ~100-200 TPS. A dedicated DA chain like Celestia targets 10,000+ TPS, shifting the bottleneck from consensus to execution.
key-trends
THROUGHPUT VS. SECURITY
The DA Bottleneck in Practice
Data Availability is the primary constraint on rollup scalability, forcing a direct trade-off between transaction speed and network security.
01
The Problem: Ethereum's 80 KB/s Ceiling
Ethereum's consensus layer can only process ~80 KB/s of calldata. This hard cap limits all L2s to a collective throughput of ~100-200 TPS, regardless of their execution speed.\n- Bottleneck: Execution is parallelizable, but DA is serialized on L1.\n- Consequence: Blockspace auctions during congestion make L2 fees volatile and expensive.
80 KB/s
DA Bandwidth
~150 TPS
Theoretical Max
02
The Solution: Modular DA Layers (Celestia, EigenDA)
Offloading data to specialized, high-throughput networks decouples DA from execution. This bypasses Ethereum's bottleneck, enabling 10-100x higher throughput for rollups like Arbitrum Orbit and Optimism Stack chains.\n- Trade-off: Security shifts from Ethereum's validator set to the new DA layer's.\n- Economics: Reduces L2 fees by >90% by avoiding L1 calldata costs.
10-100x
Throughput Gain
>90%
Cost Reduction
03
The Risk: Data Availability Sampling (DAS) Assumptions
Networks like Celestia and Avail use DAS to allow light nodes to verify data availability with minimal downloads. This scales bandwidth but introduces new security assumptions.\n- Requirement: Relies on a sufficiently decentralized and honest sampling peer set.\n- Failure Mode: A sophisticated adversary could hide data from samplers, leading to an invalid state being finalized.
~1 MB/s
DA Scale (Celestia)
Light Nodes
Verification Model
04
The Hybrid: Ethereum with Proto-Danksharding (EIP-4844)
EIP-4844 introduces blob-carrying transactions, a dedicated data channel for rollups. It increases Ethereum's effective DA bandwidth to ~1.7 MB/s while preserving L1 security.\n- Mechanism: Blobs are cheap, large (~128 KB), and automatically deleted after ~18 days.\n- Impact: Expected to reduce L2 fees by 10-100x versus current calldata, a major step-change before full Danksharding.
~1.7 MB/s
Post-4844 Bandwidth
10-100x
Fee Reduction
05
The Trade-Off: Validium vs. Optimistic Rollup
This is the core architectural choice: keep data on-chain (Optimistic/ZK Rollup) or off-chain (Validium). Validium (e.g., StarkEx, zkPorter) offers ~9,000 TPS but inherits the security of its off-chain DA committee.\n- Security: Optimistic Rollups inherit Ethereum's DA security fully.\n- Throughput: Validium sacrifices this for orders-of-magnitude higher scale and lower cost.
~9,000 TPS
Validium Scale
Ethereum
Optimistic DA
06
The Future: Volition & The DA Marketplace
Systems like StarkNet and Aztec allow users to choose per-transaction where data is posted—on L1 for high value, or off-chain for low cost. This creates a competitive DA marketplace between Ethereum, Celestia, EigenDA, and others.\n- User Sovereignty: Security becomes a configurable parameter.\n- Market Dynamics: DA costs will be driven by supply, demand, and perceived security premiums.
Per-Tx Choice
Volition Model
Multi-Layer
DA Competition
deep-dive
THE DATA PIPELINE
Anatomy of a Bottleneck: From Blobs to Finality
Rollup throughput is ultimately constrained by the data availability layer's capacity to post and confirm transaction data.
Data availability is the primary bottleneck. A rollup's execution speed is irrelevant if its transaction data cannot be posted to the base layer. The Ethereum mainnet's blob-carrying capacity sets the hard ceiling for all L2s combined.
Finality is a multi-layered concept. Rollups provide fast, soft confirmations, but users wait for L1 finality for true settlement. This delay, managed by bridges like Across and Stargate, is the security-for-speed tradeoff.
Blob fee markets create congestion. When blob space is scarce, rollups like Arbitrum and Optimism compete, spiking costs and forcing transaction batching, which increases latency for end-users.
Evidence: Post-Dencun, the 3-blob target provides ~0.375 MB per block. This shared resource must be divided among all rollups, capping aggregate throughput before execution is even considered.
THROUGHPUT BOTTLENECKS
DA Layer Capacity & Rollup Throughput Limits
Comparison of data availability (DA) layers by their core capacity metrics, which directly constrain rollup transaction throughput and cost.
| Metric / Feature | Ethereum (Calldata) | Celestia | EigenDA | Avail |
|---|---|---|---|---|
Peak Data Bandwidth (MB/s) | ~0.06 MB/s | ~14 MB/s | ~10 MB/s | ~7 MB/s |
Blob Cost per MB (Current) | $100 - $500 | < $0.01 | < $0.001 | < $0.01 |
Data Availability Sampling (DAS) | ||||
Proof System for Data Availability | None (Full Nodes) | Celestia's Light Clients | EigenLayer + Restaking | Validity Proofs (ZK) |
Throughput Limit for a Single Rollup | ~80 TPS | ~14,000 TPS | ~10,000 TPS | ~7,000 TPS |
Time to Finality for Data | ~12-15 min (Ethereum Conf.) | ~2-4 sec (Block Time) | < 1 sec | ~20 sec |
Primary Security Model | Ethereum Consensus | Sovereign Cosmos Chain | Ethereum Restaking Pool | Polkadot-Style Nominated PoS |
future-outlook
THE BOTTLENECK
The Surge is a DA Surge
Rollup throughput is fundamentally constrained by the cost and speed of publishing data to a base layer.
Data availability is the bottleneck. A rollup's execution speed is irrelevant if its state transitions cannot be posted cheaply and quickly to a secure base layer like Ethereum. This creates a direct link between DA costs and user fees.
Blobs are the scaling primitive. EIP-4844 introduced blob-carrying transactions, a dedicated data lane with separate fee markets. This decouples L2 data posting from mainnet execution congestion, directly lowering rollup transaction costs.
Throughput is a DA function. The theoretical maximum transactions per second (TPS) for a rollup is a function of blob space per block and data compression efficiency. Projects like Arbitrum and zkSync compete on data compression ratios to maximize this limit.
Evidence: Post-EIP-4844, average L2 transaction fees dropped by over 90%. The next step, full danksharding, will increase blob capacity from ~0.375 MB to ~16 MB per slot, enabling massive throughput scaling for rollups.
takeaways
DATA AVAILABILITY IS THE BOTTLENECK
TL;DR for Builders
Rollup throughput is not limited by execution; it's gated by the speed and cost of posting data to a base layer.
01
The Problem: On-Chain DA is a Cost Wall
Posting transaction data to Ethereum L1 is the single largest cost for rollups like Arbitrum and Optimism, consuming 60-90% of their total fees. This creates a direct, linear relationship between L1 gas prices and your user's transaction costs.\n- Cost Scaling: Every 100 gwei on L1 can increase rollup fees by $0.50+ per transaction.\n- Throughput Ceiling: Limited by Ethereum's ~80 KB/s block space for data blobs.
60-90%
Of Rollup Fees
~80 KB/s
Ethereum DA Cap
02
The Solution: Modular DA Layers (Celestia, Avail)
Specialized data availability layers decouple data publishing from Ethereum's execution, offering order-of-magnitude cheaper and higher-throughput data posting. This is the core innovation enabling sovereign rollups and high-throughput L2s.\n- Cost Arbitrage: DA on Celestia can be >100x cheaper than calldata on Ethereum mainnet.\n- Throughput Unlocked: Scales to MB/s of data availability, removing the primary bottleneck.
>100x
Cheaper DA
MB/s
Throughput Scale
03
The Trade-Off: Security & Interoperability
Leaving Ethereum's DA introduces new risks. You're trading Ethereum's maximal security for a lighter cryptoeconomic security model. This fragments liquidity and complicates cross-rollup messaging (e.g., bridging to/from LayerZero, Across).\n- Security Spectrum: From Ethereum's ~$100B+ stake to external DA's ~$1-5B cryptoeconomic security.\n- Interop Friction: Native bridges from an EigenDA rollup to an Arbitrum chain are not trust-minimized.
$100B+ vs $1-5B
Security Stake
High
Interop Friction
04
The Architect's Choice: EigenDA & Restaking
EigenDA leverages Ethereum's restaking ecosystem via EigenLayer to provide a hybrid model: higher throughput than mainnet DA with security backed by re-staked ETH. This appeals to rollups wanting to stay within Ethereum's security orbit.\n- Security Source: Taps into the pooled security of $15B+ in restaked ETH.\n- Target Market: Ethereum-aligned L2s (e.g., upcoming L2s from major ecosystems) prioritizing security over lowest cost.
$15B+
Restaked ETH
Ethereum-Aligned
Security Model
05
The Metric: Cost per Byte & Finality Time
When evaluating DA solutions, benchmark two hard metrics that directly impact your rollup's performance and user experience. These are your non-negotiable KPIs.\n- Cost per Byte: Drives your minimum viable transaction fee. Target < 0.1 gas/byte (vs Ethereum's ~16 gas/byte).\n- Data Finality Time: Defines your withdrawal delay. Ranges from Ethereum's 12 minutes to external DA's ~2 seconds.
< 0.1 gas/byte
Target Cost
12min vs 2sec
Finality Range
06
The Future: DA Sampling & Volitions
The endgame is data availability sampling (DAS), where light clients can verify data availability without downloading everything. This enables secure, scalable rollups for resource-constrained environments. Combined with volitions (user-choice between on-chain/off-chain DA), this creates ultimate flexibility.\n- DAS Enablers: Celestia, Avail, and Ethereum's own EIP-4844 proto-danksharding roadmap.\n- User Sovereignty: Volitions let users choose security/cost trade-offs per transaction.
Light Clients
DAS Target
User-Choice
Volition Model
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