ZK-Rollups are execution shards. They compress computation but remain bottlenecked by the underlying chain's data capacity, as seen with Ethereum's 80 KB/sec blob limit constraining Arbitrum and zkSync.
zk-rollups-the-endgame-for-scaling
Blog
Why Data Sharding Is the True Endgame, Not ZK-Rollups
Rollups are hitting a data availability wall. This analysis argues that scalable data layers like Danksharding solve the root bottleneck, making them the true architectural endgame for blockchain scalability.
introduction
THE DATA
Introduction
Scalability's final frontier is data availability, not execution, making data sharding the essential substrate for all L2s.
Data sharding is the base layer. It provides the raw bandwidth for rollups to post proofs and state diffs, a function Ethereum's Danksharding and Celestia are explicitly designed to fulfill.
The scaling hierarchy is fixed. Execution layers (rollups) scale vertically, but horizontal scaling requires an unbundled data availability layer, making sharding the true systemic bottleneck.
Evidence: A single EIP-4844 blob carries ~125 KB, enough for ~1000 simple transfers, but scaling to millions requires the parallel lanes of a sharded data network.
thesis-statement
THE DATA BOTTLENECK
The Core Argument: Data Bandwidth is the Root Constraint
The fundamental limit for global blockchain scaling is the cost and speed of publishing data to a secure base layer.
Rollups hit a data wall. ZK-Rollups like zkSync and StarkNet compress execution but still publish state diffs to Ethereum. The cost of this data availability (DA) on L1 is the primary fee component for users, creating a hard economic ceiling.
Sharding is the only solution. Execution sharding, as originally proposed for Ethereum, creates complexity without solving the core issue. Data sharding (e.g., Ethereum's Danksharding, Celestia) scales the base data layer itself, providing exponentially cheaper bandwidth for all rollups to consume.
The DA layer is the new battleground. Projects like EigenDA and Avail are competing to become the cheapest, most reliable data marketplace. Rollups will become execution clients that bid for space on these decentralized data networks.
Evidence: Ethereum's full blocks today cost a rollup ~$0.10 per transaction in pure L1 data fees. A mature data sharding layer reduces this cost by 100-1000x, making sub-cent L2 transactions economically viable.
key-trends
SHARDING VS. COMPRESSION
The Data Availability Arms Race
ZK-Rollups are a temporary fix; the true scaling endgame is scaling the DA layer itself through sharding.
01
The Problem: Exponential Blob Demand
Every new rollup (Arbitrum, zkSync, Starknet) consumes more L1 block space. Ethereum's ~1.7 MB/block is a hard cap, creating a zero-sum game for DA.\n- Costs skyrocket during congestion, making L2s unusable.\n- This bottleneck forces rollups into centralized sequencers or off-chain data solutions, compromising security.
1.7 MB
Block Limit
10x+
Cost Spike
02
The Solution: Proto-Danksharding (EIP-4844)
A transitional DA scaling via blob-carrying transactions. It separates rollup data from execution gas markets.\n- Provides ~0.5 MB/slot of dedicated, cheap blob space.\n- Enables ~100x cost reduction for L2s by using ephemeral data storage.\n- The bridge to full Danksharding, which scales to ~1.3 MB/s per shard.
~100x
Cheaper L2 Tx
0.5 MB/slot
Blob Capacity
03
The Endgame: Danksharding & Data Availability Sampling
True horizontal scaling via 64 data shards with light-client verification. This is the only way to achieve 100k+ TPS for rollups.\n- Uses KZG commitments and Data Availability Sampling (DAS) for trust-minimized scaling.\n- Enables full decentralization of rollup sequencers by guaranteeing cheap, abundant DA.\n- Renders alternative DA layers (Celestia, EigenDA) redundant for Ethereum's security domain.
64
Data Shards
100k+
Max TPS
04
Why Modular DA Layers Are a Side-Chain
External DA layers (Celestia, Avail, EigenDA) fragment security and create new trust assumptions. They are not scaling Ethereum, but creating parallel systems.\n- Introduces sovereign bridge risk between execution and DA layers.\n- Ethereum's consensus is the most valuable security asset; sidestepping it is a trade-off.\n- Their utility is as a testnet for sharding tech, not the final architecture.
New Risk
Bridge Security
$1B+
TVL at Risk
05
The ZK-Rollup Illusion
ZKPs compress execution, not data. A ZK-rollup posting 500kb proofs still needs to publish 5MB of input data for fraud proofs and state reconstruction.\n- Validity proofs don't solve data bandwidth.\n- Without scalable DA, ZK-rollups hit the same cost wall as Optimistic rollups.\n- The real race is between Celestia's early mover advantage and Ethereum's slower but secure sharding path.
5 MB
Data per Batch
0 MB
Saved by ZK
06
The Final Metric: Cost per Byte Secured
The ultimate benchmark for DA layers. Ethereum sharding aims to drive this cost asymptotically to zero while maintaining crypto-economic security.\n- Ethereum's security budget (~$30B staked) is unmatchable by new DA layers.\n- The winner isn't who scales first, but who achieves the lowest cost/byte with the highest security.\n- This metric will consolidate rollups onto 1-2 canonical DA layers long-term.
$30B
Security Budget
~$0
Target Cost/Byte
DATA AVAILABILITY IS THE TRUE SCALING CONSTRAINT
The Rollup Bottleneck: Data Costs vs. Execution Costs
Compares scaling strategies by their fundamental constraints, showing why data sharding is the only path to sustainable, low-cost scaling for all rollups.
| Core Constraint | ZK-Rollups (e.g., zkSync, StarkNet) | Optimistic Rollups (e.g., Arbitrum, Optimism) | Data Sharding (e.g., Ethereum Danksharding, Celestia) |
|---|---|---|---|
Primary Bottleneck | Prover Computation | Fraud Proof Window (7 days) | Bandwidth & Storage |
Cost to Post 100KB of Data to L1 | $5-15 (Ethereum calldata) | $5-15 (Ethereum calldata) | < $0.01 (Sharded blob) |
Scalability Limit | L1 Data Bandwidth | L1 Data Bandwidth | Global Node Bandwidth |
Inherent Trust Assumption | Cryptographic (ZK Validity) | Economic (Bonded Challengers) | Data Availability Sampling (1-of-N Honest Node) |
Supports Sovereign Rollups | |||
Cross-Rollup Synergy | None (Per-Rollup Proving Cost) | None (Per-Rollup Fraud Proofs) | Full (Shared Data Layer) |
Time to Finality (L1 Inclusion) | ~10-30 minutes | ~1 week (Challenge Period) | ~1-10 minutes (Blob Confirmation) |
Endgame Scaling Target (TPS) | ~10,000 (Bottlenecked by L1) | ~10,000 (Bottlenecked by L1) | ~1,000,000+ (Theoretical Blob Limit) |
deep-dive
THE DATA LAYER
Why Danksharding and Modular DA Win
Data availability is the fundamental bottleneck, and sharded DA layers like Danksharding provide the only path to global-scale throughput.
ZK-Rollups are not scaling. They compress execution but still post data to a monolithic chain. This creates a data availability bottleneck that limits all L2s. The final throughput of a rollup-centric ecosystem equals the DA capacity of its L1.
Danksharding decouples data from execution. It transforms Ethereum into a high-throughput data availability layer, providing cheap blobs for thousands of parallel chains. This is the modular thesis in practice, enabling specialized execution layers like Arbitrum and Optimism to scale independently.
Monolithic chains hit physical limits. Solana and Sui optimize for single-threaded performance, but face hardware and bandwidth constraints. A sharded DA layer, as pioneered by Celestia and implemented by EigenDA, scales throughput linearly by adding more nodes.
The endgame is sharded data. Execution will fragment into app-chains and rollups via ecosystems like Polygon CDK and Arbitrum Orbit. Their unified base layer must be a hyper-scalable DA platform, which is the core innovation of Danksharding's proto-danksharding (EIP-4844) rollout.
counter-argument
THE BANDWIDTH BOTTLENECK
Steelman: "But ZK-Rollups Are Sufficient"
ZK-Rollups are a scaling solution, but they do not solve the fundamental data availability problem that limits all L2s.
ZK-Rollups are not sovereign. They are data clients to an L1. Their security and finality are contingent on the L1's ability to post and verify their data. This creates a hard dependency on the L1's data bandwidth, which is the ultimate bottleneck.
Scaling is a data problem. A rollup's throughput is capped by the data capacity of its parent chain. Even with perfect ZK-proof compression, a congested L1 like Ethereum cannot post the data for thousands of high-throughput rollups. The data availability layer is the constraint.
Compare Celestia to Ethereum. A dedicated data availability layer like Celestia is architected for one job: cheap, high-throughput data posting. This is a different design goal than Ethereum's execution-focused L1. The specialization of resources is the counter-intuitive key to unlocking scale.
Evidence: Arbitrum processes ~10 TPS, but its data is posted to Ethereum which has a ~80 KB/s data bandwidth limit. To scale to 100k TPS, you need a data layer orders of magnitude larger. This is why EigenDA and Avail are being built as dedicated solutions.
protocol-spotlight
THE SCALING HIERARCHY
Architecting the Endgame: Key DA Players
ZK-Rollups are a crucial scaling step, but they merely push the data bottleneck to the underlying L1. True, sustainable scaling requires a dedicated Data Availability (DA) layer.
01
The Problem: L1s Are Not Databases
Ethereum's 80 KB/s data cap is the ultimate bottleneck. Every ZK-Rollup like zkSync or StarkNet must compete for this scarce resource, making cost reduction asymptotic.
- Scalability Ceiling: ~100 TPS per rollup, limited by L1 blob space.
- Cost Volatility: Transaction fees become a function of L1 gas wars, not ZK-proof efficiency.
- Monolithic Inefficiency: Forces every node to store all data, forever.
80 KB/s
Ethereum DA Cap
~100 TPS
Per-Rollup Limit
02
The Solution: Dedicated Data Sharding
Separate execution from data availability. A specialized DA layer like Celestia or EigenDA provides cheap, abundant blob space, turning L1s into pure settlement layers.
- Exponential Scale: 1,000x+ more bandwidth vs. native L1 execution.
- Cost Predictability: Sub-cent data posting, decoupled from L1 congestion.
- Modular Design: Enables sovereign rollups and high-throughput L2s like Fuel.
1,000x
More Bandwidth
<$0.01
Per Blob Cost
03
Celestia: The Modular Pioneer
First mover defining the DA market. Uses Data Availability Sampling (DAS) and light nodes to securely scale without increasing node hardware requirements.
- Architecture: Optimized solely for ordering and publishing blobs.
- Ecosystem Flywheel: Base layer for rollup-as-a-service platforms like Dymension and AltLayer.
- Security Model: Relies on cryptoeconomic security from its own validator set.
~$0.20
Avg. Blob Cost
100+
Rollups Live
04
EigenDA: The Restaking Power Play
Leverages EigenLayer's restaked ETH to bootstrap security, offering hyperscale DA as an AVS (Actively Validated Service).
- Capital Efficiency: Taps into $15B+ of restaked ETH instead of a new token.
- High Throughput: Targets 10-100 MB/s for data writing.
- Ethereum-Aligned: Native integration for L2s like Mantle and Celo, minimizing trust fragmentation.
$15B+
Restaked Security
10 MB/s
Target Throughput
05
Near DA: The Stateful Challenger
Uses Nightshade sharding to provide DA with fast finality and native execution. Aims to be a unified scalability layer rather than a pure DA module.
- Fast Finality: 2-3 second confirmation vs. ~12 minutes for Ethereum blobs.
- Stateful Data: Can power chain abstraction and intent-based architectures.
- Key Integration: Chosen by Movement Labs and Caldera for high-performance L2s.
2-3s
Data Finality
100k+
TPS Capacity
06
The Endgame: Execution Markets on DA Highways
DA layers become the foundational commodity. Execution environments (rollups, validiums, volitions) compete on top, creating a hyper-competitive, low-fee landscape.
- Specialization Wins: DA focuses on throughput/security, execution on speed/UX.
- Volition Choice: Apps can choose validium (DA off-chain) for cheap micro-transactions or zk-rollup for full security.
- Ultimate Outcome: User transactions are no longer bound by any single chain's limitations.
$0.001
Target Tx Cost
1000+
Specialized Chains
takeaways
THE SCALING ENDGAME
TL;DR for Busy Builders
ZK-Rollups are a crucial intermediate step, but data sharding is the only architecture that scales the base layer itself for global adoption.
01
The Problem: ZK-Rollups Just Kick the Can
ZK-Rollups compress execution but still post data to a congested L1. The data availability (DA) bottleneck remains, capping total network throughput and creating a fee market for data. This is why even optimistic rollups like Arbitrum and Optimism are exploring alternative DA layers like Celestia and EigenDA.
- Bottleneck: L1 data bandwidth is the ultimate constraint.
- Centralization Risk: High L1 fees push sequencing/DA off-chain.
~80 kB
Per Block Limit
$10B+
TVL at Risk
02
The Solution: Ethereum's Danksharding
Transforms the L1 into a pure data availability layer via Proto-Danksharding (EIP-4844) and full Danksharding. This creates dedicated, cheap block space for rollup data blobs, decoupling L1 execution from L2 scalability.
- Scalability: Targets ~1.3 MB/s to >1 MB/s per slot of blob data.
- Cost: Aims for ~100-1000x cheaper data posting for rollups versus calldata.
1 MB/s
Blob Throughput
-99%
Data Cost
03
The Architectural Shift: Execution vs. Settlement vs. DA
Sharding enforces a clean modular stack separation. L1 provides secure settlement and high-throughput DA. Rollups (ZK or Optimistic) handle execution. This mirrors the philosophy behind Celestia and EigenDA, but with the full security of Ethereum consensus.
- Specialization: Each layer optimizes for one function (security, speed, cost).
- Composability: Secure, cheap DA enables massively parallel rollup ecosystems.
3-Layer
Stack
Unlimited
Rollup Slots
04
The Verdict: Why This is Inevitable
Monolithic chains (Solana) and isolated L2s hit fundamental hardware/coordination limits. Data sharding is the only model that preserves decentralization while scaling quadratically. It's the endgame because it solves the core economic problem: making blockchain space a commodity, not a premium good.
- Future-Proof: Bandwidth scales with adoption (more nodes, more shards).
- Ecosystem Synergy: Enables ZK-rollups, Optimistic rollups, and validiums to thrive equally.
Quadratic
Scaling
All
Rollup Types
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