Calldata was a subsidy. Early L2s like Optimism and Arbitrum posted transaction data to Ethereum as cheap calldata, not expensive storage. This bypassed execution costs but still burdened Ethereum's consensus layer with raw data.
the-ethereum-roadmap-merge-surge-verge
Blog
Why Calldata Was a Temporary Scale Hack
Calldata was a brilliant, cheap scaling hack for rollups, but it was always a dead-end. The Surge's EIP-4844 blobs and dedicated DA layers like Celestia mark its end. This is the evolution from a clever exploit to sustainable infrastructure.
introduction
THE CALLDATA ERA
The Clever, Unsustainable Hack
Ethereum's scaling strategy relied on a data storage loophole that was always destined to hit a wall.
The scaling was illusory. This created a data availability bottleneck. Every L2's growth directly increased the load on Ethereum's base layer, making the system's scalability contingent on the very chain it was trying to escape.
The subsidy ended. EIP-4844 (proto-danksharding) introduced blobs as a dedicated, cheaper data channel, explicitly acknowledging that calldata scaling was a temporary hack. The era of cheap L1 data posting is over.
Evidence: Before blobs, calldata consumed over 90% of an L2's operating cost. Post-EIP-4844, data costs for chains like Base and zkSync dropped by over 90%, proving the prior model's economic fragility.
key-trends
WHY IT WAS A TEMPORARY HACK
The Three Forces Killing Calldata
Calldata scaling was a clever but unsustainable workaround, now being dismantled by fundamental economic and architectural shifts.
01
The Economic Ceiling: Blob Pricing Volatility
EIP-4844 blobs introduced a separate fee market, but their price is still subject to L1 congestion. This creates unpredictable costs for L2s and their users, breaking the promise of stable, low fees.
- Cost spikes during L1 mempool floods can still cascade to L2s.
- Creates business model risk for rollups dependent on predictable overhead.
- The solution is not cheaper blobs, but removing L1 data dependency entirely.
100x+
Fee Swings
~10 min
Blob Window
02
The Architectural Limit: Data Availability Sampling (DAS)
Full nodes downloading all blob data is the next bottleneck. True scaling requires light nodes to securely verify data availability with minimal resources.
- Celestia, EigenDA, and Avail are building dedicated DA layers with DAS.
- Enables exponential scalability by separating consensus from data.
- Forces a reckoning: pay for expensive Ethereum DA, or opt for a modular stack.
MB/s
DA Throughput
~100KB
Node Download
03
The Endgame: Volitions & Validiums
The final force is the realization that not all transactions need Ethereum-level security. Apps will split state between a secure 'slow lane' and a scalable 'fast lane'.
- Validiums (like StarkEx) use off-chain DA for ~10,000 TPS.
- Volitions let users choose their security model per transaction.
- This fragments the monolithic 'calldata as DA' model, making it a legacy option.
>10k TPS
Validium Throughput
-99%
vs. Rollup Cost
deep-dive
THE CALDATA HACK
From Subsidy to Specialization: The Data Availability Revolution
Ethereum's reliance on calldata for scaling was a temporary subsidy, not a sustainable architecture, forcing a shift to specialized data availability layers.
Calldata was a subsidy. Early L2s like Optimism and Arbitrum One posted transaction data to Ethereum's calldata because it was the cheapest, most secure storage. This created a temporary scaling illusion by leveraging Ethereum's security without paying for its full execution cost.
The subsidy expired. As L2 activity grew, calldata bloated Ethereum blocks, making it expensive for everyone. The EIP-4844 proto-danksharding upgrade replaced calldata with blobs, a dedicated data lane priced separately from gas, ending the cross-subsidy.
Specialization is mandatory now. Post-EIP-4844, L2s must choose a data availability (DA) strategy. They either pay for Ethereum blobs, use a modular DA layer like Celestia or EigenDA for lower cost, or adopt a validium model (e.g., some StarkEx apps) with off-chain DA and on-chain proofs.
Evidence: Before EIP-4844, L2s consumed over 10% of Ethereum's block space. Post-upgrade, blob fees are 90% cheaper than equivalent calldata, proving the efficiency of dedicated data channels and validating the modular thesis.
DATA AVAILABILITY LAYER COMPARISON
Cost & Scale: Calldata vs. Blobs vs. External DA
A technical breakdown of Ethereum's primary data availability solutions, comparing the legacy calldata hack, native blobs, and external DA providers.
| Feature / Metric | Calldata (Legacy) | EIP-4844 Blobs | External DA (e.g., Celestia, EigenDA) |
|---|---|---|---|
Cost per Byte (Est.) | ~68 gas/byte | ~0.125 gas/byte | ~0.01 gas/byte |
Data Persistence | Forever (on-chain) | ~18 days (pruned) | Variable (off-chain) |
Throughput (MB/block) | ~0.09 MB | ~0.38 MB |
|
Settlement Finality | Native Ethereum | Native Ethereum | Bridged / Delayed |
Security Assumption | Ethereum L1 | Ethereum L1 | Separate Consensus |
Adoption Driver | Pre-Dencun L2s | Post-Dencun L2s (Arbitrum, Optimism) | High-throughput L2s (Manta, Kinto) |
Primary Risk | Prohibitive cost at scale | Blob gas market volatility | Data withholding & bridge risk |
future-outlook
THE DATA
The Post-Calldata Stack: Implications for Builders
Calldata's role as a cheap data layer is ending, forcing a fundamental redesign of rollup economics and execution.
Calldata was a temporary hack for subsidizing L2 data availability. It exploited Ethereum's historical under-pricing of data storage, but EIP-4844 and full danksharding correct this. The era of near-zero-cost data posting is over.
The new cost basis reshapes rollup design. Projects like Arbitrum and Optimism must now optimize for blob space, not just gas. This shifts competition from pure execution efficiency to data compression and state management.
App-specific rollups gain a structural advantage. A monolithic chain like Base pays for all app data. A rollup serving only a DEX like Uniswap or a game compresses a homogeneous data stream more efficiently, lowering its fixed cost.
Evidence: Blob pricing volatility. The first 30 days of EIP-4844 saw blob costs fluctuate by 1000%, proving data is now a market-driven commodity. Rollups without a cost-optimized data strategy face existential margin pressure.
takeaways
WHY CALLDATA WAS A TEMPORARY SCALE HACK
TL;DR for CTOs & Architects
Ethereum's reliance on calldata for L2 data availability was a pragmatic but unsustainable scaling solution, creating a cost cliff that new architectures must solve.
01
The Problem: Exponential Cost vs. Linear Scale
Calldata is stored permanently on Ethereum, paying for eternal storage for temporary data. L2 transaction growth creates a quadratic fee burden: more users → more blobs → higher basefee for all blobs. This is not a viable scaling model.
- Cost Structure: Paying ~16 gas/byte forever for data needed for ~2 weeks.
- Scalability Ceiling: Pre-EIP-4844, ~100 TPS was the practical limit before fees became prohibitive.
~100 TPS
Pre-4844 Limit
16 gas/byte
Eternal Cost
02
The Solution: Proto-Danksharding (EIP-4844)
Introduces blob-carrying transactions with data that auto-expires after ~18 days. This separates L2 data market from Ethereum's execution market, enabling ~100x more data capacity at ~100x lower cost.
- Key Innovation: Blobs are cheap, ephemeral, and validated by consensus clients.
- Architectural Shift: Moves DA from execution layer (expensive, permanent) to a dedicated bandwidth layer (cheap, temporary).
~100x
Cheaper DA
18 days
Data Lifetime
03
The New Bottleneck: Blob Supply & L2 Design
EIP-4844 creates a commoditized DA market. The new constraint is blob slot supply per block (~6). L2s must now compete on blob efficiency (data compression, validity proofs) and sequencer design to batch effectively.
- Competitive Axis: Cost per user op is now driven by bytes posted to Ethereum, not gas price alone.
- Design Imperative: Architectures like zkRollups (STARKs, SNARKs) and optimized Optimistic Rollups (like Arbitrum, Optimism) have a direct cost advantage.
~6/blk
Blob Target
zk > ORU
DA Efficiency
04
The Endgame: Modular DA & EigenLayer
Blobs are still on Ethereum. The final scaling phase is modular data availability layers like Celestia, EigenDA, and Avail. These offload DA entirely, letting Ethereum act as a settlement and security hub. EigenLayer's restaking provides cryptoeconomic security for these external DA layers.
- Architectural Choice: L2s can now select DA based on cost/security trade-offs.
- Settlement Layer Focus: Ethereum's value shifts to enforcement of state transitions via proofs, not data storage.
~$1B+
EigenDA TVL
Modular Stack
New Paradigm
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