Full Danksharding's advertised throughput is a theoretical maximum, not a sustainable operational target. The system's capacity is gated by the data bandwidth of the consensus layer and the cost of data availability sampling for nodes.
the-ethereum-roadmap-merge-surge-verge
Blog
Full Danksharding Is Not Infinite Throughput
A technical deconstruction of the fundamental limits of Ethereum's scaling roadmap. Full Danksharding increases data availability by ~100x, but hardware constraints, economic incentives, and rollup design create a hard ceiling far below 'infinite'.
introduction
THE BOTTLENECK
The Infinite Scaling Mirage
Full Danksharding's theoretical throughput is bounded by physical and economic constraints, not by protocol design.
The final bottleneck is physical hardware. The network's speed is limited by the slowest participating validator's internet connection and storage I/O. This creates a practical ceiling far below the petabyte-scale theory.
Economic security imposes a hard cap. Increasing blob count dilutes the cost of data inclusion, which can degrade the cryptoeconomic security of the data availability layer. Protocols like Celestia and EigenDA compete on this exact trade-off frontier.
Evidence: Ethereum's current roadmap targets ~1.3 MB/s of data availability. This is a 1000x increase from today, but it is a finite, engineered constant, not an open-ended scaling solution.
key-trends
THE PHYSICAL BOTTLENECKS
The Three Hard Limits of Full Danksharding
Full Danksharding scales data availability, not execution. These are the fundamental constraints that remain.
01
The Problem: The P2P Network Choke
Full Danksharding's 64 data blobs (~3.2 MB each) must be propagated across the global peer-to-peer network. This is the new bottleneck.
- Bandwidth: Nodes need ~2 Gbps to keep up with a fully saturated shard.
- Latency: Global gossip imposes a ~1-2 second floor on block propagation time.
- Decentralization Tax: Higher requirements push out home validators, centralizing the network layer.
~2 Gbps
Peak Bandwidth
~1-2s
Propagation Floor
02
The Problem: The Sequencer Execution Wall
Even with infinite data, a single sequencer (e.g., the current Ethereum L1) can only process so many transactions per second.
- CPU Bound: EVM execution is single-threaded, creating a ~100-200 TPS hard cap for complex transactions.
- State Growth: Every transaction touches global state, causing exponential growth in storage and proving costs.
- The Real Solution: This is why scaling requires parallel EVMs (Monad, Sei) and L2 rollups (Arbitrum, Optimism, zkSync) for execution.
~100-200
Max Complex TPS
Exponential
State Growth
03
The Problem: The Economic Saturation Point
Throughput is not free. At full utilization, the cost to use the chain is dictated by pure supply and demand economics.
- Fee Markets: With 64 full blobs, demand still sets the base fee. "Cheap" is relative to demand.
- Validator Incentives: If blob fees are too low, validators may ignore blobs, breaking data availability guarantees.
- The Equilibrium: The system finds a price where marginal cost of resource use = marginal revenue from fees. Infinite, free blockspace is impossible.
64 Blobs
Max Supply
Demand-Led
Pricing
deep-dive
THE REALITY CHECK
Deconstructing the Bottlenecks: From Blobs to Finality
Full Danksharding solves data availability, but finality and state growth remain fundamental constraints on Ethereum's throughput.
Blobs are not bandwidth. EIP-4844's proto-danksharding provides cheap data for L2s like Arbitrum and Optimism, but the network's consensus layer still processes and attests to every blob. The 16 MB per slot target for full Danksharding is a data availability limit, not a transaction processing guarantee.
Finality is the ultimate bottleneck. Even with infinite blobs, Ethereum's 12-second finality window is a hard protocol constant. High-frequency applications requiring sub-second finality, like those built on dYdX's Cosmos app-chain, will never run directly on L1.
State growth is the silent killer. Blobs expire, but L1 state is permanent. High throughput from rollups like zkSync and Starknet forces constant state expansion, increasing node hardware requirements and centralizing the validator set over time.
Evidence: The current blob count is capped at 6 per block. Even at the 16 MB target, this translates to ~1.3 MB/s of raw data, not the 'infinite' scaling often misrepresented.
FULL DANKS-HARDING
The Scaling Stack: Bottleneck Analysis
Comparing the fundamental throughput bottlenecks of Ethereum's scaling roadmap, highlighting that data availability is not the only constraint.
| Bottleneck | Current Rollup (Base Case) | Proto-Danksharding (EIP-4844) | Full Danksharding (Post-4844) |
|---|---|---|---|
Data Availability (DA) Throughput | ~80 KB/block (Calldata) | ~1.3 MB/block (Blobs) | ~1.3 MB/block (Blobs) |
State Growth Rate | ~50 GB/year | ~50 GB/year | ~50 GB/year |
State Witness Size (Per Block) | ~1-10 MB | ~1-10 MB | ~1-10 MB |
Execution Layer Compute (Gas) | 30M gas/block | 30M gas/block | 30M gas/block |
Settlement Throughput (Proof Verification) | ~300-500 TPS (ZK) / ~100 TPS (OP) | ~300-500 TPS (ZK) / ~100 TPS (OP) | ~300-500 TPS (ZK) / ~100 TPS (OP) |
Cross-Rollup Messaging Latency | 12-20 min (L1 Finality) | 12-20 min (L1 Finality) | 12-20 min (L1 Finality) |
Primary Constraint Post-Upgrade | Expensive DA (Calldata) | State Growth & Execution | State Growth & Execution |
counter-argument
THE SCALING CEILING
Steelman: "But It's Enough for Global Scale"
Full Danksharding's theoretical throughput is immense but fundamentally capped, creating a predictable economic and architectural ceiling.
Full Danksharding is not infinite. Its design caps data availability at ~1.3 MB per slot per blob, scaling to ~128 blobs. This creates a hard, predictable throughput ceiling of ~1.3 TB/day. This is a feature, not a bug, establishing a known scaling limit for infrastructure planning.
This ceiling defines the market. A finite DA capacity creates a fee market for blobspace, similar to Ethereum's block space. Protocols like EigenDA and Celestia compete within this market, but the total supply is bounded by Ethereum's consensus.
Global scale requires off-chain execution. The ~1.3 TB/day DA layer is a data backbone for high-throughput L2s like Arbitrum and Optimism. It supports millions of TPS in execution, but only by pushing computation off-chain and settling proofs on-chain.
Evidence: Ethereum's current maximum is ~0.75 MB per slot. Full Danksharding's 128-blob target represents a ~170x increase, but this final multiplier is fixed by the protocol constants.
takeaways
THE REALITY CHECK
TL;DR for Builders and Investors
Full Danksharding is a massive scaling leap, but it's not a magic bullet for infinite, free transactions.
01
The Bottleneck Shifts to Consensus
Full Danksharding scales data availability (DA) to ~128 MB per slot, but the consensus layer (Beacon Chain) must still process and attest to this data. This creates a new, softer bottleneck.
- Throughput is gated by validator bandwidth and voting latency.
- The system is designed for ~1.33 MB/s of persistent data, not infinite blobs.
- Builders must design for realistic finality windows, not theoretical peak throughput.
~128 MB
Per Slot DA
~1.33 MB/s
Sustained Rate
02
Data is Cheap, Execution is Not
While blob storage costs plummet, executing transactions (EVM ops) on Layer 2 rollups like Arbitrum, Optimism, and zkSync remains the dominant cost.
- Blob fee markets will emerge, creating variable costs for high-demand blocks.
- L2 economics shift from paying for DA on L1 to optimizing execution and proving costs.
- Investors should evaluate L2s on proof efficiency (Validity vs. Fraud) and sequencer design.
>100x
Cheaper DA
Dominant Cost
Execution
03
The L2 Aggregation War
With abundant DA, the competitive edge for rollups moves to proving cost, interoperability, and user experience. This fuels projects like EigenDA, Celestia, and Near DA competing on cost, while Polygon, StarkWare, and zkSync compete on proof systems.
- Shared sequencers (like Espresso, Astria) will become critical infrastructure.
- Interoperability stacks (LayerZero, Chainlink CCIP, Wormhole) are essential for cross-L2 liquidity.
- Build: Focus on vertical integration (app-specific L3) or horizontal aggregation (shared sequencer).
Multi-DA
Ecosystem
Key Battleground
Shared Sequencers
04
The Verkle Proof Challenge
Full Danksharding requires Verkle Trees for statelessness, allowing validators to verify blocks without storing full state. This is a massive, complex upgrade.
- State expiry may be necessary, complicating contract design and UX.
- Builders must prepare for new RPC patterns and witness data handling.
- This is the final, critical dependency before maximal scaling is realized.
Critical Path
Dependency
~2025+
Timeline
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