The trilemma is a model for monolithic blockchains. It posits a single chain must sacrifice decentralization, security, or scalability. This is a constraint of a specific design, not a law of physics. The solution is architectural disaggregation.
Ethereum Scalability From a Systems View
Ethereum's scalability isn't one upgrade; it's a coordinated systems redesign. This analysis breaks down the Merge, Surge, and Verge as interdependent components of a single, complex throughput engine.
Introduction: The Scalability Trilemma Was a Lie
Ethereum's scalability challenge is a problem of system architecture, not a fundamental trade-off.
Scalability is a system property, not a chain property. The goal is a performant network, not a fast chain. This shifts the focus from optimizing a single state machine to orchestrating a system of specialized components like Arbitrum, Optimism, and StarkNet.
The bottleneck is state growth. A monolithic chain forces every node to process and store the entire global state. This creates the trilemma. The solution is to split the state across multiple execution environments, a principle proven by Celestia's data availability layer and EigenLayer's restaking security model.
Evidence: Arbitrum processes over 200,000 TPS for token transfers by batching proofs to Ethereum L1. This demonstrates that system-level throughput is decoupled from base-layer consensus speed. The trilemma applies to the settlement layer, not the execution layer.
Executive Summary: The Three-Layer Stack
Ethereum's scaling strategy is a layered approach, where each layer solves a specific bottleneck in the transaction lifecycle.
The Problem: The Execution Bottleneck
The EVM is a single-threaded CPU, processing transactions sequentially. This creates a hard cap on throughput (~15-30 TPS) and drives up gas fees during congestion.
- Monolithic Blockchains like Ethereum L1 cannot scale execution without sacrificing decentralization.
- High Latency for finality (~12-15 minutes) is unacceptable for many applications.
- Cost Volatility makes user experience unpredictable and dApp economics unstable.
The Solution: Rollups (L2 Execution)
Offload computation. Rollups (Optimistic like Arbitrum, Base; ZK like zkSync, Starknet) execute transactions in batches and post compressed proofs to L1.
- Throughput: Achieves 1000-4000+ TPS by batching.
- Cost: Reduces fees by 10-100x versus L1.
- Security: Inherits Ethereum's consensus and data availability, making them cryptoeconomically secure.
The New Bottleneck: Data Availability
Even compressed, posting all transaction data to L1 is expensive and limits scalability. This is the core constraint for high-throughput rollups.
- Blob Space (EIP-4844) provided temporary relief but is now becoming saturated.
- Data Cost remains the primary driver of L2 transaction fees.
- Throughput Ceiling: L1 bandwidth caps the growth of all rollups built on it.
The Solution: Modular DA Layers
Separate data publication from consensus. EigenDA, Celestia, and Avail provide cheaper, scalable data availability, allowing rollups to post data off-chain.
- Cost Reduction: DA costs drop by another 10-100x versus Ethereum blobs.
- Throughput Unlocked: Enables 10,000+ TPS per rollup.
- Trade-off: Introduces a new trust assumption outside of Ethereum's validator set.
The Final Bottleneck: Cross-Domain State
With execution and DA fragmented across hundreds of chains and rollups, liquidity and composability shatter. Moving assets and state between domains is slow and insecure.
- Bridge Risks: Over $2B+ has been stolen from bridges (Wormhole, Ronin).
- Fragmented UX: Users must manage dozens of wallets and RPC endpoints.
- Broken Composability: dApps cannot seamlessly interact across chains.
The Solution: Unified Settlement & Proving
Re-centralize trust for coordination. Ethereum L1 evolves into a universal settlement and proving layer, while interop hubs like LayerZero, Axelar, and Polygon AggLayer provide cross-chain messaging.
- Shared Security: Rollups can use Ethereum for dispute resolution (Optimistic) or proof verification (ZK).
- Atomic Composability: Protocols like Chainlink CCIP and intents via UniswapX enable cross-domain transactions.
- User Abstraction: Account Abstraction (ERC-4337) and intents hide complexity.
Deconstructing the Roadmap: Merge, Surge, Verge
Ethereum's roadmap is a phased, interdependent systems upgrade that re-architects the network for global scale.
The Merge was a consensus reset. It replaced energy-intensive PoW with PoS, enabling future scalability upgrades by establishing a stable, finalizing base layer for data attestations.
The Surge is a data availability engine. It scales execution via rollups like Arbitrum and Optimism by providing cheap, high-throughput data posting through EIP-4844 proto-danksharding.
The Verge introduces statelessness. It removes the need for nodes to store the entire state, enabling Verkle trees to make solo staking viable at planetary scale.
Evidence: Post-Merge, block finality dropped from ~13 minutes to 12 seconds, a prerequisite for secure cross-rollup bridges like Across and LayerZero.
Systems Trade-Offs: Roadmap Component Analysis
A comparison of core architectural components for scaling Ethereum, analyzing their fundamental trade-offs in decentralization, security, and performance.
| System Property / Metric | Rollups (L2s) | Data Availability Layers | Statelessness & State Expiry |
|---|---|---|---|
Primary Scaling Vector | Execution (off-chain compute) | Data Bandwidth & Storage | State Growth & Node Sync |
Throughput (TPS) Target | 3,000 - 100,000+ | Enables 100,000+ L2 TPS | Not a direct throughput gain |
Trust Assumption | 1+ Honest Sequencer |
| Cryptographic Proofs (Verkle, SNARKs) |
Data Posting Cost to L1 | $0.01 - $0.25 per tx (via calldata or blobs) | $0.001 - $0.01 per tx (optimized) | N/A |
Node Hardware Requirement | Light (sequencer) to Full (prover) | Light (sampling) to Full (DA consensus) | Constant (Verkle proofs ~50KB) |
Time to Finality on L1 | 12 min - 1 hr (optimistic) / ~20 min (ZK) | ~12 min (Ethereum blobs) / ~1 min (external DA) | N/A |
Solves State Bloat | Partially (state localized to L2) | Yes (historical data pruning) | Yes (active state management) |
Key Ecosystem Examples | Arbitrum, Optimism, zkSync, Starknet | Ethereum Danksharding, Celestia, EigenDA, Avail | Verkle Trees, EIP-4444, Portal Network |
The Monolithic Counter-Argument: Are We Just Building a Worse Solana?
The modular stack's complexity risks re-creating a slower, more expensive version of a monolithic chain.
Modularity introduces systemic latency. The data availability layer, execution environment, and settlement layer operate as separate networks. Each hop adds finality delay, making cross-domain composability slower than Solana's single-state machine.
The cost of security is fragmentation. Users pay for L1 settlement gas and sequencer fees, creating a multi-fee market problem. This often exceeds the cost of a single transaction on a high-throughput monolithic chain like Sui or Aptos.
Developer experience regresses. Building a cross-rollup application requires managing bridges like Across and Stargate, which adds failure points. Solana's single global state eliminates this complexity entirely for developers.
Evidence: A simple asset swap involving Arbitrum, Base, and Ethereum mainnet can take 20+ minutes and cost over $10 in fees. The same action on Solana finalizes in seconds for fractions of a cent.
Architectural Takeaways for Builders
Scaling Ethereum isn't just about TPS; it's about re-architecting the trust, data, and execution stack.
The Data Availability Bottleneck is the New Security Frontier
Execution is cheap; consensus on data is expensive. The core scaling constraint has shifted from block gas limits to data availability (DA).
- Key Benefit 1: Off-chain DA layers like EigenDA and Celestia reduce L2 posting costs by ~90% vs. calldata.
- Key Benefit 2: Enables high-throughput, sovereign rollups (e.g., Fuel, dYmension) to scale independently of Ethereum's data capacity.
Modularity Wins, But Introduces New Trust Assumptions
Monolithic chains (Solana) optimize for synchronous performance; modular stacks (Ethereum + L2s + DA) optimize for asynchronous, verifiable trust.
- Key Benefit 1: Specialization allows Arbitrum for gaming, zkSync for payments, Base for social—each with optimized VMs.
- Key Benefit 2: Creates a multi-billion dollar marketplace for sequencers, provers, and DA providers, but forces builders to audit new trust graphs.
Intent-Centric Design is the Next UX Paradigm
Users don't want to sign 10 transactions across 5 chains. They want outcomes. Abstracting execution to a solver network is the logical endpoint of modular scaling.
- Key Benefit 1: Protocols like UniswapX and CowSwap already route orders across L2s and sidechains for optimal price, hiding complexity.
- Key Benefit 2: Drives adoption by making cross-chain interactions feel like a single-chain experience, with Across and LayerZero as critical infrastructure.
Shared Sequencers are the Next L2 Battleground
Today, each rollup runs its own sequencer, creating fragmentation and MEV leakage. Shared sequencer networks (Espresso, Astria) will commoditize block production.
- Key Benefit 1: Enables atomic cross-rollup composability, unlocking complex DeFi strategies across Optimism's Superchain or Arbitrum Orbit.
- Key Benefit 2: Reduces centralization risk and creates a liquid market for block space, pushing costs toward marginal execution.
Verification is the Only Non-Negotiable On-Chain Load
Everything can be off-chain (execution, data, settlement) except the cryptographic proof of correctness. Zero-knowledge proofs (ZKPs) are the ultimate compression tool.
- Key Benefit 1: ZK-rollups (zkSync, Starknet, Scroll) provide Ethereum-level security with ~10-100x cheaper finality vs. optimistic rollups.
- Key Benefit 2: Enables privacy-preserving scaling (Aztec) and verifiable off-chain compute, making the base layer a pure court of appeal.
The Endgame is a Unified Liquidity Layer, Not a Unified Chain
The future isn't one chain to rule them all. It's a constellation of specialized execution environments with seamless, trust-minimized asset movement.
- Key Benefit 1: Protocols like Chainlink CCIP and Wormhole abstract bridging into a primitive, letting dApps deploy on the optimal chain for their use case.
- Key Benefit 2: Ethereum L1 becomes the anchor of trust and liquidity, settling disputes and hosting canonical asset registries, while activity explodes on L2s and L3s.
Get In Touch
today.
Our experts will offer a free quote and a 30min call to discuss your project.