Ethereum Scalability Is Not a Single Upgrade

The EVM processes transactions sequentially, creating a gas auction for block space. The solution is rollups (L2s) like Arbitrum, Optimism, and zkSync.\n- 10-100x lower fees by batching transactions.\n- ~$30B+ TVL secured by Ethereum's consensus.\n- Enables custom VMs (e.g., Starknet's Cairo) for specialized apps.

Publishing transaction data on-chain (calldata) is the primary L2 cost driver. The solution is data availability layers like Celestia, EigenDA, and Avail.\n- ~$0.001 per MB vs. Ethereum's ~$1,000.\n- Decouples security from execution, enabling sovereign rollups.\n- Critical for validiums and volitions (e.g., StarkEx).

Finalizing cross-rollup asset transfers and proofs takes ~1 week for optimistic rollups. The solution is shared settlement layers like Ethereum L1 (via EIP-4844), Arbitrum Orbit, and Polygon CDK.\n- ~10 minute finality for ZK-rollups.\n- Provides liquidity unification and trust-minimized bridging.\n- Enables interoperability via native cross-rollup messaging.

Rollups are constrained by Ethereum's ~80 KB/s blob data bandwidth. This creates a hard ceiling on total scalable throughput, leading to congestion and volatile fees during peak demand.

• Celestia and EigenDA are competing to become the dominant external DA layer.
• The risk is fragmentation: a rollup's security and liveness guarantees are now tied to its chosen DA provider.

Hundreds of L2s and L3s create isolated liquidity pools. Moving assets between them is slow, expensive, and introduces bridging risks, negating the UX benefits of scaling.

• Solutions like LayerZero, Axelar, and intent-based bridges (Across, UniswapX) are critical infrastructure.
• The derailment occurs if bridging security fails or costs remain prohibitive for micro-transactions.

Most rollups use a single, permissioned sequencer for speed. This creates a central point of failure for censorship and MEV extraction, betraying Ethereum's decentralized ethos.

• Shared sequencer networks like Espresso and Astria aim to decentralize this layer.
• Adoption is slow; the economic incentive to run a solo sequencer is currently too strong for most teams.

Even with rollups, Ethereum's historical state grows ~50 GB/year. Running a full node becomes more expensive, threatening network decentralization and client diversity.

• Verkle Trees and EIP-4444 (history expiry) are multi-year upgrades to address this.
• The roadmap derails if these complex cryptography and incentive changes are delayed or poorly implemented.

Rollup teams control upgrade keys, effectively acting as centralized governors. While many plan to decentralize, the process is slow. Users must trust the team not to introduce malicious upgrades.

• Optimism's Security Council and Arbitrum's DAO are models for progressive decentralization.
• The risk is permanent centralization disguised as temporary, creating a landscape of corporate chains.

A modular stack (Execution + DA + Settlement) introduces complexity and latency. Synchronous composability—the magic of DeFi on L1—is broken. Applications must rebuild for an asynchronous world.

• Protocols like Chainlink CCIP and Polygon AggLayer attempt to restore unified liquidity.
• The derailment is a slower, clunkier ecosystem that fails to abstract complexity from end-users.

Rollups are bottlenecked by posting data to L1. The solution is a competitive DA market.\n- Celestia and EigenDA provide ~$0.001 per MB costs vs. Ethereum's ~$0.10.\n- Enables 10-100x cheaper rollup settlement.\n- Trade-off: Security moves from Ethereum consensus to the DA layer's economic security.

Ethereum abandoned in-protocol execution sharding for a rollup-centric roadmap.\n- Arbitrum, Optimism, and zkSync act as independent execution environments.\n- EIP-4844 (Proto-Danksharding) is the key L1 upgrade, reducing rollup costs by 10-100x.\n- The L1 becomes a secure settlement and data availability layer.

Fragmented liquidity across L2s kills composability. Native bridges are slow and insecure.\n- LayerZero and Axelar provide generalized messaging for cross-chain apps.\n- Across and Chainlink CCIP use intents and oracle networks for optimized security.\n- This adds ~20-60 seconds and $1-$5+ in costs to every cross-chain action.

ZK-proofs offer instant finality and superior security, but face adoption hurdles.\n- zkEVMs (Scroll, zkSync Era) have ~5-10x higher prover costs than Optimistic Rollups.\n- Starknet and Polygon zkEVM require specialized languages (Cairo) or new tooling.\n- The winning stack will unify performant provers with seamless EVM compatibility.

Most rollups use a single sequencer for speed, creating a central point of failure and value extraction.\n- Leads to MEV capture and potential censorship.\n- Shared Sequencer networks (Espresso, Astria) and Based Rollups (using L1 for sequencing) are emerging solutions.\n- This is the next major decentralization battleground after validator sets.

Two competing models for L2 deployment: sovereign chains vs. shared networks.\n- Appchains (dYdX, Eclipse) optimize for a single app with customizability and full MEV capture.\n- Superchains (OP Stack, Arbitrum Orbit) offer shared security, liquidity, and tooling across a franchise.\n- The choice is between maximal sovereignty and network effects.