Table of contents
- What Layer-2s are and why 2025 matters
- How rollups work
- Milestones since 2024
- The two big designs: optimistic vs. ZK
- Data availability in 2025: Ethereum blobs and alt-DA
- Sequencers and the push to decentralize ordering
- How to choose an L2 in 2025
- Beyond Ethereum: OP Stack chains and SVM L2s
- FAQs
What Layer-2s are and why 2025 matters
Layer-2 networks scale Ethereum by executing transactions off-chain and periodically committing proofs and data to Ethereum, inheriting its security while delivering much lower fees. The March 2024 “Dencun” upgrade activated EIP-4844 blob space, a dedicated, cheaper lane for rollup data, which materially reduced L2 costs and set the stage for further sharding.
How rollups work
Rollups bundle user transactions and publish state commitments to Ethereum. They either prove correctness with validity proofs (ZK rollups) or accept state optimistically and rely on fraud-proof challenges during a dispute window (optimistic rollups). L2BEAT’s technical overviews remain the canonical reference for these models and their data-availability choices.
Milestones since 2024
- EIP-4844 blobs went live with Dencun, enabling rollups to post data far more cheaply; analyses after launch show dramatic fee drops on L2s and an emerging blob-fee market.
- OP Stack chains enabled permissionless fault proofs and reached “Stage 1” in L2BEAT’s maturity framework on OP Mainnet, with more OP chains upgrading.
- Arbitrum continues to operate with interactive fraud proofs that let anyone challenge incorrect execution, documented in its validation/proving specs.
- Starknet introduced governance safeguards and upgrade delays to reach Stage 1, and began publishing data to blobs post-Dencun.
- Scroll’s 2025 Euclid upgrade migrated its prover toward an OpenVM-based stack, reflecting a broader move to more general-purpose provers.
- Polygon’s 2.0 roadmap emphasizes the AggLayer to aggregate liquidity and UX across chains built with Polygon CDK.
- L2BEAT refined its Stages framework, widely used to track decentralization and user-safety properties across rollups.
The two big designs: optimistic vs. ZK
Optimistic rollups
Optimistic systems assume batches are valid unless challenged. Withdrawals to L1 are finalized after a dispute period, and the security of exits relies on live, permissionless fault/fraud proofs.
- OP Stack chains, including OP Mainnet and many Superchain members, now support permissionless fault proofs for withdrawals.
- Arbitrum documents its multi-round interactive fraud proofs and dispute protocol in detail.
- Base is an OP Stack chain integrated across Coinbase’s ecosystem; it inherits OP Stack architecture and the Superchain roadmap.
ZK rollups
ZK rollups submit cryptographic validity proofs to Ethereum so withdrawals can finalize quickly, without a challenge window. Projects vary by proof system and EVM compatibility.
- zkSync Era describes itself as a ZK rollup with validity proofs and EVM-level developer tooling; research and docs highlight its proving systems.
- Starknet uses the Cairo VM and STARK proofs; Cairo 1.0 modernized the language and security model.
- Linea and Scroll are zkEVM projects pursuing strong EVM equivalence; both maintain active 2025 docs and release notes.
- Vitalik’s zkEVM taxonomy (Type 1–4) is the common reference for understanding different levels of EVM/Ethereum equivalence.
Data availability in 2025: Ethereum blobs and alt-DA
After Dencun, most L2s post data to Ethereum “blobs” to keep trust assumptions minimal and fees low. This is the default for rollups that want Ethereum-level data availability and censorship resistance.
A growing set of chains are experimenting with external DA layers to reduce costs further or to support custom architectures:
- Celestia offers modular DA with data-availability sampling and light-client verifiability for rollups.
- EigenDA, built on EigenLayer restaking, runs on Ethereum and targets high-throughput DA for rollups.
- Avail provides DA APIs and recently announced an encryption-aware “Enigma” upgrade for privacy-sensitive rollups.
L2BEAT and community frameworks compare the risk trade-offs across DA providers and classify systems that do not post full data to Ethereum as validiums/optimiums rather than strict rollups.
Sequencers and the push to decentralize ordering
Most L2s still rely on a single sequencer per chain for ordering, which concentrates power and MEV. Shared-sequencer networks aim to provide decentralized, chain-agnostic ordering to improve liveness, censorship resistance, and cross-rollup composability.
- Espresso and Offchain Labs announced work toward shared sequencing and fair ordering across Ethereum rollups.
- Astria is building a shared-sequencer layer that multiple rollups can plug into, with public docs and devnets.
- Within the OP Stack ecosystem, “Superchain Interop” is in active development to let OP chains read each other’s state with low-latency, trust-minimized messaging.
How to choose an L2 in 2025
Use this quick due-diligence checklist.
- Maturity and exits. Check the L2BEAT Stage (0/1/2) and whether the chain exposes a real exit window before upgrades so users can withdraw safely.
- Proofs and bridge security. Confirm whether withdrawals are protected by permissionless fault proofs (optimistic) or validity proofs (ZK). Prefer canonical L1 bridges and understand bridge risks.
- Data availability. See whether the chain uses Ethereum blobs or an external DA layer (Celestia/EigenDA/Avail) and evaluate the added assumptions.
- Governance and upgrade keys. Review who can upgrade contracts and how quickly; L2BEAT project pages summarize upgrade powers and security councils for major chains like Base.
- Ecosystem and tooling. Consider OP Stack (Superchain), Arbitrum Orbit, Polygon CDK, zk-focused stacks, and available provers/toolchains in 2025 release notes.
Beyond Ethereum: OP Stack chains and SVM L2s
Ethereum’s L2 design space has spilled into the wider ecosystem.
- Celo completed its migration from an L1 to an Ethereum L2 using the OP Stack, aligning with Superchain standards while exploring new DA options.
- Eclipse is an Ethereum L2 that uses the Solana Virtual Machine for parallel execution while settling to Ethereum—an example of “beyond EVM” execution atop Ethereum security.
FAQs
What changed after EIP-4844?
Rollups can post transaction data to temporary “blobs,” a cheaper lane distinct from calldata. This reduces L2 operating costs and enables much lower end-user fees relative to pre-Dencun levels.
What does “Stage 1” mean on L2BEAT?
It indicates a rollup has met decentralization and user-protection requirements such as permissionless proofs and a minimum exit window. Different Stages reflect how much users must trust operators during upgrades.
How do optimistic and ZK rollups differ for withdrawals?
Optimistic rollups finalize withdrawals after a dispute window that allows fraud-proof challenges; ZK rollups finalize once their validity proof is accepted on L1, typically faster.
Are external DA layers safe?
They reduce costs and can scale throughput, but they introduce additional trust or liveness assumptions versus posting all data to Ethereum. Evaluate provider guarantees, light-client verification, and slashing/security models.
What is a shared sequencer and why should I care?
A shared sequencer is a decentralized ordering layer used by many rollups. It aims to reduce censorship risk, mitigate MEV centralization, and unlock smoother cross-rollup UX.
