What Are Layer 2s (L2s)? Top L2 Networks Compared

Why Do We Need Layer 2s?

Ethereum's base layer can process approximately 15 transactions per second (TPS). For context, Visa handles around 1,700 TPS on average. As Ethereum gained adoption through DeFi, NFTs, and stablecoins, network congestion drove gas fees as high as $50–$100 per transaction during peak periods in 2021–2023.

This scalability bottleneck is described by the blockchain trilemma, a concept coined by Ethereum co-founder Vitalik Buterin: a blockchain can optimize for two of three properties (decentralization, security, and scalability) but not all three simultaneously. Layer 2s address this by moving execution off-chain while relying on Ethereum for security and data availability.

The impact has been dramatic. Following Ethereum's Dencun upgrade in March 2024 (which introduced EIP-4844 and “blob” data), L2 transaction fees dropped by over 90%. Average swap costs fell from roughly $86 on mainnet to under $0.39 on L2s. By 2026, Layer 2s handle the majority of Ethereum's transaction volume, and for most users, L2 transactions are effectively free.

While Ethereum hosts the vast majority of L2 activity and TVL, Layer 2 scaling is not exclusive to Ethereum. Bitcoin has its own growing L2 ecosystem, including the Lightning Network for payments and newer projects like Stacks for smart contract functionality. For a detailed look at that side of the landscape, see our guide: Top Bitcoin Layer 2s.

This article focuses primarily on Ethereum L2s, where the rollup architecture has driven the most innovation and adoption.

What Is the Difference Between Optimistic and ZK Rollups?

Nearly every production L2 in 2026 is a rollup: a system that batches hundreds or thousands of transactions together and posts a compressed summary back to Ethereum. The two dominant rollup architectures differ in how they prove that transactions are valid.

Optimistic Rollups

Optimistic rollups, such as Arbitrum, Optimism, and Base, assume every batch of transactions is valid by default. They use a challenge window (typically 7 days) during which anyone can submit a fraud proof to dispute an invalid batch. If no challenge is raised, the batch is finalized on Ethereum.

This “optimistic” approach is simpler to implement and offers full EVM compatibility, which is why existing Solidity smart contracts can be deployed on optimistic rollups with minimal changes. As of 2026, optimistic rollups hold approximately 80% of L2 market share by total value locked (TVL) due to their mature developer tooling, deep DeFi liquidity, and first-mover advantage.

ZK (Zero-Knowledge) Rollups

ZK rollups, such as Starknet and zkSync Era, take the opposite approach. They generate a cryptographic validity proof for every batch, which Ethereum can verify in milliseconds. This means transactions are final as soon as the proof is accepted, with no 7-day wait.

ZK rollups offer faster finality and stronger long-term security assumptions. Their historical weakness, EVM incompatibility and expensive proof generation, has largely closed by 2026 as zkEVM technology matures and prover hardware improves. However, ZK rollups still face adoption challenges: Starknet requires developers to learn Cairo (its custom language), and most ZK rollups still rely on centralized provers.

For a deeper dive into ZK technology, see our guide: What Are Zero-Knowledge Proofs and ZK-Rollups?

While ZK rollups offer faster finality and stronger cryptographic guarantees, 2026 data shows that optimistic rollups still hold roughly 80% of L2 market share by TVL. The primary drivers are mature developer tooling, deep existing liquidity on Arbitrum and Base, and strong network effects that make switching costs high.

ZK rollups are gaining ground in institutional use cases; zkSync's Prividium layer with Deutsche Bank and UBS is a notable example. But for general DeFi and consumer applications, optimistic rollups remain dominant.

Are Sidechains the Same as Layer 2s?

Before rollups became dominant, several other Layer 2 architectures were proposed and deployed. While these approaches are less prominent in 2026, understanding them provides useful context for how the scaling landscape evolved.

State Channels

State channels allow two or more parties to transact directly off-chain by locking funds into a smart contract on the base layer. Participants can exchange an unlimited number of transactions between themselves without touching the main chain, and only the opening and closing states are settled on Ethereum. The Bitcoin Lightning Network is the best-known example of state channels in production. State channels offer near-instant finality and very low fees, but they are limited to interactions between pre-defined participants and are not suitable for general-purpose smart contract execution.

Sidechains

Sidechains are independent blockchains linked to a parent chain through a two-way bridge. They run their own consensus mechanism and validator set, which means they do not inherit the security of the base layer, making them technically distinct from true Layer 2s. The original Polygon PoS chain is the most well-known example. In 2026, the industry draws a clear line between sidechains and rollups: sidechains trade security guarantees for independence, while rollups maintain Ethereum's security model. Polygon itself has acknowledged this distinction by building Polygon zkEVM, a true ZK rollup, alongside its PoS sidechain.

Plasma

Plasma, proposed by Vitalik Buterin and Joseph Poon in 2017, creates a hierarchy of child chains anchored to Ethereum. Each child chain processes transactions independently and submits periodic commitments to the root chain, using fraud proofs for dispute resolution. Plasma was an important early scaling concept, but its complexity and limitations around general-purpose computation led the industry to converge on rollups as the preferred architecture. Most Plasma-based projects have since migrated to or been superseded by rollup designs.

Validiums

Validiums are a hybrid approach: like ZK rollups, they use validity proofs to verify transactions, but unlike rollups, they keep transaction data off-chain on a separate data availability layer rather than posting it to Ethereum. This reduces costs further but introduces additional trust assumptions, since users must trust the external data availability provider. Immutable zkEVM (used for gaming) operates in validium mode, and Mantle uses a hybrid approach with EigenDA for data availability. L2Beat classifies validiums as a distinct category from rollups for this reason.

Top Ethereum Layer 2s Compared (May 2026)

The table below compares the five largest Ethereum L2 networks by Total Value Locked (TVL), rollup type, average fees, and primary use cases. Data is sourced from CoinGecko and L2Fees.

Sources: CoinGecko (TVL), L2Fees (fees), The Block 2026 L2 Outlook (use cases). Data as of May 2026.

How to Choose a Layer 2

With over 50 rollups live on Ethereum, choosing the right L2 depends on your specific use case. Three questions can guide the decision:

The L2 Token Paradox

Compare current L2 token prices on the CoinGecko Layer 2 category page.

The L2 Landscape in 2026: Consolidation and Specialization

The Layer 2 ecosystem in 2026 looks very different from the “rollup explosion” of 2024–2025, when dozens of new L2s launched with nearly identical tech stacks and competed primarily on points programs and airdrop promises. In February 2026, Vitalik Buterin himself acknowledged that the “original vision of L2s and their role no longer makes sense”. That statement reflected an industry-wide reckoning. As incentive cycles ended, most of these chains saw TVL collapse 70–90% within weeks. The result is a sharp power-law distribution where a small number of networks dominate, and L2s are increasingly differentiating by use case rather than competing as generic scaling layers. On L2Beat's decentralization framework, Arbitrum, Base, and Optimism have all reached Stage 1 (functional proof systems live), while most ZK rollups remain at Stage 0.

The Big Three: Arbitrum, Base, and Optimism

Arbitrum remains a major L2 player by TVL, with approximately $1.7B in TVL. Its deep DeFi ecosystem, including Uniswap, Aave, and GMX, creates strong network effects. The Arbitrum DAO holds over $150 million in non-native assets, and the Foundation reports margins above 90%. Its Stylus upgrade adds a WASM/Rust execution path alongside Solidity, attracting developers from outside the EVM ecosystem. Track Arbitrum (ARB) price and market data on CoinGecko.

Base, built by Coinbase, is the fastest-growing consumer-facing L2. It leads all Layer 2s in daily active users (surpassing 1 million daily active addresses) and transaction count, fueled by Coinbase's 110+ million verified users and seamless fiat on-ramps. Base does not have its own token; revenue flows to Coinbase and contributes to the OP Stack ecosystem. In early 2026, Base announced a move to its own unified codebase, departing from the shared Optimism OP Stack. Browse all chains on CoinGecko for current Base metrics.

Optimism holds approximately $365.5M in TVL, but its influence extends far beyond its own chain. The OP Stack, Optimism's open-source rollup framework, powers Base, Mantle, Zora, Mode, and dozens of other “Superchain” networks. However, OP Mainnet experienced a significant TVL decline in early 2026 following shifts in its relationship with Base. Optimism is pivoting from competing for retail users to positioning itself as infrastructure: a shared sequencing layer and Retroactive Public Goods Funding (RPGF) mechanism for the broader ecosystem. Track Optimism (OP) price and market data on CoinGecko.

ZK Rollups: The Specialization Play

Starknet is the leading ZK rollup by TVL at roughly $210 million. It uses STARK proofs (which do not require a trusted setup, unlike SNARKs) and its custom Cairo programming language. In 2026, Starknet has pivoted toward BTCFi with its shielded strkBTC wrapper and introduced native on-chain privacy via the SNIP-36 upgrade (“Shinobi”). Its technology is arguably the most advanced in the L2 space, but the Cairo learning curve has slowed developer adoption compared to EVM-compatible alternatives. Track Starknet (STRK) price and market data on CoinGecko.

zkSync Era took one of the boldest pivots of any major L2 in 2026, shifting its primary focus from retail DeFi to institutional finance through Prividium, a privacy-preserving, permissioned enterprise layer. Deutsche Bank and UBS are among the first partners, exploring tokenized asset settlement and cross-border payments. Track zkSync (ZK) price and market data on CoinGecko.

Emerging L2s to Watch

MegaETH is one of the most-watched emerging L2 projects heading into mid-2026. It targets parallel, high-throughput execution on an Ethereum-aligned L2, where independent transactions can be processed concurrently rather than through a single global queue, aiming for 100,000+ TPS. MegaETH is still pre-production and does not yet have meaningful TVL or adoption data, but its architecture represents a potential next step in L2 design for high-frequency trading and latency-sensitive applications.

Key Layer 2 Terms

Rollup: A scaling solution that bundles (or “rolls up”) hundreds of transactions into a single batch and posts compressed data to Ethereum for settlement.

Fraud proof: The mechanism used by optimistic rollups. If a batch is suspected to be invalid, any network participant can submit a fraud proof during the challenge window to dispute it.

Validity proof: The mechanism used by ZK rollups. A cryptographic proof that mathematically demonstrates every transaction in a batch is valid, verified on-chain by Ethereum.

Blob (EIP-4844): A new data type introduced by Ethereum's Dencun upgrade in March 2024 that allows rollups to post data more cheaply. Blobs are stored temporarily (roughly 18 days) rather than permanently, dramatically reducing L2 operating costs.

Superchain: Optimism's vision for a network of interoperable L2 chains built on the OP Stack, sharing security, communication protocols, and governance. Base, Mantle, Zora, and Mode are current Superchain members.

Layer 3 (L3): Hyper-specialized chains built on top of Layer 2s. An L3 settles its transactions to an L2 (which in turn settles to Ethereum), enabling application-specific execution environments for use cases like gaming, high-frequency trading, or privacy. Arbitrum Orbit and zkSync Hyperchains are frameworks for deploying L3s.

Conclusion

Layer 2s have evolved from an experimental scaling concept into the primary execution layer for Ethereum. In 2026, the vast majority of Ethereum transactions happen on L2s, fees have dropped to near-zero for most users, and the ecosystem has consolidated around a handful of dominant networks. Base leads in consumer adoption and DeFi liquidity, while Arbitrum remains a market leader, and Optimism's OP Stack underpins much of the infrastructure. ZK rollups like Starknet and zkSync Era are carving out specialized niches in privacy and institutional finance.

For users, the practical takeaway is straightforward: L2s are where the activity is, and the cost of participating in DeFi, NFTs, and on-chain applications is a fraction of what it was on Ethereum mainnet. For developers and projects, the question is no longer whether to deploy on an L2, but which L2 best fits their liquidity needs, finality requirements, and target audience.

Track the latest L2 token prices, market caps, and rankings on the CoinGecko Layer 2 category page.